The AM Forum

Did some CAD. Based on the old 'Bonadio' speech LPF but some values simplified and all items found the cheapest.

Nice! What's the insertion loss at 1kHz? That would go well after the limiter/clipper I'm putting together.

It would be handy to see variants for other bandwidths, and as a bonus, a parts/price list assuming all the components can be sourced from one distributor (saves shipping charges).

ask and ye shal..
board 6x6 inch. the coils are critical as to dc resistance. It has to be low or the response suffers.
It is not required to use a board, just would make it neater. No board will be ordered anyway until a breadboard test is done.

attenuation:  dBv
0.14 dB @ 10 Hz
0.15 dB @ 100 Hz
0.2 dB @ 1000 Hz
0.23 dB @ 2000 Hz
0.3 dB @ 2500 Hz
0.45 dB @ 3000 Hz
0.87 dB @ 3330 Hz
03 dB @ 3477 Hz
06 dB @ 3539 Hz
09 dB @ 3584 Hz
12 dB @ 3629 Hz
18 dB @ 3713 Hz
24 dB @ 3793 Hz
30 dB @ 3865 Hz
36 dB @ 3927 Hz
42 dB @ 3975 Hz
47 dB @ 4000 Hz
max atten:
46.8 dB @ 4650 Hz

Boy that was fast! Custom boards are always going to be $$ so that's no surprise. Sure makes me appreciate the Radio Shack project boards.

Lots to think about here...thanks!

For that price, buy a CNC and never pay the outrageous 'setup' fee again.

I did.

--Shane
KD6VXI

What's the impulse response?

It took a few hours to do this. Mostly learning the program's limitations. For example, you can't globally 'change all pads to dimensions xyz', etc. However once the board is laid out, circuit value changes are simple enough. The original Bonadio's filters used odd values, taking 3 or 4 caps per 'capacitor', I cut it to two. It also uses odd inductances which are not off the shelf, I chose some that are. All 10mH, using three in series for 30mH, etc. Defilement of that 'perfect' design didn't hurt anything by more than 1dB. So maybe a filter design program using the same basic schematic could be made for any use. That is up to others to do. I would not mind zipping and posting the cad files and all info, with the understanding that the board has not been made or tested, perhaps some holes are too small, but I did make a sincere effort especially concerning the inductors, for which a special ccustom footptint had to be made - was easy. BTW it cost $250+ to make 'just one'.

The green picture is the bottom side/solder side. There is also a full coverage ground plane on top. Was thinking of RF avoidance.

The huge 600mA inductors are not so the thing can sit in the 500 Ohm line between a 10W speech and and a class B driver transformer. They are for low DC resistance. Above 7 Ohms in this circuit and the curve gets less efficient.

That said, it's a function of in/out impedance -vs- inductor/capacitor series resistance. A filter made for 10K Ohms could well use small SMD coils with low L and higher R. Perhaps a 500 Ohm to 10K transformer at each end could enable a board to be made with 10K in and out impedances, with the smaller parts. As an example, look at Kahn's rotator.

If I redo it, after any input is all collected, I might want to make a board design that would be long and narrow so it could fit behind a 19" panel, or someone else might.

I posted this as a possibility only, but if 20 of us went in, took the risk, after it was proven on a breadboard, and after the layout got a very serious review (hah like at work, how tedious but necessary) the final cost would not be so bad. Could silkscreen the AMFone url on it.

Going farther, The pre-emphasis curves could be applied before this. I explored passive networks but they are extremely lossy, so an active circuit would be needed. Care must be taken not to give so much pre-emphasis that the good attenuation at 4KHz suffers.

One thing noticed in LTspice is that it can be driven from a 1 Ohm or 500 Ohm source no matter to the filter shape, and the input Z changes with frequency so the signal source should be well regulated, but the output needs terminated halfway decently.

