The AM Forum

I've decided to build up some audio filters for transmitting that will dynamically manage bandwidth for several of my homebrew and analog AM transmitters.  (non-SDR) They are low level low pass filters placed in the audio chain, so can be common to all rigs. They basically roll off the extreme highs at a preset point.  Since many transmitters have "wider than normally used audio"  to generate a clean signal with headroom, we need some kind of control to keep our bandwidth in line with band activity and crowding.

I've selected 10th order Butterworth designs with a 600 ohm input and 600 ohm output. (The impedances can be changed to anything else with the software below.)

Being 10th order, the roll offs are quite steep. I want a filter for three kinds of band conditions on AM. For example, a 4KHz filter starts to roll off at 4 KHz of spoken audio.  The total bandwidth is 8KHz because of two sidebands on AM. Here are some roll off examples of the filter designs attached below:

1) 4 KHz  filter =  down -20 DB at 5KHz  (somewhat tighter than AM broadcast bandwidth)

2) 5.5 KHz filter =  down -20 DB at 7.1 KHz

3) 7.5 KHz filter =  down -20 DB at 9.5 KHz


These filters will be built on one board and switched in and out in an instant.

The reasons for three filters:

1) When there is 5 KHz spacing between stations, the band is very busy and the 4KHz filter can be selected. At 5KHz the audio is rolled off -20 DB for the neighboring station's benefit.  (and sharper rolloff at 5.5 KHz, etc.)

2) When the station spacing is 7-8 KHz (common) then the 5.5 Khz filter is selected.

3) When the band is quiet and stations are spread out with 10 KHz spacing, then the 7.5 KHz filter can be run.

*** For these filters to do their job, the transmitter and amplifiers must be clean - low IMD and a flat response is important. ***

The Butterworth filters appear to have the least amount of ripple and are passive filters.

I plan to order the parts and give them a try within a week.  Until there is a DSP audio filter designed for this kind of job, the Butterworth using inductors and capacitors may suit the bill.


The parts are very cheap and small and standard values.  At Mouser Electronics, inductors the size of a pea will work FB cuz there is really little current in the circuit.

Here is the web calculator in case you want to run your own filter calculations:
https://rf-tools.com/lc-filter/


Comments?

Tom, K1JJ

Thanks to Frank / GFZ for his guidance...

Actually, an all-in-one processor is in its final design stages. The MAX processor incorporates several stages of envelope processing primarily geared to the AM market:

(http://www.internetwork.com/MAX/images/20190616_211757a.jpg)

The optional stages are: Low Cut, All-Pass Filter, Pre-Emphasis, Compression, Limiting, SCAF (Switched Capacitor Audio Filter) for 3, 5, and 7kHz or user selectable), Peak Clipper, and Low Pass Filter.

(http://www.internetwork.com/MAX/images/BlockDiag_3.0.jpg)

(http://www.internetwork.com/MAX/images/3-5-7kHz_sm.jpg)

The bandwidth display shows white noise used as the source and two MAXIM 295 Butterworth SCAF chips in tandem synced with a 555 timer. These are well worth considering for your project.

https://www.maximintegrated.com/en/products/analog/analog-filters/MAX295.html (https://www.maximintegrated.com/en/products/analog/analog-filters/MAX295.html)

Outstanding, Clark!

We have been waiting for such a product.  The spectrum scope roll offs look impressive!

I have used switched capacitor audio filter technology before. At least with the chips I tried, (one was a Maxim)  there was a hint of "cell phone" ringing that we often hear when a cell phone gets overloaded. The filter did a tremendous rolloff, but the slight "artificial digital" distortion was not good, so I abandoned it.  Hopefully you have solved this problem or have a different approach to using it.  

I also tried a switched cap commercial product from Indiom? or some name like this that Stu AB2EZ was experimenting with about 15 years ago. He sent it to me to try. It rolled off sharply too as desired, but had artifacts that a critical AMer could hear.

Steve/QIX mentioned he is going to be working on a DSP version of audio filtering too.  I am looking forward to testing that cuz we know DSP works well in the Flex, Anan, etc.

At this time I see no products on the mkt that will cover ham AM audio ranges, just CW/SSB stuff at 2.8 KHz, 1.2 KHz, etc., so I am rooting for you...

I will experiment with my passive Butterworth filters until one of you guys come out with something better. Let me know if you need any beta testing. I would be more than happy to help and test it on various AM rigs..

T

I run a stock Heathkit DX-100 with an original crystal D-104 non-amplified microphone.(high impeadance)
would your homebrew filter designs work for my application?
you mentioned 600 ohms.
AG5UM

Sure, they should work FB. The filters work best when the input and output impedance is matched. (less ripple, etc.)

For the DX-100, find a stage AFTER the mic and preamp that is at least a few volts of swing. Determine the input and output impedances and run the calculator using these values. The calculator is easy enuff to use after 10 minutes of playing around.  Be sure the capacitors are rated high enough in voltage for tube environments. 

Maybe later you will lose the D-104 and go in with a processor into a later stage. Then it will make it easier to slip a filter in the low level solid state 1 volt area.

Before you do anything, run some tones thru the DX-100 to insure it is clean without IMD splatter problems.

T

K1JJ,
sounds interesting, keep us updated on your progress. The calculator is easy to use,very cool.
I have done scope tests,etc. on the DX-100, I like and plan to keep the D-104.
I think I remember reading an article that you wrote that was very good about audio testing,what I remember
is the warning about overdriving the audio input and damaging the mod tfrm. Good article.
I think K4KYV,Don in Tenn. has one of the best sounding signals on the air, from what I gather he uses a
D-104 into a push-pull tube audio input.
I thought your idea is interesting for times when the bands get crowded.
73's
AG5UM

Tom,

Be glad to whip out a PCB for you. You going to relay switch the 3 filters?

John

Super, John!!!

I was just thinking of how I would do a point-point prototype board, but this would be great.

Leave big pads for the outputs and inputs.  I'll have to pick out the inductors to use, probably no bigger than a pea.

Interesting idea on the relays. I was going to use six toggles, three on each end of the board for input/output selection.   Not sure yet.  I will decide by Monday.

Thanks for the offer, OM.

How's the new rig coming along?  I made another 813s + parts   "offer he can't refuse" to Daryl, KD2AFL on the air the other day. I still have  four tubes and parts left.  He owns a modified T-368 and was talking about buying a KW-1. I asked him why not build a pair of 813s X 813s?  He is thinking about it. I mentioned your thread here on AMFone.  I told him you are approaching about 70% completion with a very FB rig build.

T

SMT or thru hole, I can do both. 1206 size SMT is really not very hard to solder.

I can use a footprint that can use either 5V or 12V signal relays (or 24V for that matter). That way you have options on how to wire it. I was thinking in my case, I could tie it into the Arduino sequencer. I have a touch screen on it that I could enable to select the bandwidth. Another guy could just wire up a simple rotary switch to select one of 3. Simple.

What did Frank say about his little op-amp board. Does it fit in this scenario?

Yes, around 70-80% of physical assembly. We both know that there is still a LOT of work after that on bring up. I have just about finished the mechanical work and will then wire up the RF deck. Takes a lot of time to get all of those big components mounted just right.

John

John,

I like the idea of thru holes so we can modify things without lifting SMT pads.

The relay holes option that can covert to a rotary switch or toggles is a good idea for versatility.

I think Frank's board will stay separate until we can test these filters out first. They will have some phase shift and I want to make some careful measurements to see if it is tolerable. All filters have some phase shift and more as they get more poles.   DSP is the way to go, but we are not there yet.

I think these filters will work FB when needed for crowded condix, and when conditions are quiet, just bypass them completely and do as we have done in the past - use a Desser, roll off the EQ, etc.  But I do like the idea of switching in a known filter set of roll off numbers to know exactly how far out we are transmitting vs: a pot knob that estimates bandwidth.

I'll order up the parts later today. I gotta see who has values near the suggested ones and their physical size for you to work with.  I did request standard values in the software run, but who knows what is really out there.

I've been adding a lot of bells and whistles to the shack lately.  I am going to wire Steve/QIX's REA mod monitor to show received signal pos and neg peaks.  I found a good 455 KHz tap on the FT-1000D and will buy a tiny amplifier to bring it up to 20 V p-p. This is a valuable service to give out receiving reports, IE what are my actual pos and neg peak numbers?   Also, Jeff/W2NBC sent me an SDRPlay RSPdx  SDR receiver which arrived today. Also, the filters we have been discussing.  I am also waiting for a Chinese mic cartridge to mount on a headset. That should be cool to be able to walk around away from the mic boom.

I also put a Henry 2K on line driven by the FT-1000D for easy, bandswitching 150 watt AM. Also added back in my dual 4-1000A linear amplifier for medium power service.  I even have a new laptop in the shack for all this stuff.   I am looking at multiple "power levels" for my rigs. Instead of having five 100 watt boat anchors to chose from, I have five rigs that gradually rise in power levels to match conditions and my mood... :-)    Fun days.

Later -

T

Tom,

Sent you an email.
There are tons of products on the market that do exactly this job, very well,
and are inexpensive, work on line level audio... :D

                     _-_-

Here is a bit more information about the audio processor that Clark / N1BCG and I have been developing over the past few months.  We are on the fourth revision of PCB artwork, and all that remains is testing and fine-tuning.  On-air testing has proven very rewarding.

As Clark mentioned above, it includes modules for Low Cut, All-Pass Filter, Pre-Emphasis, Compression, Limiting, SCAF (Switched Capacitor Audio Filter) for 3, 5, and 7kHz or user selectable), Peak Clipper, and Low Pass Filter.  Each of these functions may be enabled or disabled, either by panel-mounted controls, or set by on-board jumpers.  The modular design is thus very flexible.  

In addition, several inputs are provided for balanced, low impedance studio microphones, or single-ended high impedance units such as the Astatic D-104.  Balanced and unbalanced line inputs and outputs are provided for interfacing other signal sources and driving multiple transmitters.  Utility outputs for signal monitoring or a VU meter are included.  LED indicators are included for normal and peak input, as well as showing whether positive or negative peaks are predominant.  A special feature is included in the compressor module to eliminate the pumping effect when the operator's voice is paused, thus automatically adjusting the compressor level only when speech is detected.  A pad with AC coupling is provided, such that no transceiver modification is required to interface with microphone inputs.

Views of the PCB are attached.  In the very near future, more information will be shared as on-air testing is completed.

The MAXIM chips are the cat's MEOW. I didn't a little fooling around with them myself. I must, say though, your project looks awfully good. Would a 180 degree phase switch be appropriate?

Oh, one more thing. I had a problem with spurs. Are you having a problem with spurs? I seem to see some in the spectrum display. Maybe they're an illusion.

Yes, there is a phase inverter in the signal path, which may be enabled either by jumper connection or a SPDT front-panel switch.

In earlier versions of the artwork, we did notice the 50-X clock signal was not fully suppressed.  Two stages of LPF following the SCAF clean it up, and there are no apparent spurs produced by the SCAF.  It is my assumption that artifacts others have experienced with the Maxim chips may also be clock signal feed-through.

Hi Rick,

Tnx for the email.  Those pictures above - wow, that is some effort!

You may be right about the Maxim chip and clock feedthrough / artifacts.  A LP filter afterwards could be missing with the SCAF products I tested before. It may be the equivalent to Steve/QIX's anti-aliasing LP filter in the PDM generator.

After you added the LP filter did you hear any delay/overload sounds like a cell phone makes?  If that is gone then that would be FB.

T

Yes. The clock was getting through for me too, I see now, having found my sketchy notes. The clock or some mixing product from the clock was causing the spurs I had. You solved it the same way I did. I bet you figured it out a whole lot faster, too.

Your project really does look good.

Tom and Jon, thanks so much for the kind words. 

I don't recall hearing any delay or overload sounds emanating from the SCAF stages.  The SCAF chips are post-compressor in our implementation, so the range of signal levels to the SCAF was well-controlled.  The sound was crisp and clean, and scope display did not reveal any anomalies. 

Our LPF post-SCAF is just a simple RC filter in conjunction with existing op-amps, no additional stages were needed.

Folks:

The "MAX" audio processor looks quite good. Wonder of a kit will be sold? Looking forward to hearing it on the air.

On the SCAF (I used a Maxim Elliptical filter chip-two in series), I also found that feeding it well controlled audio level wise was the best as it's S/N is not as high as an op-amp or L/C filter, but for the SCAF the brick-wall filtering was excellent. The phase shift near the cutoff point will be high, but the clipper afterwards will take care of that.

Tom: it will be interesting to hear how you like the LC filter. I bet it'll work FB!

73,
Dan
W1DAN

The MAX processor was bandwidth tested on 1540kc at night with an SDR on an antenna and the transmitter on a dummy load. What appears as spurs in the bandwidth comparison image above aren't in any of the other tests (see below).

The SCAF chips do use a 555 timer to generate a clock frequency but that's between 100 and 300kc thus easily filtered out. Also, the bandwidth filter is after the limiter and before the clipper, which is followed by an LPF.

It's important to note that the limiter stage is very tight and prevents the clipper from getting over-driven, thus avoiding the clipping artifacts that have given this function a bad rap.

I've been using the prototype on 75M for most of the Fall and have gotten good reports.

Seems like the “Max” offers the important tools for effective AM control.
I have listened to N1BCG and can report his signal as clean (no audible artifacts), and loud!

Thanks for your efforts and looking forward to the W8KHK/N1BCG-type accepted version!!

I am ready for the "MAX" to put in in front of my Bauer. And hopefully the Collins 21-E soon. Please keep us updated on availability
KC4MNE

From a few emails I have received, I believe that the SCAF CAN be made to work well with no problems as well as long as the LP filtering and clock feed thru issues are addressed.  I look forward to comparing the Butterworth L/C filters to the SCAF for IMD, etc...  John / W9JSW and I are working on parts and he is doing a board layout, so it will be soon.

T

*** UPDATE:

I've decided to order parts for just ONE 5.5 KHz Butterworth L/C filter to mount on a prototype board. I will test it and use it on the air.  At that point I want to build and test a SCAF Maxim circuit to compare and decide which one is best to use.

Other than the data sheet circuit, does anyone have a MAXIM SCAF circuit that has the LP filter and other precautions we have discussed here to eliminate the distortion?  Maybe Rick will show me his SCAF-related circuitry since I may be able to improve on it and it will become public anyway in the future once the MAX is released in kit form.

T

Tom, I sent email to you this afternoon.....

Thanks, Rick!  That is a very generous offer and I believe everyone interested in good bandwidth control will benefit in the end. I will get back soon.  BTW, the schematics and artwork are very impressive and I can see you guys did a pro job.


I received an email today backing up our current beliefs about the Maxim SCAF:

K6JEK with his permission:  "I started with [xxx] mods to the SCAF-1 which uses the MAXIM chips. An HP distortion analyzer showed quite a lot of distortion. It was from the LM386 which [xxx] left in line -- completely unnecessary for line level. [xxx] also defeated the LP filter that was in the original SCAF-1 design which led to the clock getting through and mixing it up with the audio signal and causing all sorts of havoc. Fixing those two things resulted in no distortion I can hear."

I am encouraged as ever to try one again...  I was thinking that maybe my L/C filters could be switched in after the Maxim SCAF (replacing the existing LP  R/C filter) as an additional brick wall, or at least experimenting to see how effective it becomes. I notice the MAX processor's rolloff is sharp, (as shown in the screen shots posted earlier) though not as sharp as a DSP circuit. Maybe a more elaborate LP filter would help.

My new SDR RX (compliments Jeff/W2NBC) is working FB and I am always impressed and able to spot SDR transmitters on the air by their brickwall bandwidth. It's so intense that I can tell the difference between two SDRs running 3 KHz or 3.5 KHz ssb.  Amazing control.  Wouldn't it be cool to be able to do the same with all the big BC transmitters, the plate modulated homebrew rigs, modified riceboxes and the thousands of 100 watt boatanchors of the world?  Right now most AM bandwidth on the band is like the wild west but could be as smooth as driving the Autobahn.   ;D

T
 

Over a year ago I experienced the same shortcomings with the LM386, and removed it from my design.  Instead, interfacing to balanced 600 ohm input and output via op amps proved to be superior.

It would be interesting to see a comparison of the SCAF implementation to your multi-pole brick-wall filter solution, and this could be analyzed using the existing MAX processor PCA with minimal effort.  I am not sure there would be an advantage in using both the SCAF and the passive multi-pole filter.  It will be enlightening to see results of the two filter methods in tandem!  In any case, it is crucial to remove all remnants of the clock signal, post SCAF, from the remaining stages of the processor.  

One consideration is that the passive approach is keyed to a single cutoff frequency per filter, whereas the SCAF is continually adjustable, and may be preset to the desired bandwith(s) by switch selection of multiple clock frequency trim-pots.  Our testing previously indicated that a single SCAF was insufficient, but two SCAF filters in cascade provided an exemplary brick wall.  Further confirmation of this assumption would be invaluable.

Our goal is to produce a complete solution to the AM processor challenge, in a single device, and that may be based upon SCAF or passive filter approach,  determined by lab instruments and on-the-air performance testing.  Discussion thus far has focused on just the filter; but the processor includes all the necessary elements to produce a strong, clean, high fidelity signal, from the microphone all the way to the modulator input.  

It is certainly the right time to get more folks involved; so far we have been a team of two working for several months to arrive at the present configuration.  Results thus far have been very encouraging, but there is always opportunity for improvement!

I will be in touch and provide some beta hardware for your continuing analysis.  

73, Rick

Hi Rick -

Well, I just ordered the parts for the 10th order 5.5 KHz Butterworth LP filter. The bill of materials for Mouser is below if anyone wants to build it. Or let me get the arrows in my back first...  ;D  It's only 5 inductors and 5 capacitors. A few parts needed to be paralleled to get exact near values.  $10 plus shipping ain't a bad price to pay for a tighter, more controlled signal, right?

I should be testing it in a few days.  I desperately need  control of my extreme highs and hope to leave it inline for all my rigs until we make progress on the Maxim chip.
Yes the 10th order filter in cascade may just be too much adding excessive phase shift that is not a good trade-off compared to a smaller LP filter when using the SCAF. We'll have to see.

And yes, we should focus on the other areas of the Max shortly and maybe by the time it is ready for production we will all understand it enough to feel comfortable running it.

I do like the possibility of getting rid of some redundant audio processors in the shack. For instance I have a noise gate for ssb blower noise in one box and a desser in another box. The 31 band EQ is necessary. My favorite box is a CRL PMA-300A limiter  (compliments of my Secret Santa, W2NBC) which limits the positive peaks at 150% while the negative peaks at -99%. I love the way it is invisible until the highest peaks are hit and then it's soft knee gets to working. Hopefully the Max's limiter will work this way.  I don't like audio compression much and have only about 3 DB in... it brings up blower noise and is always working.  So as you can see the possibility of getting rid of the clutter with one Max box and an EQ sounds nice.  Who knows, maybe a later version will have an optional Desser and noise gate module.

You mentioned other modes. Does this have some utility on ssb or essb?

T

5.5 KHz Filter Mouser Electronics Bill of Materials below:

Tom, that is good information to help us move forward. I will await test results on your passive filter.

The goal of eliminating much of the rack processor equipment has been a priority in the design  of this "all-in-one" processor, but we realize that folks will still want to use multi-band equalizers and other tools.  This is the main reason we provide several line in and line out circuits compatible with balanced 600 ohm devices.  Our inputs are higher, and outputs are lower impedance than 600 ohms, which is the norm today for inter-operability.  All the modules fit together like an erector set, such that you may use an input/output pair for insert send and return functionality.  I believe this "feature" could be leveraged to interface your passive filter, using resistive loading so that it "sees" the proper input and output impedances, and let the existing op-amp circuits bridge to your existing chain with balanced 600 ohm links for noise and RFI immunity.  Your single-ended filter would lie inside the "circle-of-protection" afforded by these interfaces.

Each module in the processor includes a four-pin header to allow the user to include or exclude that stage from the signal path.  Pin one is ground, such that shielded cable may be used, but so far we have found that to be unnecessary.  Pin two is the signal into the current module, and pin three is the "send" to the next module, while pin four is the output of the current module.  To include the current module, connect pin four to pin three, or to exclude the current module, connect pin two to pin three, either with an on-board jumper or a panel-mounted SPDT switch.  

With this modular approach, you will find it very easy to insert your passive filter array in place of the existing SCAF, or in cascade, your choice.

The Maxim chips include an additional op amp.  On the first Maxim chip, we use the op amp as an input interface to avoid loading the previous stage, and to provide unity gain through each module, as necessary to avoid level shifts when inserting or removing a module on the fly.  (Due to the operation of the internal voltage controlled amplifier, the Maxim compressor input exhibits a rather low and varying input impedance.)  The op amp on the second Maxim chip is used as an output buffer to avoid loading the  Maxim chip, no matter what module is connected following the SCAF stages.  In this op amp circuit we have added R148 and C148 to attenuate the clock.  This should have minimal impact on the audio, because the clock is 10 to 30 times the frequency of the voice components.

I have attached a few "snips" of the circuit to illustrate these components that remove the SCAF clock from the audio.  You will also notice fixed resistors R145 and R146 in parallel with trim-pot R147.  This pot is used to fine-adjust the SCAF module to unity gain, and the fixed resistor pads may be used to lock down the attenuator, thus eliminating the pot in the final design.  You will see this combination in several other areas of the device.  The extraneous components will be removed from the final artwork, thus reducing the setup and calibration steps to a minimum.  EDIT: For clarity, I should mention that either the pot, or the fixed resistors will be present, but both will not be present at the same time.  The fixed resistor values (ratio) may be determined based upon the required pot setting during fine tuning.

Two additional optional LPF for clock elimination are in the main output driver stage, R221 and C222 and the post-clipper stage, R209 and C201.  All these are optional, and the goal with this board rev is to determine the bare minimum of filtering with total removal of the clock signal.

As it is obvious we will have at least one more revision of the artwork to nail down all fine-tuning issues, it is certainly possible to add more features needed to eliminate other external components in the audio chain.  

Currently the features following the input buffers include a low-cut filter to reduce the amount of bass power and thus improve readability in heavy QRM or QSB situations, but this is done "gently" to maintain good voice quality.  Following this filter is an optional phase inverter, to allow selection of high positive peaks.  The all-pass filter, AKA phase rotator is the next optional module.  Prior to compression and limiting, we added pre-emphasis, and when used in conjunction with the low-cut filter, the signal is significantly more understandable  when band conditions are rough.  A separate op amp and comparator are used to visually indicate normal voice level and high peak level, as an aid in adjusting microphone level.  The comparator also controls LED indicators to identify whether positive or negative peaks are predominant.  The compressor/limiter is so tight, a clipper might seem unnecessary.  But for the occasional spike that might not be caught by the limiter, a soft clipper is provided.  The clipper may be set off, or 100% for pos and neg peaks, or 125% pos and 100% neg.  You mention using 150% pos, and this could easily be added as a third switch position.  With soft clipping, minimal LPF is required.  As described above, this LPF may double as a clock attenuator as well.  The compressor response time may be adjusted for open or dense performance, and the novel "gain gate" allows the level of compression to remain static in the absence of voice.  This might preclude the necessity for the noise gate you are now using. Lastly, several output circuits are offered to drive multiple transmitters, monitoring functions, metering, etc.  This allows the input level to each modulator to be set for each transmitter, thus simplifying band QSY.

Your suggestions on documenting voltage and signal levels, as well as troubleshooting information and circuit functions, are all in the works and will be available on the web as the project is finalized.

While we are primarily focusing on a device for the avid AM'er, I see absolutely no reason why it would not work for SSB, FM, or even digital voice.  LPAM is another possible application.  After all, we are simply providing a clean, consistent signal to a modulator, nothing unique to AM-only whatsoever.  

One area where this little box might shine is in the schools when hams set up for students to chat with the ISS astronauts.  In cases I am aware, the teachers ask students to submit the questions they would like to ask the astronauts.  Then students vote on the questions, and the persons who submit the most popular questions get the opportunity to actually talk to space!  Some are shy, and barely whisper, while others may be so excited they almost yell.  

I recall when testing, Clark tried speaking softly, then he yelled loudly into the studio mic.  The compressor and limiter performed admirably.  Then he took a long pause.  He told me the noise I was hearing was not a bug, but it was his dog snoring in the living room.  Hence the development of the "gain gate", adjusting compressor gain only when voice is detected.  Might this preclude the need for a traditional noise gate?

I will coordinate the non-technical details of getting test hardware to you offline.

Hi Rick,

That's a lot of FB info. I read it over a few times to absorb it all.

I see instead of one LP filter for the clock you have scattered filter components around to work as an overall system. I also like the module and in/out approach to keep things versatile and also give users the chance to bypass things they don't need or add in things they do.

I made my first mistake on this prototype passive filter by ordering five ceramic capacitors. It turns out film polystyrene caps are better for audio. I will get the right ones for the real circuit boards assuming the tests work out reasonably well. I plan on using a 600 ohm resistor shunt on both the input and output to match the filter's requirements.   Turns out my own insertion point is high impedance in/out, so this will work to reduce the ripple.

Let's look forward to testing your SCAF version soon and later on to the whole unit.

You obviously have invested a lot of expertise, time and money into the project so far and a successful outcome will surely be coming.

  T

Tom:

The Maxim Elliptical SCAF is just about as as sharp as the sharpest DSP filter. My dual-SCAF was something like 36dB/octave after the cutoff point. Really you don't need that sharp of a filter.

Rick: The LM386 suffers from high amounts of crossover distortion, and noise. Have you considered using a second order Sallen-Key filter for your Main LPF?

https://www.edn.com/a-sallen-key-low-pass-filter-design-toolkit/
http://sim.okawa-denshi.jp/en/OPseikiLowkeisan.htm

Enjoying watching both of your developments!

73,
Dan
W1DAN

Dan, Thanks for your input and interest.

The LM386 has many known limitations and is not used in the audio processor.  Thanks for your suggestion on the LPF.

I attach a snip of the LPF section we used in version two.  It is no longer included, because the SCAF performs the required bandwidth filtering, while a simple RC added to an op amp following the SCAF eliminates the clock remnants.

I also attach a snip of the current low-cut filter, which is the complement of what you suggested, with the resistor and capacitor positions exchanged to provide a high pass filter instead of a low pass filter.

73, Rick

Rick,

In your Max trials using the Maxim SCAF, did you or Clark notice that the natural asymmetry of the voice was stripped to become  more symmetrical; something similar to what an all-pass filter (phase rotator) does to audio?

A friend of mine suggested that a brickwall filter may do this to a signal in the form of phase distortion.  Did you do any phase shift in/out measurements?

I know some guys like symetrical dense audio. I prefer high dynamic range and assemtrical positive peaks that naturally form.  So I will be looking closely at any phase distortion effects.

My friend felt that a more mild roll off would preserve the asymmetry.   Until we run some tests, what do you think?

T

Tom, you bring up an interesting point.  I have observed the phase distortion in the all-pass filter, as expected.  I have not specifically tested the SCAF for phase distortion.  I will do this test soon and report back.  Perhaps the resulting signal is more symmetrical because some of the higher frequency components are the major contributors to the asymmetry?

If the SCAF proves to cause phase distortion and creates a more symmetrical output, there certainly may be cases where a passive multi-pole brick wall filter is optimal.  In this case, perhaps the new processor could include both the SCAF and the passive filter as operator-selectable options.  Time will tell....

Hi Rick,

OK on all.

Unfortunately all filters including passive filters like mine will cause some degree of phase distortion.  That's why I am worried about my large L/C filter.  I have tried an all-pass filter (phase rotator) before and hated the effects cuz I lost my big positive peaks. But I know guys who run an all-phase filter deliberately to get symmetrical audio and they love the big carrier and dense audio at 110% positive.

It this happens with mine, I will cut it to half size and try again.