On that note, Two or more of these circuits can be put in series, to gain much steeper dB/octave cut at the same frequency without much loss in the pass band, as long as a 500 Ohm network is placed between sections, and that will cause a few dB loss if it is the typical 3-resistor setup. The dB responses just add. One of these= -47dB@4000Hz, two= -94dB@4000Hz, etc. So pre-emphasize away!

Just more food for thought.


@ Shane, how would a mill help? I guess hole-drilling, but how are the boards made with plated through holes at home? I am not familiar with making this kind of nice board, this process, just the old ink etch and drill.

@ Steve, No idea what the impulse response is. Untested. The LTspice file is attached as .TXT, just rename to .ASC  I might get to it, what pulse parameters did you have in mind?

That filter has quite a sharp cutoff. Impressive.

A pulse replicating a 3 kHz clipped waveform could be instructive. I've done this on a few passive LC filters used after clippers in amateur radio transmitters. There was usually a fair amount of overshoot which would reduce, or eliminate the effectiveness of the clipper. Just something to think about.

Pat,

It won't plate through holes, nor will it flow them.  You have to do that yourself.  I haven't done a ton of boards, but hasn't been a problem thus far.

I suppose, if I where using them commercially, I'd let the end users know.  I just make sure both sides (pads) have solder on them.

I don't believe 'old school etching' would address the plated holes either.....?

Anywho, the technology is there, and for less than 400 bucks and a couple evenings of assembly time, you can have a CNC that will etch up to 6x9 inch boards, as well as has the capabilities for up to a 10 Watt laser for stainless, etc.

Next I want a 3d printer.  Want to print my own polystyrene caps! Lol.  Just cuz I could!

--Shane
KD6VXI

I'll see about the 3KHz pulsed signal at some point. Bonadio was not clear about clipping that I know of, but he liked pre-emphasis and then to use the filter (wish I had the pre-emphasis schematic of his from the "EAM" article just to see what was done).

I didn't really like the square board. Also the footprints of the coils were off almost a mm, so that was redone for this re-do, which is a board about 13"x2". unused caps etc were eliminated. Nothing was corrected on the previous 6x6" board.

,

Impulse response.

It’s a passive network of RCL components.
As known, such networks are integrators slowing down or averaging incoming incoming transients. The very act of limiting higher frequencies displays this.

I can do the math for a simple integrator but for this assemblage, um uh, I’ll leave that as an excercise for others.  ;D

Hey back to the future TIM wars, don’cha know.
Those were somewhat latter than the music power wars.
Remember “straight wire with gain?” Yeah, that was for active devices.

A pulse replicating a 3kHz clipped waveform is a square wave; betcha it doesn’t come out square.  For that matter getting a perfect real world square wave input is impossible except within proscribed limits. Mathamatically, of course easy and precise if you can handle infinity so to speak. So an impulse response is already imbedded in a good CAD program, by mathematical construct. It should be back trackable.  Even in the slow gain loss, lower frequency portion of the passband, you can see a dA/df (differential gain/d freq.) relationship.  The submitted table clearly shows this. Reduce it to loss/decade or octave, logarithmically if desired from 10 to 3k Hz.
This can be directly correlated to impulse loss as a function of frequency for the operative part of this filter.

There, I’ve trotted out about all I know to sound erudite.  ;D

It sure is a nice sharp cutoff filter though. Very nice.

I think it's a good design, the PCB Any suggestions?
I can't afford to have the boards made and if anything will have to do a point to point build.
- but may as well leave a nice item for others. I like layout of little things like this, it's fun.
the setup fee per order is more or less fixed. The rest is the square area of PCB.

There is some ringing at about 3.3KHz when a 300Hz square wave is input.
The square wave rise/fall time is 100us. Previously it was 0.1us. Did not make much difference but is more realistic.

Is that ringing going to be an issue? I sorta doubt it, as long as it is inside the passband.

Your graph shows both input square wave and output ringing wave at .0033 s periods for your 300Hz.  I see it so far. Is the 3.3 KHz ringing your talking about from the period between crests on the superimposed damped waveform? Oh, think I answered my own question. Looks like about 1/10 period of the base wave.

And the 3.3 KHz is Inside the filter passband (maybe 3dB down or so) but much smaller.  Almost has to be a little harshly audible unless a fairly exact harmonic of the base 300Hz or it’s of very low amplitude.

Does the filter show ringing at just 300 Hz input or did you sweep it and found only this one case or at even multiples of 300, etc. ?

For our purposes, perhaps lesser order, less bricky filters (if we’re going the all passive design route) would be more pleasing sounding. .. not that I can hear much difference any more.  Most of the distortion I hear these days is between the stirrups and the stapes deterioration or some such.  ;D

If the cost from ExpressPCB is too much, try what I did. There are some free PCB design programs such as KiCad or DIPTrace that produce Industry standard Gerber file output. Then upload the files to an overseas PCB house like allpcb.com, easyeda.com or jlcpcb.com. These all make full double-sided plated thru hole boards with solder mask and silk screen for dirt cheap. Extremely high quality and less than a week turnaround to your door.

An example: I ordered five 4" x 6" boards from allpcb.com on a Sunday evening, had to re-submit on Monday evening because I forgot to include the NC drill file. The boards were already in DHL's hands Wednesday morning and delivered to me on Saturday afternoon. The cost? $33 + $1.44 PayPal fee. That includes all setup fees. Smaller boards are even cheaper. And they sent me 8 boards instead of the 5 I ordered. This is typical!

Yes, you have to learn a bit more complicated program than the EasyPCB program, but you get industry standard Gerber files and a much lower cost. The quality is incredible.

August KG7BZ

I just chose 300Hz randomly because the simulation at 3KHz seemed too good to be true.. The ringing period is as you say, in the passband. it's at the high end of it.

consider adding high value resistors across C5 C6 and C7 .... I am not a filter guy but intuitively these resistors would smooth the stop band parallel resonances at 4 5 and 8 kHz and dampen the ringing exhibited ...John

IF your system happens to have feedback wrapped around a filter like this, be careful,
look at the phase shift!

This could happen if you put the filter into a speech amplifier, and you have feedback.

No feedback, no likely problem...

If you reduce the Q by using resistances you'll also will also reduce the stop band depth(s)...
The filter, iirc, is also sensitive to input and output Z...

Probably for typical ham use the multiple sections are excessive... being ~30dB down above the
roll-off frequency is likely more than sufficient, given the "brick wall" slope of the initial roll-off.
Since most AM ham gear rolls down anyhow as one gets above ~10kHz, it seems likely that
this is sufficient... adding one more 2nd order pole above the roll-of freq will pull the ultimate
HF response down about 24dB below that area above the initial "cliff" by two octaves higher.
In the example shown, that would put it down 24dB by about 20kHz, from the initial plateau that
occurs after the initial "cliff", which is due to the filter in essence being a high Q notch.

Fyi, these are elliptic filters...

If the DCR of each inductor was raised to 7 Ohms, the sharp downward spikes go away, but the filter is not as sharp. (still sharp enough) but the ringing is not affected . I have not tried shunting the caps with resistors.

You need to shunt the inductors

shunting the inductors  (same as C5 C6 C7) enough to eliminate the ringing softened the cutoff too much. 1K is shown.
2K was about the max. and the curve was still pretty soft.
5K the ringing was as much as before.

There are lots of commercially made 'can' stye voice frequency filters in the older decades. I wonder if they have the same pulse response issues.

On the other hand, if the filter is used before any clipping is done, the problem seems to not exist much.