A DSP filter module for the Max is do-able, though not easy without some sophisticated work. My bet is a DSP filter will have little to no phase distortion, but I am not sure.  

I'm very interested in what you see with the Max SCAF.  To test it, looking at your peaks with the filter in and out will show the difference in peaks.   Measuring the phase distortion with tones is something I'd have to think about.  Maybe a dual trace scope looking at the the input and output phase difference would do it.  

T

Well, Tom, your assumptions appear to be valid regarding phase shift.

I ran a test today with two MAX296 chips in cascade, with a clock freq of approximately 285 kHz, which should yield a bandwidth of around 1/50 of the clock rate, or 5.7 kHz.  The two SCAFs seem to behave a bit like the phase rotator, with the following shifts recorded:

Starting at 0 degrees phase shift at 210 Hz, I recorded

degrees      frequency
00                210 Hz
90             1.42  kHz
180           2.60  kHz
270           4.10  kHz  
00             5.60  kHz
90             6.60  kHz
180           8.40  kHz
270           9.60  kHz
00           12.40  kHz

That is just about 1.5 revolutions of phase within the desired audio bandwidth.  I would assume one stage would generate half the phase shift, but unfortunately a single stage does not provide a steep enough cutoff characteristic.

We have a couple MAX295 chips available for test, but no MAX294 elliptical devices in stock.

I do not know how many degrees our four-pole phase rotator adds.  I can measure it as well, if that information would be of interest.  Perhaps another test of value could be to pass some asymmetrical audio into the SCAF, and see how symmetrical it comes out the other side.

It would be very interesting to see a comparison looking at the phase shift you record, per pole, of your passive L/C bandwidth filter.

Much of the MAX audio processing is based on the same principles as broadcast processors. Why reinvent the wheel with over half a century of proven R&D that's implemented in the field? Here are the main points:

The All Pass Filter (APF) reduces asymmetry as a means of increasing loudness. The problem with "as-is" microphone audio is that there is no proper polarity, only majority polarity. It is quite common for lower vocal frequencies to have the *opposite* asymmetry of higher vocal frequencies. You can see this on the polarity indicators when saying "aaaahhhh" and "eeeeeeee" which will light opposite polarity LEDs.

Even when an air chain is carefully checked for polarity it is only correct most, but not all, of the time. Many have noticed that adding a low-cut filter dramatically increases positive modulation. What is actually happening is that the lower vocal components with opposite polarity are getting suppressed.

Another benefit of the All Pass Filter is that it increases average modulation without an increase in compression. This happens because gain limiting stages such as compressors and limiters act on the highest peak of audio regardless of polarity. More gain reduction and lower modulation will occur if the audio is highly asymmetric.

Finally, the use of All Pass Filters yields more consistent asymmetric modulation since the clipper stage acts on each half of the waveform to different degrees, re-creating asymmetry regardless of originating content.

Users who need to be mindful of component ratings greatly benefit from the use of an All Pass Filter since it decreases the chance of arc-overs in output stages or modulation transformer winding damage.

Despite all this, the APF, like every function, can be bypassed in the MAX processor.

I've used the Maxim elliptical filter chip in the Rx channel of a couple homebrew transceivers and have never noticed any "phase distortion" effects.  Is it really fair to call it phase "distortion" when its audible effects are unnoticeable? 

I've read several times that phase shifts through audio filters produce inaudible effects.  The all-pass filter (phase rotator) seems proof of that.

What's the big worry??

Rod

FB on the tests, Rick.  Thanks.   Good info that we didn't have before.

Clark: I'm still open to trying an all-pass filter again and seeing the benefits you described. I have seen MOST of these effects, especially the lower freqs having the opposite polarity as the highs. And it will even change as the day goes on. Early morning deep voice phase morphs into late day regular voice phase. Trying to optimize polarity can be like chasing our tails sometimes. But once we randomize it, I can see how it is more controllable.

I'll see what my passive filter does to the audio.  I think it will rotate the phase probably as much as the Maxim SCAF did for you, Rick. If so, I'll try a different approach and see what more symmetry does to the overall signal.

Right now I am seeing 150% positive peaks with -99% negative. Maybe it's just an illusion with low average power. It even required switching in the lows cut on the mic as Clark mentioned. Overall, big audio peaks seems too fragile and temperamental to me.  I could be swayed very easily when I try symmetrical audio again.  It would solve my filter roll off problem if I didn't care about phase shift in the filter.

There are guys who do very well like Steve QIX and several of the class E rigs using big asymmetrical peaks. And my good friend Chuck K1KW put in a big effort to get rid of polarity and run denser audio as a result. And he uses a DSP filter system in the ANAN and could have chosen "no phase shift" and high peaks if he wanted. There are several ways to get there.  It's not an easy decision for me but I'm still flexible and could end up with a totally different audio processing approach a month from now..

Clark: When you use an all-pass filter and lose the big peaks, is there still an optimum phase to use? Does it become more stable thruout the audio spectrum and as our voices change?   My guess is if it is really symmetrical, it will be difficult some times to determine the proper phase, if any.

The REA QIX mod monitor makes it very easy to see the proper phase right now for me. I wonder if the REA is still reasonably clear with more symmetrical audio.

**Clark said:   "Another benefit of the All Pass Filter is that it increases average modulation without an increase in compression. This happens because gain limiting stages such as compressors and limiters act on the highest peak of audio regardless of polarity. More gain reduction and lower modulation will occur if the audio is highly asymmetric." **   Very interesting.


T

Clark: When you use an all-pass filter and lose the big peaks, is there still an optimum phase to use?

An APF will greatly reduce asymmetry but not eliminate it. Whatever is left will be slight and you could optimize it so that negative clipping is minimized.

A big advantage to broadcasters is ease of handling a wide variety of audio content. Many different mics from different studios, different announcers, tape recorders in the field, ad agency recordings... it's a mish-mosh of inconsistent audio that is "made right" by an APF.

Does it become more stable thruout the audio spectrum and as our voices change?

Yes. Using an APF thoughout adolescence will help stabilize your modulation during those formative years and avoid shame and embarassment.

**Clark said:   "Another benefit of the All Pass Filter is that it increases average modulation without an increase in compression.

Clark says a lot of things, but at least this stuff is useful. Not only do you lose modulation but the extra recovery time adds to the effect. It's important to note that every broadcast processor made in the past 40 years includes an APF. Success leaves clues.


Some important APF usage notes:

1) All Pass Filters must be installed *before* any compression stages.

2) Those that like to listen to themselves with headphones while transmitting will hear a significant difference after switching an APF in-line. This is due to bone conduction where you are hearing your own voice partly through the headphones and partly though your upper jaw. These two paths may be in or out of phase.

This is why switching mic phase makes a HUGE difference in headphones, and an APF will have a similar but less pronounced effect. It's interesting to note that only the op hears the phase effect and no one else.


The SymmetraPeak unit was the original All Pass Filter developed for the broadcast industry decades ago. I've attached a PDF of the brochure. For those wanting a more technical description, the patent can be read here:

http://www.w3am.com/SymmetraPeakPatent.pdf (http://www.w3am.com/SymmetraPeakPatent.pdf)

K1JJ,
What happened to your simple passive fiter idea to NARROW BANDWIDTH for AM transmitters??
Did they talk you into working on digital audio processing instead?
I know from the guitar-p.a.-studio world that it is a never ending-subjective-expensive pursuit.
The guitar world has gone to very expensive NEW tube equip. designs, tube compressors etc. etc. etc.
So, be prepared to sell the house, if your going down that road...(it is Fun though).
ANYWAY, what happened to your simple design to narrow bandwidth, I thought that was interesting.
73's to all,
AG5UM

Scope creep. Happens all the time without proper project management and firm requirements/goals...

"Does it become more stable thruout the audio spectrum and as our voices change?"

Yes. Using an APF thoughout adolescence will help stabilize your modulation during those formative years and avoid shame and embarassment.

ROFLMAO!

The deeper I dive into this subject, the more I realize how heavily divided amateur AMer opinions are regarding using an all-pass filter (symmetrical audio) or letting the natural voice positive peaks go where they may. (asymmetrical audio)  I have received several emails from some very infomed AMers that lean both ways.

I have no skin in this debate, so I will be trying several approaches at this point:

1) The natural approach:  No LP or all-pass filters at all.  I have never heard this done before, but I have a DSP Berhinger 9024  six band compressor-limiter that will be put to use in a unique way. I was playing around with it the other day to VERY heavily limit and compress the 5KHz to 12 KHz band, thus producing a brick wall at those freqs. The other lower audio bands (< 5KHz) would have little to no 9024 band processing. It was not quite DSP brick walling, but close... I saw some very desirable look-ahead DSP effects to limit my extreme highs while at the same time preserving the positive peaks that my own voice generated.  I should know the results of these tests soon.

2) Natural with mild LP filter rolloff:  Same heavy Behringer 9024  5KHz highs limiting/compression but in conjunction with my L/C passive filter using only 5th order filtering, a smaller filter - slower roll off, less phase shift.

3) All-pass:  Try my all-pass Chinese filter board again, right after the 528E mic preamp to randomize the rotation, looking for symmetrical audio.

4 All-pass and LP filters:  Use the all-pass filter in conjunction with my 10th order passive L/C filter (parts to arrive next week) and see what symmetrical audio AND a sharp roll-off does.


Again, personally, I have been looking for a natural sound with high dynamic range where people in person might say they recognized me by my voice. (with a sharp 5 KHz roll off)  I don't want to be in a morning drive used car commercial with tiring audio..

The last time I used an all-pass filter, (10 years ago)  I was disappointed. I read the two articles posted above, which is what got me interested. But for whatever reason, with symmetrical audio I felt like I was fighting soft modulator tubes or I had the linear amp loaded too lightly or I was always out of phase.  It may be due to old habits - because I am used to seeing and hearing high positive peaks over the years. So I switched back to the more natural sound since.

But I have an open mind and will be trying the four techniques listed above and will report back my findings.  I may find the more agressive set of parameters are better for nightime QRM condix and the other smoother sound better for daytime quiet condix.  It may mean there is a hard rock AM radio audio and a lighter ham rag chew audio that is the issue.

In these kinds of discussions we all learn very quickly and cover a lot of technical ground, which is a good thang.

T

I'll bet (Hi Dave!). Try this topic on-air and all hell will break loose. Have your pitchforks and torches ready. As an example of eliciting angst...


Multiband processors were the solution to music and mixed content audio where bass notes from drums, etc, were causing mid and higher frequencies to be "ducked"  (Optimod 8000). Separating the lows from everything else eliminated this problem (Optimod 8100). In contrast, studio voice processors are all wideband.

Compressing the snot out of the sibilance frequencies will guarantee angst on the bands because it ensures that these frequencies are at 100% modulation at all times whether you have vocal energy there or not. Hiss, background noise, harmonics, anything that exists in that spectrum will be maximized regardless of the rest of the vocal frequencies. Band gain coupling goes a long way to reducing this effect as does careful use of multiband processing for voice processing.

FWIW, the folks who have built up switched capacitor filters and that I have heard on the air
were all quite audibly noticeable in a negative way... ymmv.

K1JJ, passive audio filters require stable input and output impedances in order to get the
desired response. So, buffers and load resistors may be mandatory.

Yes, I was just told the same thing about compression.  Thanks.   So heavy limiting of the extreme highs band instead will probably work well?  The 9024 gives me this choice.


Bear:  I have 600 ohm shunt resistors across the filter input and output, so shud be close.

T

I use two compressor/limiters.

The first compressor/limiter offers gently compression, and hard limiting only of excessively high levels (cough, sneeze, bumped mic). It is connected post EQ so as not to respond to frequencies outside of the bandwidth I wish to transmit. The EQ offers a gentle rise from 1kHz to 5kHz and rolls-off above that.

The second, and final compressor/limiter, has an EQ connected as an insert. That EQ is set to cause the compressor to heavily compress/limit a only peak around 6.3kHz. Connected in this way, this compressor operates without make-up gain and, in effect, is a de-esser. A form of multi-band compression. It causes gain reduction only... no angst.

It comes down to how much gain reduction (in each band) occurs with normal program content.

Limiters are fast acting compressors. In an oversimplified example, compressors deal with average levels while limiters handle peaks. The UREi BL-40 Modulimiter is a great example of each function put together.

Unless there's a platform function that separates compression (gain reduction) and expansion (gain increase), all compressors and limiters reduce gain when an input level exceeds the set threshold.

What goes down must come up...

Compression and limiting will reduce gain when needed, but when a high input signal is lowered, the circuits will "follow" this change with a corresponding increase in gain back to unity. The "heavier" the action of compression or limiting, the more dB this gain change will be. That's why heavy limiting (or compression) at all times will result in dramatic increases in whatever range of input frequencies that circuit is fed *unless* there is no gain reduction with normal content.

This is how the arrangement sounds with KK4YY's de-esser, that is, no gain reduction of sibilance frequencies unless needed.

Now I know another thing that doesn't work...   I got the Berhinger 9024 multiband DSP audio box to roll off the extreme highs, but it was not sharp enough. As hoped it did not effect the positive peaks, but I still need something sharper.

So nix the 9024 idea.  I've also tried two high cut modules in the EQ and the 528E with poor results. I also tried the de-esser in the 528E and not sharp enuff to kill the 6-7 KHz extreme highs.  The sss's are killing me even though the rigs are clean.  

I'll try the passive filter next week when the parts come.  I will focus on seeing how many poles of the 5.5 KHz filter I can use before it affects the positive peaks phase excessively.

There is a solution that I will eventually find to effectively control bandwidth one way or another.



** "The second, and final compressor/limiter, has an EQ connected as an insert. That EQ is set to cause the compressor to heavily compress/limit a only peak around 6.3kHz. Connected in this way, this compressor operates without make-up gain and, in effect, is a de-esser. A form of multi-band compression. It causes gain reduction only... no angst."  

Don / KK4YY, could you elaborate more on the practical setup and adjustments of this idea?   I might like to try it...