I don't use clipping, just some compression more like AVC for the mike, and and soft limiting. There are no sharp corners to cause the problem. That's probably the best answer.

nice work Pat .... 25 or so years ago National Semidestructor made a switched capacitor filter low pass filter with steep skirts and predictable phase response and were inexpensve ... can't remember the model number so going to Google it ... MF10 rings a bell .... a copy of the applicatons handbook is available to download for free ( mf10_AN-1) .... something to feed the ravenous Bunker of Doom ....more input ...more input  :-*  :-*

Your softer, recent bypassed inductor design seems to be about equivalent to a simple RC low pass filter.
https://en.m.wikipedia.org/wiki/Low-pass_filter#/media/File%3AButterworth_response.svg

I have a low pass filter built into my HB xmtr.  I used a shielded UTC audio inductor and some caps.  Works very well.  Slight rise at 5KC and then drops off sharply at around 5.5KC.  No audio past 6KC.

Fred

I disagree about equivalence to a simple RC low pass filter. You can check them together in LTspice.
If that RC curve was the same it would be -20dB @ about 1080Hz, not the 10KHz in wikipedia.

If anyone wants the ltspice and expresspcb files for the long board shown, they are here. Free for anyone to use or change.
Simple matter to change the freq., footprints, etc.
I have not had time to do anything, have boards made, etc. It might be nice paired with a pre-emphasis circuit.
bunkerofdoom.com/__scratch-monkey/3500Hz audio LPF LTspice and ExpressPCB.zip

OSHPARK for PCBs chaps, they're in the USA so why look 'overseas'?

JB.

Pat,

Can't seem to access the files. Can you post the zip here?

I use EasyEDA and it works well with any of the PCB houses. ExpressPCB is a US based house with heavy setup fees and their software locks you into their house. When I did the 25 board run for the Mosfet driver board I used a US house and it was $100 for setup and then so much per board. China is much cheaper with no setup. $250 setup is robbery.

John

OSHPARK has no setup fees.

I just tested the file, it's OK. Perhaps something is improperly blocking it.
I would have put it here in the first place but ZIP files are not allowed to be uploaded.

Most of what I did on this is a matter of personal preferences. design tools cost money, so there are may ways to go. Some people have access at work  others can buy a seat, others want a free program or a $30/month subscription.

I don't want to use EasyEDA because it is online / cloud based.
I would rather buy a month of Autodesk Eagle or something.
Want files and program on my computer and not forced to work online.


Here is why I chose what I did:
1.) Wanted a free CAD & turn key PCB.
2.) Considered a run of 20 pcs PCB to minimize the one-time costs.
3.) As much USA-made content as possible to support my countrymen first, not foreign interests.
4.) Able to get EMS sales engineer on the phone in clear English during my business hours.
5.) Per unit costs a reasonable $65, as complete kit.
5a.) BOM: ~$40 less PCB.
5b.) PCB shipped: $24.17 each total.
6.) Assembled/tested/verified version could have been offered as well.
7.) If I was going to have PCBs made, I want the best quality control over them.
8.) It's not a commodity product. For myself I don't care about $24 for the PCB.
9.) Nothing's stopping me from doing one piece for myself on a home made PCB or perf-board.
10.) Can this even be bought online? Passive 500 Ohms, not a board with ICs.

Hope that makes sense. Main goal is quality.

------------------------------------------------------------------------------------------------
Inductor selection:
One should use lowest DCR inductors on a 500 Ohm filter (2.3 Ohms per 10mH).
Even the $1-2 inductors (10 Ohms per 10mH) are poor performers.
Using cheap 10 cent (40 Ohms per 10mH) inductors is terrible.
(see plot to compare)
also .CSV as a text file for the BOM with part numbers.

I used Precision Technologies for the driver board. US house based in Chicago. $100 setup and then a sliding scale based on board size and quantity. Express was considerably more.

I used to use KICAD which is free, but can be complex for the average user. Probably on par with eagle.