T

The compressor/limiter I'm using is the Symetrix 425 with a Symetrix 532E graphic EQ as an insert (sidechaining). This is how the Symetrix 425 operating manual explains it:

---

8.7  Sibilance Control

Patching an equalizer into the sidechain can cause the 425 to respond more or less to selected frequencies, giving it the ability to make sibilance problems less apparent. Fine tuning between the compressor threshold, ratio, and the EQ boost applied in the sidechain will have to be made to arrive at premium results.

To find basic settings, start with a fairly high ratio (5:1 or so), and a compressor threshold setting between -20 and -5. Cut the low frequencies on the equalizer and give a 15 dB broadband boost to the EQ at around 5 or 6kHz. Now, carefully "tweak" the threshold setting as you count "four, five, six." What you're looking for is no compression on "four,five," and somewhere around 9 dB of gain reduction on the word "six."

You can refine the setting by listening to the equalizer output and adjusting the EQ to emphasize the sibilance in the source. Remember that you're equalizing the signal to emphasize the sibilance, not to sound groovy. Let the 425 do that.

Do you have a recording where the cymbals drive you nuts?...try the same technique on the overall mix.

Set the peak limiter threshold for 6 dB of gain reduction when Sam Screamer is on the system.

All normal signals will be slightly compressed, and really loud signals will activate the peak limiter. With these settings a shy person will be audible, and the guy who thinks he has to shout won't be too loud, or cause distortion.

---

BTW, the Symetrix 425 is not my favorite compressor. I used it because I had one. But I wouldn't recommend it to a friend.

Sounds like you are proficient in making pcb’s? Have you ever considered making a pcb for Steve’s audio peak limiter? I tried, but there were a few things about the pcb software that confused the heck out of me. I bet several people would be interested. I’ve already collected all the parts.

Don, thanks for the sibilance control info.   I am still experimenting and will try it soon.

I removed my PMC 300A limiter and found a cleaner, more natural sound, as expected. Even the extreme highs seemed 1 KHz narrower and cleaner on sss's. Not sure what that is all about - maybe some IMD from the limiter doing its job. The positive peaks are also about 10% higher than before.  

So right now I am using just the Symetrix 528E -   the Berhinger Ultragraph 31 band EQ side-patched in, then 3 DB of 528E compression... that's it.  I really like the natural line up of positive peaks phasing based on my own voice. The A/B mic phase switch is dramatic. If this station were used by other people it would be a different story.  Maybe just a small touch of LP filtering will do the trick when the parts arrive.

Bob, could you post the QIX limiter URL or schematic?  Is that his 5-band homebrew version?

Someone mentioned to me last night that Steve has one of his low freq audio bands phase-flipped to add in better to the rest of the lineup. That sounds like an intereting idea to tailor ones own audio phase when one of the bands is 180 degrees out, etc.  I can imagine a 5 band processor with a seperate phase toggle switch when entering each band - rather than a single mic switch covering all bands. Customized to one voice.


Just thinking out loud...   ;)

T

I printed the schematic out at home. I recall Steve posting it within a post about something totally different. I did a quick search but could not find it. Perhaps Steve can repost it?

Well, I received the parts for the 5.5 KHz LP filter and built it on a terminal strip prototype style.  I measured each part for capacitance and inductance and found them off a little. I padded or subtracted small values to get them as close as possible. The filter is not perfect but still built reasonably close to design.

On the bench, I ran the tones thru and was very pleased at the roll-off.    With a full scope screen of signal there was no amplitude change when going from DC up to about 4.7 KHz. At this point going higher in freq made the amplitude decline very slightly until about 5.0 KHz. Then it dropped like a rock after 5.5 KHz and continued down to a faint signal past 6.5 KHz and was gone at 7.5 KHz.  


I didn't make any phase measurements  yet, but at least the filter design worked very FB.

This is exactly the rolloff I would want if I had a wish list... :-)

So now the question is how much it will affect the transmitter being such a bad-ass filter of 10th order?  I'm hoping that if it works reasonably and has a roll-off like this, I could use it when condix get crowded and leave it off when the band is quiet.    Having a 4.0 KHz and 7.5 KHz version of this would round out the harem.


I'll try it on the rig later tonight and see...  I remain 50-50 optimistic.

T

Get the QIX schematic and I will take a look.

Now it's Schaefer time!

(please excuse the rough camera to screen shots for now)

I installed the new passive L/C 10th order Butterworth LP filter into the last position of the audio chain before it goes to any transmitter or modulator.  With it bypassed I set up the audio for the cleanest sound and normal  high fidelity with no high cuts or any other type of roll off. Just the 528E mic preamp driving the 31 band EQ and then 3 DB of compressor action...   The first picture below is the spectrum of the audio (NO FILTER) when socking yallos, chhhhh's and ssss's. IE, whatever I could do to talk it up and make some splatter, though I did not overmodulate.


As you can see, this is why I was concerned about bandwidth. The highest highs were extending out to 10-12 KHz.  Sounds clean in the monitor and the IMD tests out good. This is simply what is going into the mic without restrictions.  The waterfall is abominable.

Now, the next two pictures show exactly the same types of modulation activity except the filter is in-line. The filter is the last thing in line in the audio chain before it goes to the modulator, etc. Notice the bandwidth cuts off like a haircut at about 6 KHz. The waterfall is sharp and clean. The spectrum shows a beautiful roll off slope.

The good part: When A/Bing the filter in and out,  I could not see any meaningful difference in the positive peaks. If any, maybe 5% less but hard to see with voice modulation. I am still able to get the near-sharkfin audio pattern I like to see and the peaks bounce up to positive 150% with -100% negative. These pics are not a good representation of the true extent of the positive peaks because I am doing a camera to screen shot with poor control of everyting.

Bottom line is I think this is what I have been striving for. The filter is rolling the transmitter off sharply just as it did on the bench WITHOUT adversely affecting the positive peaks. With the filter in, both the scope and the REA mod monitor show the peaks hitting similar levels as without the filter.

Needless to say I am estatic about these results and will now give it some time on the air doing A/B tests. Once finished, I will see if John/JSW still wants to make up some filter boards. I definately want to make a 4 KHz and 7.5 KHz board too; maybe all three on one board in a single box.  I like the idea of switching in a brick wall filter number and living with it rather than turning a knob. It's a man thang.

BTW, notice in the last two pics the roll off on the spectrum scope. I like the drop off a lot. I'm not sure why the phase shift (not measured yet) is not adversley affecting the peaks, but that's OK by me. Using the REA mod monitor I often see 140% positive peaks with -85% negative. Some may not agree with this approach, but this is exactly what I have been trying to achieve.... big positive peaks and controlled narrow bandwidth when desired.

SLAP ME FIVE!

T


Actual RF transmitter shots taken while modulating 1500w pep into a dummyload...

Pic 1 is bad, no filter:  10-12 KHz highs spreading out from 3760 to 3749 area


The only change is the filter....

Pic 2 is good, with 5.5 KHz filter:   6 Khz highs haircut shown on waterfall, from 3760 to 3754 area

Pic 3 is good, with 5.5 KHz filter:    6 KHz haircut on waterfall

To better show the roll off on the spectrum scope, I injected a hard series of ssss's simulating white noise. Notice it brought the bandwidth out one more KHz. Normally these should not show up in normal speech too often. Notice the positive peaks are pinned while the negative are under -100%.  This is a sign that for my voice, anyway, the extreme highs are in phase.

A word of warning for those building a filter... during sweeping I did notice a slight hesitation pause in the roll off at about 6.7 KHz. All other freqs were smooth. I think it may be due to part tolerances.  I also noticed that the inductors should be at right angles to one another or they couple magnetically and effectively shunt themselves. I saw an inductor change from 25 mH to 28mH just by moving it in alignment to another.

Notes:

*** BTW, my mic polarity switch stays in the same polarity whether the filter is in or not, further indicating there is little phase shift happening, at least to frequencies that favor my own voice.

Also, this filter was populated with undesirable ceramic caps. I will use film-poly caps next time which are better for audio. Can it get any better?

** The roll off shows a -45DB drop starting at 5.5 KHz and ending at about 7.2 KHz.   IE, if someone were +40 DB over S9, their side products at 7.2 KHz up the band would be down to about S8, near the average 75M noise floor.  The proposed 4 KHz filter will be even tighter.

**  The audio quality of 5.5 KHz bandwidth sounds really FB to me in the monitor.  It's really more like 6KHz before the filter really kicks in. That is plenty of fidelity for average band activity 75M voice operation.

*All tests were made at about 250 watts carrier or 1500 watts pep on highest peaks showing the amplifier(s) IMD is good and we are dealing mainly with low-level audio bandwidth control.

** The passive filter insertion loss is only about 3 DB or so. I will measure more accurately later.  The L/C inductors keep the feedthrough resistance low compared to R/C resistors.

T


Below:   ssss's white noise with filter in   -  Check out the spectrum roll off at 6 KHz and above:

Nice. A passive Courtesy Filter. Every man should own one!

A good source of broadband test noise is from an FM broadcast receiver with no antenna and tuned to the end of the dial where there are no signals. It's close to white noise, but technically pink noise due to the 75uS de-emphasis in these receivers.

Nice work on the filter, Tom!  It will be interesting to read more about the tests and hear it on the air.

Thanks for the comments guys. I just knew it could be pulled off in the end... :-)

A "Passive Courtesy Filter"   (PCF) I like it!   ;D

I'm building a small aluminum box for the filter with toggles to do fast A/B tests. I'm curious what a trained ear like the Tron has to say about its insertion invisibility.

I'll take a picture of the unit, but it is just a mass of parts on a terminal strip.  John/JSW and I are talking about a PC board at the moment.


Here's some more notes about the filter PCB layout regarding actual exact parts availability:

When two inductors (like the barrel style I am using) are put in series or parallel, they need to be physically separated or they will couple together. Also previous stages using one inductor need to be away from each other. They use magnetic cores here, so are like mod transformers for audio. This applies to all inductors in the circuit…. Keep them separated by an inch or more if possible.  One way to isolate them is to keep them apart at least a few  diameters from each other or place them at 90 degree angles so that their poles are opposing coupling.

I was thinking that every part should have extra pads for each lead to allow for series or parallel insertion in case we need to adjust values or use parts that need parallel or series to help availability.  It’s a simple circuit but the parts availability makes it tough. I had to struggle getting everything close to the exact values in the design.

There’s no problem making the board bigger to allow for good filter isolation since we are building three filters on one board. IE, not a time to make the board tiny.   Unwanted poles coupling will degrade the roll off skirts and probably create unwanted ringing or ripple effects.  

T

Tom,

Is the plan to make separate filters and switch them in/out, or to connect narrower sections in series with wider sections to narrow the bandpass as new sections are added? You may be able to reduce the part-count by adding sections in series that way. Just wondering.


Don

Hi Don,

Good pondering...

I don't know enuff about filter design to answer that with any expertise, but according to Frank/GFZ, the filter should be used in its design entirety. To tap off the middle to simulate a smaller filter will upset the design. IE, use it from beginning to end with 600 ohm loads. I use 600 ohm resistors at the in and out right now. The insertion point will add some more loading, but I did not optimize that yet.

As for putting them in series/ cascade, I think the same problem occurs since the overall filter doesn't adhere to a "systematic" design from a single job optimization. I believe all parts of the filter work together and do their jobs no matter where in the circuit they are. But you could be right about making a 5th order set of smaller filters (each terminated) and putting them in series, why not?


That said, I will surely try your idea with the finished filters to see what happens when it goes into a 7.5 KHz filter and then enters a 5.5 KHz filter. Will the rolloff be cleaner? I dunno yet. How about two 5.5 KHz filters in cascade? Shud be an interesting test.  My goal would be to approach DSP skirts until the audio showed signs of artifacts or any kind of modification in the lower freqs below cutoff.

For now, I plan to keep it simple and bullet proof - make three separate filters, all optimized via the computer and selecting parts to exact values, and either use relays or toggle switches at each end to completely disconnect one filter and add in the next one.  I want poles and input/outputs well isolated from each other, just like we do with RF amplifier design.  

Based on this 5.5 KHz working filter, I know the 4 KHz and 7.5 KHz version will work for sure now.  


Last night I tried it on the air for a minute and had to smile as I watched the skirts cut off at +- 6 Khz. Before this I would cringe and worry who was nearby and talk more softly just in case...  ;D  Just look at my first pic again (showing 10-12 KHz  pre-filter bandwidth)  a few posts back to appreciate what I mean. And that was using the same clean transmitter as shown in the next pics with the filter in line.


T

That's what motivated me, too. I always got glowing audio reports using my QIX designed class E but then I started using the KFS web SDR to listen to hard-to-copy guys. Holey moley I was as wide as the Montana sky. Hence, my SCAF box.

I'm curious, though. Why put the bandwidth limiter at the end of the audio chain? I put mine near the beginning on the naive assumption that it made no sense to do compression and equalization on frequencies that I planned to chop off later. Maybe it makes no difference.

Your results are darned impressive.

Jon

Thanks very much Jon!

That's a good question that I haven't thought out yet.  Where to place it in line???
I put it at the end thinking that if there were a chance of ANYTHING generating frequencies above 5.5 KHz, I wanted them cut. It won't help IMD generated by the amplifiers later on, but I wonder if there are frequencies generated within the audio gear due to mistakes such as overdriving, the EQ too hot or whatever, that this would be the ultimate audio trap.  I'd like to hear other opinions on this. Where do they usually place high cut filters in commercial gear?  It would be easy for me to move it.

Tnx for the question.

T

And I finally put it in a beat up, used mini-box. This is a prototype AKA very JS, but solid and works FB.

I used a toggle switch at each end of the box for in/out.  This will be in line for daily use until I build a PCB version for three filters and put them in a nice box.  This unit can also be used later to test the effects of putting two filters in cascade, etc.


The terminal strip is actually mounted upside down so most capacitors are hidden. The electrons don't seem to mind.


Additional testing notes:

** After testing the 5.5 KHz filter more with fast A/B in-out switching I have determined there is no effect on the positive peaks or asymmetrical character of the audio.  In addition, I am beginning to think that a 4.2 or 4.5 KHz filter would do a better job at guarding the "sacred" broadcast +- 5KHz bandwidth guideline.   Each filter starts really working a little past its design rolloff freq and the 5.5 filter is more of a guardian of 6.5KHz bandwidth.  So a 4.2 might be a good guard of 5 KHz during busy times and still give 5KHz of fidelity.


T

On Pic 3 of the station:  Notice I am running out of room to sit. My legs go where the plastic trash can is now.   I do NOT want any more rigs in the shack...  ;D

I EQ ahead of the compressor for that same reason. Although I use some pre-emphasis in the EQ, bandwidth limiting is the primary reason for my EQing. A passive LPF, as is being designed here, would be very useful as a first stage after the mic, or maybe better yet, just after the mic preamp. I'd try those two places first.

One of the problems encountered with a passive LPF is differing I/O impedances in the audio chain. Typically, modern audio processing gear have high input impedances and low output impedances. This is by design, and is called impedance bridging. There's no standard (that I'm aware of) so a filter may yield different performance when used with different pieces of equipment. This might negate the additional effort put into "tuning" the filter by careful selection of component value tolerances. I suppose that loading the filter with a terminating resistor would mitigate the problem by stabilizing the response into different loads, but at a cost of additional loss and reduced S/N ratio. This may not be as much a problem in radio communication as it would be in a studio recording environment. In the end, close enough is probably close enough.

Hi Don,

I changed the filter from the last position in the audio chain -  to right after the preamp feeding the EQ.   For whatever reason the bandwidth was about 1 KHz wider under normal speech and sss's. It may be because of a poorer impedance match for the filter, I dunno.   I have the EQ peaking out at 5 KHz and rolling off after that.   I am using 600 ohm terminating/loading resistors on the input and output of the filter.

I put the filter back at the end of the chain and like what I see again.   So I will continue testing with this lashup. I plan to fire up the 4X1 plate modulated rig and see how well behaved that is next.

T

Yeah, 1 kHz is a lot. That's more than I had expected.

If the cause is impedance source/load related it makes it difficult to produce a design that will work across the board. If you built ten of them and sent them to ten friends, they'd all have different results when used in their setups. This may be the bug that can't be built-out of a passive design.
As a test, you could put attenuators (perhaps 10dB) fore and aft of the filter, to reduce the loading effects and try it in different positions. I think that would confirm the theory. As a custom design, it works like a charm. Mass produced, maybe not so much.

If your equipment is generating harmonics that are being caught by the filter (with it in the last position) that's a whole 'nother story.


Don

Hi Don,

That's good info, thanks.

I can probably improve on this design by adding buffers, like emitter followers to the input and output easily enuff.  I'll ask Frank/GFZ if he can come up with a simple design in his software. I have room and terminals to add something.

In the meantime I'll make some measurements to see what the impedances are where it works best and doesn't.

My gut feeling is that the EQ is generating the sss's that need to be tamed. The filter works well after the EQ, at the end of the chain, but we'll see.   I like the idea of isolating this very effective passive filter anyway. Maybe as simple as a 2N2222  in and out.


Another add-on is I want to design and implement a filter for ssb DXing too. My main ssb exciter is about 3 KHz wide due to wider filters I added in the I.F.  The norm is more like 2.4 KHz when the band gets crowed, so I could use a 2.0 KHz filter or so to roll it off at about 2.4 KHz.   When the 75 DX window is busy and the noise floor is low this little difference of 600 Hz can make a big difference, especially when everyone is clustered close together and running a combination of sdr and 2.4 KHz rigs. Some even run 2.1 Collins type mechanical filters or equivalent.  When in Rome, use Roman filters... :-)


T

Hi Tom,

You may be out of room but you have double plus good transmitters.

I Like those filters very much, very nice job. I'm one who appreciates the results of straightforward analog circuits.

Have you found that the high resistance of the coils (e.g. the 33mH ones are 90 Ohms) reduces the sharpness/Q of the filter sections?

I ask because I'm curious about the response and the effect of inductor resistance on that kind of filter. I'm not sure how to interpret the waterfall displays dB-wise. Do you have spice curves on them with the chosen inductors?

Some time ago, I built a 500 Ohm in&out passive one for 3.5KHz. It's based on the "Bonadio" article circuits but with standard L values. It worked very well.

Low resistance coils (6.9 Ohms for 30mH) were used to get the sharpest cutoff diverse situations like handling up to 5 Watts input, and limit insertion loss to 6dB which was a bonus, and also sharpness of cutoff.

Models were also made in LTspice for -3dB@ these other frequencies of 2.9, 4.0, 4.45, 5.0, and 5.25KHz.
Slope is 3dB/58Hz on the 3.5KHz one and 3dB/110Hz on the 5.25KHz one, so it is comparable to the ones you designed (3dB/113Hz).

I built only the 3.5KHz one. It was put in the 500 Ohm line right before the input to the driver of the modulator. There was no effect on clarity, but I have not tested with any pre-emphasis either. Termination in and out seemed important.
3.5 KHz LPF: http://amfone.net/Amforum/index.php?topic=43634

PJ

Tom,

Yes, buffers! Maybe op amps at each end set to unity gain? They could be powered by +48V phantom voltage allowing placement just after the microphone. A nice simple desktop accessory. Perhaps a characteristic impedance for the filter could be selected that makes component selection easier too. This just keeps getting better...


Don

Some good ideas!

Pat:  That Bonadio filter using inductors down into the 10 ohm range sounds appealing for better Q and selectivity.  Yes, I am dealing wth 90 ohms + on some of these inductors. I didn't consider more than insertion loss at the time.  I read thru the thread and very interested.

Don:  Yes, an op amp at unity gain might work well too.  I notice Rick/KHK and Steve/QIX both have used them to isolate their audio filters.  

Well, my bandwidth problem is solved for now - I can run this passive filter in the meantime and relax in the catbird seat while we put together the next improved version.  I am very happy with the present rolloff and invisibility of it now - so it can only get better.  Tonight I try it on the 4-1000A plate modulated rig. The only difference in the set up is it will be driving Frank/GFZ's MOSFET driver board with a 2.6K input load. Probably no problem at all being a terminated 600 ohm filter.

My farther out goal is to get an analog filter to work "almost" as well as a DSP filter. And clone that design for the many bandwidth jobs I have.

Circuits welcome for evaluation...   

BTW, I will be testing out the SCAF portion of Rick/KHK's project soon. I have not closed the door on using this switched capacitor  approach and have verified my passive filter has no meaningful effect on the positive peaks, assymetry, etc.  It may just be a myth that these filters affect the freq below the cutoff adversely.

T

Here you go, third post down the list
http://amfone.net/Amforum/index.php?topic=42552.0

Don, 

I read over the thread.  Looks like you were the lead horse building the NPL / filter.  Are you still using it and did it eliminate the wavelet foldback problem?  I had interaction between the wavelets and NPL limiting in my trial using the older PDM board section - kinda a crossover distortion.   The wavelets appeared mixed with the NPL line... strange.   

So I now live with not naturally going over -99% negative. There is still plenty of pos peaks doing so. (+150%)       Maybe I shud try the latest version.    Did anyone make a board or wanna sell a working point to point?

T

Tom,

Per my on-the-air comment... the input of the filter "looks back" at the source. If that
is the output of the 528 compressor box, it's likely a 5532 or similar opamp at the output.
Nominally it is designed to "drive" a 600 ohm load rail-rail. So whatever that output Z
works out to be is like a shunt to ground. If you put a 600ohm resistor to ground at the
input, then the filter "sees" the output Z of the opamp + the 600ohm in parallel as the
input Z. Aka much lower than "600 ohms."

Probably won't change a Butterworth filter very much. Would change an eliptic filter.

That's why I said 600ohm resistor in series. This assumes ur driving it with a source that
is actually substantially lower Z, as is usually the case.

On the output side, just an emitter follower running enough current to drive the next
stage is all that is needed. The base or gate would get a 600ohm to ground, to give
the filter the requisite design load.

If there is some insertion loss, either an opamp or a transistor at the input or output
configured for a little gain is ideal.

Btw, it sounds really good with the filter...

           _-_-

Well, tonight I met up with a big group on 3885 AM and asked five of the experienced "refined ear" AMers what they thought about the 5.5 KHz filter performance.  (Tim/HLR, Bear/GCR, Tina/IA, Ron/ RON, Bob/KBW.... did I miss someone? Most of these guys have SDR spectrum scopes.

I did a transmission testing the filter in and out. As expected, without the filter I was 10-12 KHz wide per sideband. With the filter in all of the guys were reporting a well-defined 6.5KHz bandwidth. I could hear the excitement in their voices telling me it was working very FB.

The surprising subjective comments were that it sounded BETTER with the filter in. Some described it as sounding smoother and more focused in the frequencies that counted, rather than spreading out the extreme highs.  Not one had a negative comment with all very positive. Tron added that with his narrower receiver he could hear no difference in audio at all.. a good thing.

Ron/RON sent me a recording off his SDR. I could hear a slight difference in extreme highs.... they sounded basically good with the filter in or out. There WAS some more sparkle in the 'T' sounds with the filter out as expected.   So this is also good news because the highs were being chopped off nicely looking on the spec scope.   There is not a lot more voice intelligence or ambience above 6-7 KHz anyway but on quiet days, why not?.


I'd say the "Courtesy Filter" is a success.  (Thanks Don)     I will switch it into a different load tonight; into the GFZ MOSFET audio board to drive the 4-1000A plate modulated rig.
  

Bear: Thanks for taking the time today to run the tests on the air.  OK on the filter loading, resistor input in series and other comments. I will try some of the things you mentioned.

*** The more I listen to Ron's recording in +- 10 KHz bandwidth, the more I can hear the high cut when going from 11 KHz bandwith to 5.5 -6.5 KHz.  There is definitely some ambience gained without the filter. That is why there is a place for the 7.5 KHz filter on quiet days.


T

Wow, It seems that I've named a piece of JJ equipment. My place in AM history is now secure. ;)

Yep, that's a great honor, Don.  

We just received some additional help from the JJ marketing dept. The brilliant man (who wants to remain nameless) running the logos dept came up with this today;

The new APF "Courtesy", by Trojan:  

UPDATE:  I tried the Courtesy Filter  on the 4-1000A plate modulated rig.  I found the insertion loss was higher than I thought and had to add some more gain before the GFZ audio driver.

Running some voice audio tests I found the bandwidth to be exactly like earlier tests on the FT-1000D and 3-500Z amplifier. (about + -  6.5 KHz bandwidth)  The IMD appears to be superb. I ran some tone tests (with the filter in) and saw -45DB 3rd order !!!  when the 4X1 rig was run conservatively at 1/2 power.  That is better than the Ft-1000D and amplifier lash-up by about -10 to -12 DB.  This pleased  me greatly because the 4X1 is louder and needs to be cleaner. This means that the filter is not adversely affecting the IMD of the rig.

I tried adding in a low level NPL. (Negative peak limiter) and found when the modulation hit the preset -95%, the IMD went to hell. (inter-modulation distortion)   It was down to -15DB 3rd which is abominable. That word means it really sucked.  So I'm back to simplicity and find as long as I stay away from the baseline, the big rig has great fidelity, is narrow, clean,  and bad ass with great APF Courtesy to all on the band.  Thank you. Thank you very much. I'll be in Vegas all week.

T

Here is a picture of Tron when he first found out about the APF Courtesy:

Love the idea of the all-in-one AM processor. Looks like a brilliant piece of work.

But I hate to see all this narrowing up of formerly excellent-sounding signals. Maybe people with severe hearing loss or old unmodified stock receivers can't tell the difference, but I sure can.

My default audio bandwidth is 7.5 kHz. Anything less than that doesn't sound high fidelity to me. If I can do it without causing interference, I'll go 10 kHz.

Anything under 6 kHz starts to sound pretty bad to my ears; just getting off the air until things clear up starts to look attractive. Maybe as a stopgap, during a temporary interference situation, it's tolerable -- for a few minutes.

It's not hard with a good SDR (especially at 48 kHz sampling rate, where the filters are truly awesomely sharp) to keep on receiving 7 kHz audio or better from AM signals, even in crowded conditions. Just narrow up the sideband (as little as needed, of course) that's getting the most interference, and keep the clearer sideband as wide as you can. This works 75 per cent. or more of the time for me.

I'd rather listen to 7.5 kHz audio with a few sibilants of interference from a nearby station than listen to 5 kHz of audio interference-free. The narrower audio bandwidth is not only less natural -- it's also less intelligible. Contrary to what the ARRL and ham lore tell us, narrower audio is always less intelligible, all things being equal, not the reverse.

No one needs to explain crowding to me, or the rationale behind going narrower. I understand it completely. I just disagree with it. If AM operators would stop thinking that 5 kHz spacing is acceptable, and never do that, the problem would be 99 per cent. solved without ruining the AM fidelity that many have spent years achieving.

73,

Kevin, WB4AIO.

Hi Kevin,

That's the problem. AM stations do come in within 5 KHz spacing quite frequently. Maybe we should be addressing that problem more aggressively. We can either have an RF battle with them or tighten up our bandwidth or sign out. When it happens, I prefer to tighten up with a transmit filter. It's so easy to do.  This will probably never happen down on 3730, but on 3870 to 3885 there is often a crowd at prime times.  The AM rallies can be bedlam with no spacing at all.  People like to flock together.

The filter idea is all about having the option to run full-boat 7.5 KHz audio at anytime...or narrower when we choose. Flexibility.

T

Yes, Tom, I agree. It's good to have that option to go narrower when absolutely necessary; it's something I do myself when the situation demands it. I do think that we should run 7 to 10 kHz of audio normally, though. It sounds better, amazes anyone who takes the trouble to widen his receive passband and listen through a decent sound system, and potentially sparks interest in AM (or eSSB) experimentation. It's also a rebuke to that tired old lie that narrower is always better and more effective. It isn't.

73,

Kevin, WB4AIO.