John

In the days when I had an actual radio room, antennas, and iron, I had a UTC LML-4000 in the audio chain (pre-compression).  Still have it, if my inventory list is accurate  :o   Not quite as sharp rolloff — but seemed to do its thing even with some Z mismatch.

https://drive.google.com/file/d/1ERNuEPvB3wKUfkAkTk_cuKl3XSdh_NCw/view?usp=sharing

Grant NQ5T

Explain to me like I am the newbie I am --

Why do you want such a filter? It has tons of phase shift, time smear, etc., for the sake of a super deep fast cutoff, but why? I get that there's distortion and overload and other reasons to band-limit the audio, but  a roll-off of 12 or 18 dB is a huge voltage/power difference, and this thing is like 40+ dB down, even with the higher resistance inductors. Wouldn't a simpler filter have just as good real world results as far as modulators, transformers, etc., are concerned? Or just use a couple 12AX7s, and a handful of caps and resistors, as in the attached schematic?   Help me understand...

Ed

Ed poses a good question (s)  ....  I am second guessing with this:  the sharp cutoff is an effort to keep preemp audio within a 10kHz rf frequency using am.   does this answer your question or does it need to be spelled out ?

Since you have loaded the params into Spice orone of its fellows, why not run the phase shift curves

My ancient filter training stated that an eliptical or Bessel response had the least phase shift.  I dont know about group delay.

Hi Ed,

It's just a preference.

Here are my own reasons:
sharp cut-off.
analog
passive
>1dB insertion loss out to 3KHz.
500 Ohm impedance common in audio stuff.
Can pass a little audio power as well. 5W easily. Durable.


I'm not too concerned about phase shift. It's predictable in LTspice, but it's not for modern precision hi-fi use where minute frequency-based timing differences in harmonic sounds from some instruments would be more offensive than the delays caused by crossovers in speaker systems. It is more like mid-1950's hi-fi quality in regard to phase shift through the system including bass and treble controls and the like.

In fact, the Kahn Symmetra-Peak all-pass filter intentionally shifted the phase according to frequency and it was successully used in broadcast AM and FM to 'unstack' the high and low frequency elements that occur at the crest of the modulation cycle so that the average modulation is more symmetric and could be made higher.  There had not been complaints about phase shift from listeners as far as I know. The Symmetra-Peak is applicable to ham use since the offending stuff is well within the usual ham baseband. I have never compared the phase shift of the Symmetra-Peak to this simple filter since they are for different purposes.

A cursory look shows that both 'phase-smearing' devices are reasonably similar in that one respect:
In the LC filter there the phase slopes from 0 to 425 degrees at 0 to 3.7KHz.
The Kahn product shifts from about 490 degrees from 100 to 4Khz.
So there should not be an issue for ham radio.

What amount of phase smear would the string of RC filters in the cathode follower circuit have if designed for a 100-3500Hz +0/-3dB passband response and -20dB@4KHz? (more easy than the -43dB of the LC filter and a good question)

Gentlemen,

On behalf of the OBC (Old Buzzards' Club) and the AOAL (Acronym Overload Action League), please provide the meaning of the following TLA's (Three Letter Acronyms):

LPF
SMD
TIM
DCR

HAND. YMMV. TIA.

73,

Bill, W4EWH

LPF = low pass filter

SMD = surface mount device

TIM = Transient Intermodulation

DCR = DC resistance

Each of these may have several other meanings and must be taken in context.

Here is an article about inexpensive low pass filters that have a gentler slope.

In my world 3.6khz rolloff with such a sharp slope is a bit "tight"... ymmv.

I designed a similar filter for a Globe Champ 350 - since it had an inductor there already.
Fewer stages... higher cutoff freq... I think it was ~5.5khz.

But I agree, one could put gentle first order or even second order rolloffs into successive
stages and obtain a satisfactory result as well... but that wouldn't work for a single "audio
chain" that was to feed multiple rigs.

Otoh, today one can buy rather inexpensive digitally based xovers (nominally for PS/SR)
and cascade two channels for impressive stop band action and be very "frequency agile".
But, it's not analog. :D

                         _-_-

Do you have the filter design available?

Here is a nice Chinese pcb manufacturing video from PCBWay.
https://www.youtube.com/watch?v=_GVk_hEMjzs