Kevin,

It's good we have this discussion...  I consider you one of the sophisticated audio guys on AM, so I am listening closely...  ;)

In regards to the unique 5 KHz spacing problem between 3870 to 3885, I know, I know....we should all spread out beyond these walls, but few have for the last 40+ years.  I think it all begins when the first station starts up on 3880. This is a mistake. The next station then starts up on 3885 because they want to stay away from the 3890 ssb gang.  The next station is at 3875 to stay away from the 3870 and lower ssb groups. And there we have it... 5 KHz spacing and QRM.  (unless we switch in 5 KHz filters)  

The better alternative is for the first station to grab 3885, then the next station grabs 3878 and the next station hits 3872.  (or 3885 and 3875 for 10 KHz on quiet days)  This is more reasonable spacing and I've seen it work many times.  Especially when someone has sharp receiver skirts like you described using your SDR.

You make a good point about using a very hi-fi receiver that can actually hear + - 10 KHz.  I must admit before 2 weeks ago my receiver was limited to about + - 7.5 KHz so I paid little attention to higher audio. But with the new SDR I have going into a stereo and big speakers, I can hear the difference when someone really opens it up. The subtle breathing, lip and tongue noise are brighter and more articulated. And I can hear the difference in my own transmit monitor when I go from ~6.5 KHz to no filter at all. (+- 10 KHz)   It is worth it for sure, like a treat.

There are some stations that run 10 KHz audio at quiet times and it sounds great. As long as the highs terminate in wispy sibilance and are not real artifacts and splatter we are golden. There are a few stations on with some splatter mixed in and I have hesitated to say anything to avoid being a complainer. I sometimes think they will figure this out themselves. I have tried hard to increase AMer awareness of running IMD tests, triangle / sinewave tests and being self-aware of our own signals.   With today's cheap SDR spectrum scopes and a simple 2-tone test, we have all the tools to monitor ourselves. This was not available in the past. The SDR competition to older rigs is intense and has made a new standard. We need to be on top of it all with our older rigs. My whole station consists of five analog, plate modulated and linear amps so I'm in that boat.

There is nothing worse than another AMer telling us we are splattering all over the band... nothing worse... :o   Lately I just love the feeling of confidence - getting on with the new filter and watching the SDR spectrum and waterfall scope with my sibilance just kissing the edge of an adjacent QSO knowing it is not bothering them. I just hate the feeling in the past when I would unkey and hear the adjacent QSO AMer say to his friend, "there is someone up the band taking you out OM."  Being able to forget about myself and focus on my QSO is what it's about. It's akin to going to a party dressed like a bum. We focus on ourselves and worry about what we look like. But wear a nice suit and we forget about ourselves and just have fun.

For the most part AMers are very accommodating to one another. I think it would be great if AMers in general were looked upon by the ham world as being considerate and aware operators that were always willing to slide up or down the band a  wee bit when congested, run less bandwidth when the conditions suggested and generally did not dig in our heels when a little QRM starts up. I look at it like giving up my seat on the subway to anyone who asked. But it's rare, and it's more about having an easy-going "turf" attitude I am talking about.

I've been having more fun in the last two months then ever before on the air and working on new projects.  It's all about the friends we have made over the years and getting excited about improving our stations.  Helping newer Amers who we may never meet through this AMFone BB is a legacy I hope to leave for years to come.. Just think if we had this great BB resource of like-minded AMers years ago, Kevin...

Take care, OM.

T

We're making more progress...

We've decided to advance the  passive 10th order Butterworth design.  It is working well.   The only change is to add buffer stages to the input and output making it an active Butterworth filter. So don't build anything yet.

Frank / GFZ is modeling simple 2N2222 transistor stages that requires only 13.8VDC to work.   This will permit us to put the filter anywhere in the audio chain without worrying about disturbing the filter's design characteristics.

I have the parts and hope to test it soon.

John/ JSW has already made good progress on a professional PC board holding three different filters.  We expect to use the same buffers for all three to reduce parts count..

I will be testing Rick/KHK's SCAF (switched capacitor audio filter)  in a few days too, so stand by for more...

T

Thank you, Tom. I mess around with audio a lot (my latest fun and learning experience is using VST plugins in my all-digital audio chain) but am not 100 per cent. satisfied with my signal. You are one of the Renaissance men of radio in my opinion, far beyond just audio. Been following your career since 1973, sonny!




True.

The AM window interference problem is complex, partly technical and partly psychological.

Some people live and listen in a bubble in which only "their gang" matters and those people "over there" in the other part of the country don't matter (even though they're propagating in just fine), so parking 5 kHz away from them and narrowing the (probably stock and horribly narrow) receiver even more is justified -- because, I suppose, tuning to another part of the band is too much trouble. Sometimes this psychological bubble is so impregnable that they'll park ONE or even ZERO kHz away and just fire up.

Then there's the very real technical bubble -- the tiny high-angle reception bubble created by people running inverted vees at 20 feet and the like, and exacerbated by high local noise levels. Such folks have a hard time hearing anything other than NVIS semi-locals and falsely assume that others 500-plus miles away can't hear them -- wrong! But they plop 5 kHz away anyway, and evidently think that's perfectly fine.

In some ways, channelized operation would be better -- but, then again, amateur radio is experimental and wild and free, so let everyone be free, even with their wobbly VFOs, bizarre ideas of how close one can get to another QSO, distorted audio in which the IMD-hash is almost as strong as the desired audio, 300-3000 Hz audio that was a stupid idea to begin with, QSOs in which participants are spread out over several kHz, and all the other things I don't like too much. That freedom to do almost anything is also what allows _us_ to experiment with high fidelity wideband audio, AM, DSB, eSSB, and all the things I DO like very much. (If we did have channelized operation, I would want AM/DSB channels to be spaced 16 kHz, SSB channels 8 kHz. This would allow 7.5 kHz audio with a small guard band. Wow, would that be nice!)

To totally avoid interference with 5 kHz spacing, we'd need to chop our audio off at 2.5 kHz -- 2 kHz to account for BA drifting. Sheesh! Might as well run one of those awful 2.1 kHz SSB rigs -- or, better yet, just sign off and spend more time with my wife and my collie.



Studies have shown that limiting speech bandwidth to less than 8 kHz reduces intelligibility. The old canard that "3 kHz is more intelligible" was nonsense from the beginning. Three kHz was chosen because it's as narrow as you can get without _severely_ reducing intelligibility, and the telephone engineers that came up with it (actually, they chose 3.5 kHz with a 500 Hz guard band) were counting beans (dollars) to save on landline equalizer coils and, later, get as many channels as possible in their carrier multiplex systems. New phones with HD Voice (finally!) now open that up to more than 7 kHz and it's _astoundingly_ better. The ham radio establishment of the 1950s, unfortunately, followed their lead, and eventually even refused advertising space to transmitters that had broadcast quality, of which there were quite a few before the moronic "narrow is better" idea took hold.

Then there are also the subtleties of the human voice that come across in wideband audio, but are lost with "communications quality." These have a role in conveying tones of voice and psychological cues of personality and expression and shades of meaning that should not be disregarded. Another factor is the aesthetic factor, the happiness that beautiful audio can bring -- just as valid a human goal as a beautiful building or oil painting or sculpture.


I couldn't agree more. As W1AEX says, in many ways we are living in the best age of amateur radio that has ever existed.

73,

Kevin, WB4AIO.

The bandwidth issue needs to be looked at from microphone to speaker while considering the factors in between.

AM is unique in that everything from 3kc to 10kc frequency response (not bandwidth) in both transmission and reception has its place. When the bands are quiet and lightly occupied, nothing sounds smoother than a cleanly modulated broadband signal. But since this is not always the case, particularly after the sun sets, ops are faced with a decision.

I agree with Tom about narrowing up when things get crowded, but not just to be neighborly. Even if a transmission remains "Hi-Fi", it's unlikely that other ops on frequency are going to leave their receivers set wide and tolerate the artifacts from adjacent signals. This leaves the question about being "wide" when nobody can appreciate it.

Also, concentrating modulated energy into a narrower spectrum contributes to intelligibility, a point not lost on equipment manufacturers who want customers raving about piercing pileups and scoring difficult DX contacts. That's not going to happen with 20-20kc audio.

SEE? YOU AMers NEED TO TIGHTEN IT UP!!!

That's the popular argument, but it makes no sense. AM's strength is not in DXing. AM is the "QSO Mode" where ops don't need to constantly twiddle with a RIT control each time someone keys up and we can recognize each other at hamfests by voice.

Bottom line: Hi-Fi AM is fantastic and it's the hallmark of the mode, but it's only good if others can appreciate it too. There's no need to be a part of adjacent QSOs if nobody can hear those highs.

Side bar: For Pete's sake (and everyone else's) if you have to operate in the 3875-3880-3885 segment of 75M, 3880 is NOT the first choice, it's the last. Think of it as the "middle seat" on an aircraft. Choose the window or aisle first. Then again, there is no "AM Window" any more so spread it out folks...

Very good on all....

Lots of interesting points regarding the psychological and technical features of the AM Window, audio bandwidths, etc.  An important point mentioned by Kevin is that if two AM stations are spaced 5 KHz apart, then both need to run +- 2.5 Khz audio JUST to barely keep out of each others' way. That is an eye-opener to guys who get within 5 KHz and then complain of splattering neighbors....  That's why broadcast stations are spaced 10 KHz apart and run +- 5KHz audio.

I've been having a lot of fun lately with the SDR spectrum and waterfall scopes as I get involved in a QSO.  I've noticed a few things as I get better at it...  (I'm pretty good at pattern recognition due to my job.)  I have a new appreciation for clean, well contained and managed AM signals, no matter the bandwidth..

Yesterday, an old friend Dave K2IJY came into our QSO using a BC transmitter and was very loud. I noticed his unique spectrum of exactly + - 10KHz. It was a beautiful spectrum like it was highly intentional and showed an "intelligent being" was controlling it. The audio was dense thru-out the full spectrum and cut at 10 KHz like a crewcut. I asked him how was he doing this?? He said he was using an analog Orban processor, something like a 9xx?  IIRC.  He knew what he was doing and it showed.

I later heard a 100 watt stock DX-100 and could see it was struggling. The bandwidth was somewhat random... wide, then narrow then some splatter, then not. It was not an "intelligent alien" controlling his signal...  ;D

An analogy is when there's a thunderstorm and a lightning bolt strikes. The spectrum will show many megacycles of garbage, no rhyme or reason.  Then compare this to the commercial AM broadcast band and there is absolute order... +-5 KHz with highly intelligent data. And we  clearly see it stand out on the spectrum scope and waterfall.

I was looking at two stations - one talking on essb the other day. One was 2.4 KHz wide and the other was 4 KHz. Both had sharp skirts. It was perfectly easy to see the difference. The 4 KHz station sounded like a million bux. The 2.4 KHz ssb station sounded like a regular ricebox with a stock ssb filter. Who had the best contained and controlled signal? I suppose both did but the 4 KHz guy was stretching himself to go the extra mile to sound spectacular and he did.

The point I'm making is years ago we got on AM and hoped for the best. We hoped our signals were clean and orderly. We depended on the LACK of complaints to tell us we were OK. Except for an oscilloscope that's all I had to go on.

Nowadays with the spectrum tools we have, we can take control of our transmitters whether they are 60 year old boatanchors or newer SDR rigs. I just love doing the detective work and training myself to pick out what a station is doing with their rig by the order or disorder of their signals.  This has nothing to do with absolute bandwidth, rather how are they managing their signals to achieve the bandwidth they choose to use and keeping it clean.

In general I'd say the majority of the hi-fi experienced AM stations who operate between 3870 to 3890 are running about +-  6.5 KHz audio and most signals are well contained. They are very self-aware. Then there are a few who are not sure what bandwidth their rigs are transmitting. But nowadays, when a rig breaks and starts splattering, it ain't long before they are informed.

All good.

T

Today I finished the 5.5 KHz prototype.  I added a 2N2222 emitter follower at the input and a 2N2222 buffer at the output of the 10th order Butterworth passive LP filter, making it an active filter.

Frank did the modeling changes as posted below.

It worked right away, with no mistakes - which is unusual for me... ;)   I was pleased with the initial bench test. It was perfectly flat from 2 cycles up to about 4.8 KHz. It then started dropping off and got faster beyond 5.5 KHz and was way down at  6.5 KHz.  Hardly a residual trace at 7.2 KHz.

I noticed a few improvements as a result of the filter being better isolated from the external world. The previous passive filter had a slight ripple hesitation at about 6.7 KHz as I reported last week. This slight ripple/hesitation was gone.  The flatness and fall of the skirt was near perfect in shape. I noticed on the bench the drop off had moved out about 500 Hz farther than before, but on the air, it appeared more behaved and very sharp with the spectrum waterfall cutoff of 6.5KHz.  This is a 5.5 KHz filter, but the brick wall effect is around 6.5 KHz in practice it seems.

I ran it into a dummy load.  Without the filter it was the usual + - 10-12 KHz. The filter in-line actually looks cleaner and sharper than before at a razor 6.5 KHz.  Another thing I noticed is the highs seem slightly cleaner in the monitor.. they sounded similar in quality when the filter is off line.  The filter is near invisible below cutoff as far as I can tell.

Butterworth filters have a great reputation for several reasons. I am glad I went this route.  We are looking at doing 3 or 4 filters with buffers on one board. At the present moment I have plans for a 2.5 KHz ssb filter for DXing when very crowed, a 4.0 for AM or ssb  (+- 5KHz)     a 5.1 (+- 6.2 KHZ)  a 7.0 (+- 8.5 KHz)  and a 10 KHz (+- 12 KHz)

The option will be either an ssb filter or the 10KHz since only four filters will fit on the board.  I plan on the ssb filter rather than the 10 KHz. We will have to see what the final decision is for the number of filters in total..

Oh, another benefit is the slight residual 120Hz hum I had using the passive filter is totally gone. The audio line did not like the passive filter alone, but with it isolated, it appears cleaner, less noisy and works like a champ with no observable ripple or ringing.

It uses 13.8 DC now and 600 ohms in and out. Now on to the good looking PCB version.  Frank wants to normalize the filters so that they have the same gain for ease of fast A/B switching in and out without readjusting the audio gain. Oughta be a slick box when finished.

One of my goals with this project is to get the 4.0 filter to be brickwall at +- 5.0 KHz, just like a BC station... ;)  I will add on some more poles just to see how far I can take it at 5.0 KHz.

T

 
The finished 5.5 KHz prototype:

Very interesting points on order versus disorder. Disorder is entropy, the most powerful force in the Universe so far as we know. And Life in general, and we in particular, are entropy-fighting creatures. So, by ordering our transmitted intelligence and energy well, we are serving the Life Force. Bet George Bernard Shaw never thought of that!

One correction, though: It's mostly in a few big radio markets, especially in the Great Eastern Megalopolis, that _some_ AM broadcasters have adopted the execrable +-5 kHz audio standard, something that is not required and that they never should have done (the REAL AM interference problem is not from adjacents or second adjacents -- it's from a zillion stations co-channel on every channel, and reducing bandwidth helps not at all with that). The vast majority of AM broadcast stations (and ALL of them in my area) are, thank goodness, +-10 kHz.

73,

Kevin, WB4AIO.

Understood. But even if you feel you cannot open up your receiver much, and find it hard to believe that your QSO partner could either, there may be other people listening in who can. I sometimes find that an op has made a choice to use a narrow transmit bandwidth when I could have easily heard 7 or 10 kHz from him in my location. Disappointment!



Actually, concentrating your energy into a narrower passband does improve the signal to noise ratio (a LITTLE), but that does not always translate into increased intelligibility. One day, I'll post one of the studies here that show wider audio is more intelligible than narrow audio.

I agree 100 per cent. about AM (and I would add eSSB) being the QSO mode, with the sound tailored for enjoyability, naturalness, and conveying subtleties of character, meaning, and feeling in the sound.

I think about it this way: If you cut out all your bass, and heavily clip your audio, it will probably help your intelligibility when your signal is barely detectable, below the noise. So, for people who seek out such situations, such as snagging the rare one using a manpack and a whip on the far side of the Himalayas, that kind of audio is preferred.

But 95 per cent. of my operating involves nice, healthy signal-to-noise ratios where space shuttle audio does not help at all -- in fact it hurts in more ways than one.

Many stations run space shuttle audio all the time. But if I do that, and cater to the 5 per cent. of stations on super-weak paths just so they can understand half of my words if I repeat them three times, then I am simultaneously torturing 95 per cent. of the people I talk to by making them listen to wretched audio -- for no reason at all.


Agreed! I'd go even further. There are only two seats, aisle and window. 3880 is sitting on the arm rests between the seats.

73,

Kevin, WB4AIO.

Yesterday I received  Rick/W8KHK's SCAF, Switched Capacitor Audio Filter,   (contained on the Max board)  and I started testing it on the bench today. It initially had some residual clock feedthrough crud, but a simple low pass filter fixed that and it is now clean. This is an older version board and that has been fixed and Rick has been on top of it..

It swept just like the Butterworth on the bench. Very close in smooth response and sharp skirts.  After all, two Maxim SCAF chips in cascade is pretty bad-ass.

I then hooked the SCAF up to the rig and ran some more tests. It was easy to adjust quickly to the desired filter freq and it sounded as good as the Butterworth in the monitor.   I ran it on ssb at 2.1 and worked fine.  I tried it on AM at +- 4.0, 5.0 and 8 KHz and I was able to adjust it on the fly to give perfect skirts at those freqs. I figured +- 8 KHz is as high as I will ever run my bandwidth. It did sound sweet at 8 KHz!  And it sounded like crap on ssb at 2.1 KHz due to the bandwidth... ;)   The difference between my FT-1000D on ssb running 3KHz and 2.1 is unbelievable, so I am going to stay with 3.0 on ssb..

For my use, the easiest way to quickly adjust the filter wud be to modify the board and mount three larger 1-turn pots on the front panel with a switch for narrow, medium and wide. This combo would give continuous control from ~ 2-12 KHz. (or more if needed)

I am impressed but first I need to run some IMD tests with the rig and check it more carefully against the Butterworth. Also, I want some on air A/B tests to see if Tron hears any difference or artifacts.

 On the air while watching the SDR spectrum analyzer and waterfall, I notice it is definately in better control of the skirts than the Butterworth. There are no rouge spikes when hitting sss’s.  It actually resembles SDR DSP on the waterfall when using the FT-1000D on hi-fi AM which is only -30DB 3rd order.

We shall see.  Our Butterworth is a fine filter and a lot of work, but it will be a simple matter of which one works and sounds the best in the end.

Rick has been very helpful getting me up to speed on the filter - so it went very smoothly with no errors so far. No smoke or board arc-overs so far...

Oh, one last thing...  FWIW, when the filter was adjusted to 3.5 KHz or lower, the mic polarity switch needed to be flipped 180 degrees to give the best positive peaks.  Above 4.0 Khz, the mic polarity remained the same as when the filter is out of the path. My Butterworth is fixed on 5.5 KHz, so I cannot compare the two on this basis. No big deal - just wanted to mention it.  My voice is different from yours, so you may or may not see the same thing. I do not use an all-pass filter for symetrical audio.


More soon.

T

The tandem Maxim 295 chips are Butterworth filters, known for their sharp cut-off. The Butterworth downside is overshoot which is why they are located prior to the clipper stage. We tried Maxim 296 chips with Bessel filters to address the overshoot but the sacrifice in slope wasn't worth it so we reverted back to the 295s.

The MAX processor has abundant customization notes. Most might build and use it as-is, but the AM crowd in particular are experimenters and like things their way.

Plans for purple chaser lights and chrome hubs are in the works.


This happens because vocal asymmetry is not consistent across the vocal range. It's quite common for the dominant polarity of lower frequencies to conflict with upper frequencies. Using a low-cut filter or limiting higher frequencies will both reduce the opposing polarities. The use of an all-pass filter eliminates this issue.

Split highs and lows to separate channels, flip the phase of one channel and recombine. Positive peaks galore!

Butterworth filters best characteristic is passband flatness. They are called maximally flat filters. If you want a sharp cutoff, go with a Chebychev. If you want good impulse response go with a Bessel. I'd go with a Bessel to avoid overshoot. A lowpass filter at the tail end of the processing chain can put back in peaks that you've tried so hard to remove.

This nice aluminum box has been sitting around for years waiting for a job.  I usually don't waste time. I received the old version 3 MAX with the SCAF MAXIM chips and got it working today.  I asked Rick about my adding filter frequency control pots and toggle switches on my front panel rather than trimmers on the board. He thought it was a cool idea and told me to solder right to the header pins and add a switch and pot in series for the three trimmers.. Easy enuff.  He warned me about potential instability with RF, but all is stable which is a sign of quality design.

I first did a lot of tests before the mods to make sure everything was stable in an RF field and it was. The clock can be sensitive to RF so I kept the chip and connections close to he front panel. The pot wires are the brown Teflon wires you see at the front of the cabinet.  This is obviously a custom SCAF implementation and as Clark suggested, shows the versatility and modification potential of the overall MAX unit.  The board has in and out headers all around and you can add or subtract whatever you need. In my case the MAXIM SCAF and clock are activated.

I added in a LP filter to kill any remaining clock crud and it was completely gone.  I will drill some heat vent holes tmw.

I ran some tests with the panel pots mods. Bottom line is the SCAF roll off is about as sharp as the discrete Butterworth, but the ease of changing filter frequencies is amazing. With a flip of one of the three toggle switches I can go to preset filter frequencies of +- 4.5, 6.0 KHz or 8 KHz.  (or whatever you want preset. The three knobs on the front panel allow me to overlap the three filters and tune them up or down in bandwidth.  So I have it set right now to cover CONTINUOSLY +- 4.0 to +- 8.0 Khz transmit bandwidth. What a difference between 4.0  and 8.0 KHz in audio quality!

The isolation is good - I hear nor see any hum when the mic is unplugged and on the air.... none at all. The 120VAC isolation power transformer did the trick for the 60 Hz hum and the MAXIM chip appears to have good buffering internally according to Rick. So no 120 VAC hum either.

I have the knobs set at the middle top as the presets - and can swing the filter freq monkey back and forth just by eyeballing the knob position. I will probably label the knob points to signify frequency.
 
Now that the Butterworth and SCAF are both in a box with plugs, I get to try them in cascade or compare them separately for tests. That will be fun.  I don't hear or see any issues at all with the SCAF so far. The SCAF parts count is greatly reduced compared to building four Butterworth filters.

I'd like to push the envelope and try a 3 or 4 MAXIM chip SCAF in cascade experiment.  There is surely a limit that I want to explore to see how close to DSP skirts I can get. Right now the RF bandwidth on the waterfall is pretty sharp and probably limited by the small IMD of the amplifiers.


Bottom line is it appears that the Max designers, Rick and Clark did a great job getting the filter to function well.  It oughta be an interesting box in the future if this is any indication.

T


Latest SCAF implementation:
  

I tried something I always wanted to do.  

First of all the two filters are about the same in performance when run alone.

I wanted to test the discrete Butterworth filter with the SCAF filter in cascade.

On the bench using a sig gen and scope - the discrete filter working alone,  I swept  the filter from 5.0 KHz to 7.4 KHz and got a certain DB drop.  I then optimized the SCAF MAXIM filter for the same frequency coverage. Both in cascade,  I swept them over the same frequency and compared results of the two skirts.

The results were quite good!  Dropping the same skirt depth, it  finished at 6.5 KHz.  IE the same skirt drop was sharper by almost 1 KHz.  This is significant.
The roll off looks closer to a DSP skirt. Later I will try the same lashup on the air and give it a good audio listen.

This tells me there is more to go in performance. I want to try four MAXIM SCAF chips in series next.   The Max board will let me insert an outboard pair on a perf board.

A general rule I've observed: The lower the cut off frequency of the filter, (3.5 KHz)  the more the asymmetry is affected in the audio lows.  When using high cutoffs, the lows asymmetry is hardly affected. (6.5 KHz)

T


The two filters in cascade:

Tom,

IF the design impedance for the "passive" filter is 600 ohms, you don't have a 600 ohm load on either end
that I could see. Maybe if you set it for 600 ohms the frequencies will magically fall into place? Just an idea.

                _-_-

PS, if you want more of a brickwall effect, turn one section into an elliptic filter by putting a cap
across the inductor. Yes, it needs to be simulated and tweaked in situ to get the proper "dip" at
the right frequency.

Hi Bear,

The discrete Butterworth now has 2N2222 buffers going in and out so the loads are not as important as when passive before.  The two in cascade are really unreal.

But before I fool around any further, I want to see what FOUR MAXIM 295 SCAF chips look like in cascade.   We could wrap this project up quickly and reduce the parts count by a barrel-full ... :-)

Before we spend more time on it, has anyone modeled or tried four chips?

T
 

you're going fast enuf.....hard to keep up....glad you're having fun.....Steve

BTW, my idea that lowering the filter cutoff below 3.5 KHz is affecting the lows or phase shift affecting the lows adversely seems invalid. I swept the filter cutoff up and down from 2 KHz to 8KHz and could NOT see any change in the positive peaks or the asymmetry.  So looks like filters in general are not the villains we make them out to be.


I could not hear any difference between the SCAF and Butterworth during tests tonight into the dummy load.

We will be needing a buffer for the in /out SCAF, however. There is a little 120 VAC hum. It was the same with the passive filter too and the 2N2222s buffers fixed it.  Rick already uses buffers on the MAX board so it is not an issue for the MAX.   Maybe I can use those myself. I was just trying to get by without them... a form of laziness.


T

You were originally correct. Lowering the cutoff frequency will likely affect asymmetry because you are filtering out the predominant polarity vocal frequencies leaving the opposing lower frequencies alone. The opposite will be true when switching in a low-cut filter. In this case, the opposing polarity lower vocal frequencies will be filtered out and your positive polarity should increase significantly.

None of this will be apparent when using a sinewave of equal positive and negative amplitude. This simple circuit will demonstrate this effect. You'll never think of asymmetry the same way again:

http://www.internetwork.com/radio/n1bcg/RMS_Ratio_Monitor.htm

(http://www.internetwork.com/radio/images/RMS_Ratio_Monitor.jpg)

Thanks for the info and circuit, Clark -

What you say makes perfect logical sense to me. But what I don't understand is if I sweep the filter frequency up and down from 2 KHz to 8 KHz, I see no evidence of the transmitted waveform changing. I use my own  asymmetrical voice and sock an extended yaaaaallo and see no change at all... ?


While on the subject:

I made some careful two-tone IMD and THD tests tonight with both filters on the rig.  My normal hi-fi FT-1000D IMD is about -33 DB 3rd order, and the harmonic distortion is about -40DB.

I tried both filters on AM and SSB and could not see any signs whatsoever of IMD or THD changing with the filters in or out. They are apparently cleaner than the power amplifiers in the rig and below my measuring floor, which is understandable.

I tried many combinations of signal change and filter frequency changes with good, smooth results with both.  No signs of clock crud or strange signals at all. This is with the longer pot leads too.

T

It actually takes a bit of practice to isolate the opposing vocal frequencies since normal speech or even a hearty yaaaaallo involve a complex mix of sounds. That circuit, or the level/polarity indicator on the MAX board, makes it easier to visualize and experiment. Everyone is different, but I've found that "aaaaaahhhhh" (lower frequencies) and "eeeeeeeee" (higher frequencies) generate completely opposite polarities for most folks. This is best done off the air of course unless you're really comfortable with yourself and the inevitable "What the hell are you doing?" that you'll hear when you unkey.

The main point is that when someone's mic is "phased correctly" it's really just phased for the majority of their speech.

This video demonstrates the effect:

http://dw.convertfiles.com/files/0490507001582771942/6af520b2-1c73-4e42-a89d-e6d79421b709.mp4

Almost finished with this phase of the job - a good working filter, sharp as a razor, adjustable on the fly, no hum and a great waveform!

The last problem was some 120AC hum in the audio line but Rick told me how to borrow a buffer stage on the MAX board to isolate the input. Worked like a charm and there is absolutely no 60 Hz or 120 Hz on my signal.


What is pictured below is the Max board Maxim 295 SCAF filter  set ON THE FLY in real time to  +- 4.5 KHz,  +-6KHz and +-8 KHz  sssss's of my voice.  I can switch in narrow, medium and wide filtering and then vary its bandwidth even more with the panel pots. Overlapping coverage from 2.1 KHz to 10KHz.  Notice how sharp the skirts are. They are very close to this sharpness in normal voice speech, thought not as densely packed.

The only critique is I see a faint bit of blue wisp indicating very low energy high freqs in the form of ssss's leaking out when the waterfall sensitivity is increased. Most of this stuff is invisible under normal band conditions, but on a BIG local condition day running the big mawl, I'm sure some sibilance will be heard past the obvious filter limits.


This is why the next step is to try four Maxim chips in series rather than the present two. Rick offered to send me some parts so I can experiment away. As said, he has been very helpful on this project.  I was telling him that my "technical highlight" of the year was seeing my transmit AM signal (on the Butterworth filter) waterfall with tight sidebands, almost like an SDR rig. I've always had issues containing my sssss highs, but for the first time I have total control.  Amazing technology.

Now lets see how far we can push these chips in cascade.  The "QUAD SCAFs"  as coined by Rick.

T


Below -  *switched on the fly, real time*   4.5KHz, 6Khz and 8KHz transmit audio bandwidth on the 4-1000A plate modulated rig:

Yeah, sorry. This forum doesn't allow videos to be uploaded, so I converted it to mp4 and provided a link. The problem is that these converter links expire.

The screenshots below show the polarity and level indicator on the MAX processor boards. The red LED shows predominant negative polarity, the green LED shows predominant positive polarity, and the yellow LED is a normal level indicator. The righthand red LED flashes on with peak levels.

Symmetric tone (Red and Green LEDs are lit):

(http://www.internetwork.com/MAX/images/Polarity-Sym.jpg)

Positive polarity with "eeeeee" sound (Green LED is lit):

(http://www.internetwork.com/MAX/images/Polarity-Pos.jpg)

Negative polarity with "aaaaahhhh" sound (Red LED is lit):

(http://www.internetwork.com/MAX/images/Polarity-Neg.jpg)

With normal complex speech, the red and green polarity LEDs flicker alternately but the green is on most of the time, indicating the predominant polarity. The negative polarity LED is usually blue, but for some reason my phone shows that color flushed, so I replaced it with red for more clear display.

"...normal complex speech..."  ???

Yes dear friends,

When you want that syrupy smooth audio, just open the top, and pour it on!

ON AIR:

I got some filter audio reports from the locals on 3885 AM today. I switched filters, switched bandwidths, fished for SDR spectrum and ear reports.  Here's what I came up with...

Most said it sounded FB using either filter. W1IA said the bandwidth was well contained on the spec anayzer and sounded great. Jay/ N3WWL liked what he heard. Bear had me try the SCAF filter at 8KHz and then out of line. He said he could hear some faint spectral artifacts with the SCAF filter in the extreme highs.  I then tried the discrete Butterworth filter and he thought it was cleaner, but was not sure.

The +- 2.5 KHz bandwidth sounded too restrictive, while most said they could live with +-4 KHz no problem.  +-6 KHz was a good compromise while others thought there was a big difference in audio quality going to 8 KHz. Much of the opinions depended much on the receiver bandwidth, of course.

I plan to get more listening reports over the next few days to get a consensus. I didn't hear the Tron on today who is my main man for audio reports.


SQUAREWAVE BENCH TESTS:

Bear later suggested I run a square wave thru the two filters to compare them.  I hooked the dual trace scope to both filters and fed them together to get an overlay. The old scope actually sync'd.  Amazingly, except for the low freq tilt,  they were almost identical from about 80 Hz up until about 3000 Hz when the square wave turns into a sine wave as the filter starts to affect the signal at 5.5 KHz.  Again, both filters tracked closely.  

The first two pictures below were taken at 500 Hz using a squarewave.   However, the third picture taken at 20 Hz shows the SCAF  has a tilt slope while the discrete is a pretty good squarewave. I would attribute this to using smaller coupling caps in the SCAF. I am using 150 uF couplings in the discrete, simply because I had them. I think Rick said he was using 10 uF coupling. Bigger caps would even out the response I think.... assuming that's what they want to do with Max.

The swept response of the discrete Butterworth goes down to 2 Hz.  They both have close to the same ripple and slope but there is a slight bit more overshoot and ringing in the SCAF filter, but probably not of concern..

I'm very pleased so far with both filter's transparency and peformance.

T

Below: The two filters independently compared at 500 Hz, 500 Hz and 20 Hz  - (TWO TRACES TOGETHER):

Here is a clearer view of the SCAF Filter  (top) and the Discrete Butterworth (bottom) using a 300 Hz square wave signal.

T



Below: Notice the SCAF has a little more tilt and the ripple and overshoot is slightly higher.  My guess is once the tilt is fixed it will be hard to hear the difference in the tiny ripple vs: the added versatility of the SCAF.  Opinions?

The overshoot amplitude and the number of cycles of ringing does tell a story about
the sound. While not unexpected, the fact that the two filters differ on overshoot indicates
that they may sound different.

Ideally there would be little or no overshoot.

In the abstract, the leading edge is a high frequency, and the goal of the filter is to
roll off high frequencies. If the filter could do this perfectly, only the rise time of that
leading edge and the squareness of the corner would be rounded (here it shows
overshoot and ringing).

I don't recall if it is possible for such a high order filter to not exhibit the ringing,
and/or if it is possible to control/damp it. Would have to pull out a text...

I thought they did sound somewhat different.

Subjectively on my R-390A @16kHz all of the filter widths sounded different, as did
both filter circuits...

It's certainly an interesting experiment!

 

I changed the coupling caps from 1 uF to 150 uF and the tilt is almost gone and the filter response looks like a square wave down to 10 Hz or so.  The ripple is smaller.  Big improvement.


I'm going thru voice tests now but this should bring back the lows and maybe make the performance more transparant.

With this change I think we can now tolerate a bit more ripple with four chips in cascade.

* BTW, I just found that part of the slight tilt is caused by the signal generator and the scope.  So this SCAF filter is really performing better than pictured.

T


Both Pics show squarewave at 16 Hz -   SCAF Filter:

After more on-air reports and bench tests, I have decided to go with the MAXIM 295 SCAF chip approach to the LP filter. It was a tough decision because  both filters had advantages and disadvantages. I am not convinced that the SCAF has any artifacts enough to hear. It's such a close toss up in performance. I think going with the SCAF is best. It has the best skirts and has more control of the frequency changes.  The tilt has been optimized and the ripples are about the same in both filters. I have no transmit hum at all using either now that buffers have been used.

I ended up with just ONE panel pot that controls the overall filter cutoff frequency. With four pre-marked dots I can set the filter quickly to  +- 3 KHz,  +-4.5 KHz,  +-6.5 KHz and +-8 KHz.  These settings are where the haircut starts -  where the actual sidebands end.

After a week of actual on-air practice I am able to see band conditions on the spectrum waterfall and set the appropriate bandwidth in seconds.  I can see if my sidebands are hitting the adjacent QSO or are clear. In fact today I was demonstrating the +- 8KHz position to Bruce/W2XR and I could see it was too wide and hitting the close QSO down the band. So I quickly switched back to 6.5 KHz and I was clear. This is a new double-check way of operating for me and I feel more self-aware of what's happening on the band around me.

The only thing left is to build up the extra two Maxim 295 SCAF chips and add them in in cascade, doubling the size of the filter.  I still see a slight bit of energy that leaks out past the skirts... everyone has it due to IMD or simply not a sharp enough audio roll off. If the four chips kill this extra fuzz, the extra ripple added to the filter may be worth it.  I should know sometime next week.

*** As a perfectionist failsafe, I have a filter in/out switch that will completely take the filter out of the system for those days I feel like letting the freak flag fly at at +-10KHz, using the natural bandwidth of the transmitter I'm using.  

All in all, I'd say the W8KHK / N1BCG Max implementation and MAXIM SCAF chip, when set up right, is a good design and will do the job quite transparently. My thanks to  Rick/KHK for his email suggestions and parts, Frank/GFZ for related filter modeling and the many guys on the air who took the time to run tests with me from time to time.

T

Tom, after you finish testing with two additional Maxim Max295 Butterworth filter chips in the chain, you might like to substitute a quad of Max296 Bessel filter chips and compare the results.  The cutoff slope is not quite as steep with the Bessel chips, but that may no longer be an issue with four in cascade.  

The  Bessel devices do not have the overshoot issues exhibited by the Butterworth, so impulse performance (as demonstrated with your square wave test) might be significantly cleaner.  I will provide four 296's if you think you would like to try them.

Depending upon the results of your testing, we may consider adding two more Maxim devices to the artwork before the next rev PCB of the Max Processor goes to fab.  

Your efforts testing, as well as making suggestions for improvements, are very much appreciated!

Sounds good Rick!  Glad to help.

I think the 296 chip swap test - sacrificing some skirt steepness for the chance of a cleaner filter response is well worth a try.

It makes sense and we could end up with a really sharp squarewave with little to no ripple. Who could ask for more?

I continue to test the two MAXIM 295 chips at various bandwidths on the air and the reports are good.  Tron was listening to the filter and CRL PMC-300A clipper tonight and said he liked the sound very much.

Tim was running hi-fi audio and his articulation was quite pronounced. I could hear the subtle sounds of "THHHH" like in the word "The" and the sounds of his lips, tongue and teeth. It's subtle but very plain when using a good SDR receiver and sound system, even with band conditions. The effort is worth it to have this kind of transmit bandwidth versatility. In this case it's about listening enjoyment rather than raw communications.

T

My hearing is not what it was 60+ years ago when the horizontal output on our 21 inch round kinescope color box sang in my ears.  But still the higher frequencies make the difference in getting the message clearly on the first pass, or asking for a repeat. 

Sibilance is SO important in understanding!  If the bandwidth is available, we should use it.  When we are through, someone else can have it.  Fully renewable resource!

OM, you are 5 by 9 here in South Podunkville.  Would you kindly repeat your call, handle and QTH, preferably with phonetics?????  Yeah, Right!

The latest SCAF filter test:

I got the four MAXIM 295 chips running in cascade and made some rough tests for now….

The ripple is doubled as we expected.   The skirts definitely got sharper. Before, the TWO chips would start dropping at 5 KHz and be down to a premarked low level at about 7.2 KHz.   With the FOUR chips, the same drop was finished at 6.5 KHz.  

In comparison, the same passive filter drop starting at 5KHz ended at 7.7 KHz.

The four-chip narrower skirts are significant, but the ripple is now starting to get out of hand. I plan to hook it up to the rig and listen carefully.

The plan now is to plug in the MAXIM 296 chips into the same sockets.  They have less ripple and less rolloff - a good compromise.


T

Update:

After trying many combinations of sequence using four Maxim 295 and 296 SCAF audio chips I have determined that the original two Maxim  295 chips in cascade are the best. There is a compromise between ripple and skirt selectivity.  Two chips appears to be the point of diminishing returns.


I tested the board with two 295 chips used in the Max processor. (on air)  I found it to have good on-air skirts with no noticeable ripple or clock generated IMD.  Rick tells me they will stick with this design and move on to other testing.

T

Am I seeing some overshoot there? This is what I was talking about in my impulse response post.

Does it mute for a few ms (the little gaps) when shifting gears or do you do that in your design?
If so, Can it be 'speed shifted' so muting is not present, or it doesn't work that way?
No one would notice, just curious.

Hi Pat:   The gaps resulted from intentionally pausing my voice ssss's to make a clean separation between filters.  

It can make seamless excursions with no carrier interruption if wanted, producing sloping lines during adjustment.

Steve: Yes, there is some overshoot in the SCAF filter. The best filter uses two Max 295s in cascade for skirt and overshoot compromise.

I have been experimenting with some newer DSP audio filters that have the same 8th order Butterworth performance without the overshoot. There are many ways to accomplish a tighter, more contained signal.  Passive, SCAF, DSP, etc.

Presently, I set my AM transmit filter to start roll-off at about +- 4.5 KHz and the normal sibilance sidebands stop at about +- 6 KHz, which is a good compromise for fidelity and reasonable bandwidth. I use it all the time now and occasionally get comments from SDR waterfall users saying even though the signal is BIG, (4-1000A) it is tight and well behaved. That was the goal. It was a lot of work experimenting, but I'm very happy with the final results.

T


I usually run the bottom or center filter curve unless the band is quiet:

Nice spectra there Tom!

Check out Bessel filters for no/minimum overshoot.

 Another possible option is to use  DSP  filtering.  Then one can  set the TX  BW  to  anything you want.   Brickwall  cutoff,  no phase shift, and no ripple,  no distortion.   I use this technique when narrowing a  6 khz  ESSB signal  down to  a  narrower TX  BW... like 3.5 or  4 khz... or  5 khz  etc.  DSP can also be used to limit the lowest TX freqs,  like 40 hz,  50, 80, etc.   Most of the  DSP  boxes will  cover  20 hz to  20 khz..and anything between. It can  be used for AM /  FM / ESSB,  cxr and one sideband, etc.

Shane put me onto this thread a long time ago, and I 'forgot'  to offer a comment at the time.

Jim  VE7RF

Hi Jim,

Yep, that's exactly what I am doing now.  Since this thread was written about analog filters I've changed over to DSP control for bandwidth, limiting and EQing - works FB. Yes, I use mine to make a sharp cut at below 40Hz too.

I am using a DBX DriveRack speaker management system. It lets you input a low level audio signal and tailor it in DSP.  The only limiting factor is hi-level IMD from the final that can cause wider bandwidth than desired.  I've worked hard to make my rigs clean so the low level DSP stuff works well. No pre-distortion here yet... just old school homebrew tube AM transmitters or standard linears for ssb.

I think I've seen some of your stuff around related to hi-fi ESSB and it looks informative.   Thanks for stopping by.   

Tom, K1JJ