What is causing this audio highs attenuation/ distortion problem?

[K1JJ]

Jeff/ W2NBC made a recording of the 4-1000 rig tonight running at about  75% power. No negative peak limiter yet.  Have a little acoustical feedback from the mod xfmr being out of the cabinet - and occasional arcing from the spark gaps.  

Notice the high end sounds reasonable now.

It's getting closer.

T

[KA2DZT]

Tom,

I have been thinking about the plate blocking caps.  I was never 100% sure (in my mind) that the cap has anything to due with the freq response of the xmtr.  All this discussion about the audio being shorted by the RFC that's on the pie network output doesn't make any sense.  The cap is not really passing an audio signal but a RF signal that varies in amplitude at some audio frequency.  The RF signal is not being shorted by the RFC on the output.

This is not true when it comes to the bypass cap on the plate choke.  This cap is seeing an audio signal (modulated B+) and it has an effect on the audio highs.

Maybe we can get some other thoughts and comments on this point.

What does have an effect on audio bandwidth of the xmtr is the Q of the tank coil.  Too high a Q and the outer limits of the bandwidth are reduced.

Fred

I thought about this some more,  maybe the blocking cap does affect the audio highs,  not really sure.

Back for a third comment,  I guess the blocking cap does affect the highs after all.  The bypass cap and the blocking cap are both to ground only separated by the plate choke which has little resistance at audio frequencies.

It's after 2AM, maybe it's time for me to go to bed.  But, when I wake up later, I'm sure I'll have more brilliant ideas.

[Opcom]

This topic is just about like a tutorial!

You are using 160M? then your blocking cap may be large.. but what you said, that the plate choke may pass the audio right through to a big blocking cap. I have 2500pF there. I only use 80M and up, and have a HF roll-off with high modulator drive needed to compensate and high mod plate current results. Maybe it's the same kind of issue. My audio path through the choke, and then to GND would be a 2500pF blocking cap (five 500pF door knob caps in parallel) and an RF choke for 'dc on the output' safety. When the modulator was tested into a resistive load of 7000 Ohms with the RF side dark, the roll-off was not an issue. When the transmitter operates, it is an issue, despite people say it sounds fine a spectrograph shows it up. Could it also stress the modulation transformer?

Fabio's RF is clean I assume. If you do not mind, please tell about Fabio's plate suppressors so I can make one just like it for my single 4-1000.
I've never totally got rid of a parasitic and its cause may be the original home brew parasitic suppressor I'm using.
It's not a harmonic, it's something in the plate area with resonances at ~100 and ~165 Mhz.
The LPF at the output absorbs it OK but it's visible on the scope as barely visible little "notchies" on the RF waveform, and so it's too much and it drives me crazy because it is not perfect.

[flintstone mop]

Tom.......where can we hear all of this beautiful TX audio?? And what times? I don't waste time on 75M and the zoo.
Fred MOP radio

[KA2DZT]

Opcom,

2500 is probably is way too much.  Try removing at least two of the 500s and see if things improve.  Also what cap are you using to bypass the plate choke?  This is a plate modulated rig that you're referring to??

Fred

[K1JJ]

Opcom,

Yes, I agree with Fred. Take some careful audio sweep measurements and then disconnect a few paralleled caps. Do both the plate bypass and coupling caps to see what it does.  I'll bet 500 pF is all you need for both places with one tube.  You will need less C than my rig.  Look at the last pictures I posted for an idea of the parasitic suppressors I used. I just wound 5 turns with 50 ohms and that was it. 500 pF at each screen pin. I loaded each grid with a 5K with a .001 cap in series to ground for more stability. It is neutralized too.


You menti0ned that your modulator is drawing big current to handle the high freqs. That is exactly the symptom I had and it went away when the caps were reduced.  I'll bet you are seeing some waveform tilt and harmonic distortion when you run the modulation up at those high freqs.  It is definitely a strain on the whole modulator system.  It's really putting a big cap across the mod xfmr, isn't it?

Today I'm working with Frank / GFZ and his SS driver. I see a very slight slope in the triangle waveform coming out of the driver itself. He doesn't see it on the simulation so it is in my implementation here. Gonna try some different things.

The bottom line - when striving for pristine audio with a tube rig, it's solving a combination of many small details until they all add up in one positive direction.   As W2NBC told me, "refine, refine, refine."    (Don't tear it down)


Fred/ Mop:  I'm just on 75M for now. Plan to try 160 and 40M once things are running right, Catch ya then.


T

[W2VW]

With all that energy dumping into bypass caps when do they start smoking?

[K1JJ]

With all that energy dumping into bypass caps when do they start smoking?

I wondered the same thing.  I didn't remember to give 'em a feel after a long carrier at 8Khz.

Is the capacitive power factor the reason?   I'll bet if the capacitive reactance was instead pure resistance, it would have smoked by now...  


Added a Hall effect shutdown circuit to the Dual Quads PDM rig today. Nice - if it goes above 800 ma, it shuts down as fast as  a Monty game on Broadway.


Still thinking about an NPL for Fabio II. I don't know if it's smart to be leaning on an NPL with that mod iron. Might be better to run a fat carrier and do 110% positive and call it a day.

T

[W2VW]

Time for one of those IR thermometers.

[Opcom]

Attached is a schematic simplified for this audio stuff.

Screen is bypassed at the socket with 2000pF. (two 500pF on each SG pin)
  It's fed DC by a 8H Chicago choke and 2K dropping resistor from a 0-700V supply having a 10uF oil cap at the output.
  I seem to have enough screen modulation voltage. I have not tried to check its phase relative to plate modulated voltage phase.

Plate choke is a Raypar RL-100 (never gets hot)
  (96uH, 18KV, 3KW AM rated, self-res 22.5 to 26.5MHz)

Plate suppressor is 4 pcs. 10R 10W Koolohm NI
  (in parallel with Burnstien Applebee 17B684 suppressor coil)
  So, I could exchange that for the 5T coil with a 50 Ohm resistor across it? I could not really see the resistors there but will look again. I see yellow things, they looked in series.

I forgot about the big caps across the mod transformer.
  There is a .003 across the primary and .001 across the secondary.
  Should one of them go, or both? I thought they are needed to protect the transformer.

Yes the DC blocker for the plate is 2500pF. It's five of those TV-type doorknobs.
  I was reading that they handle RF current poorly and can heat up and fail.


It is effectivly a 6500pF load. 0.0065uF. 
  The load presented to the modulator is 7500-9500 Ohms, depending. I'm using only about 300mA @2500-3000V.
  If I did the math, those caps all add up to a 7000 Ohms reactance at 3500Hz.

OK I never noticed how hot any of them get, I usually stay away from them! An IR thermometer may e hard to use safely because of how it focuses.

[N2DTS]

I never run caps across the mod transformer, and always removed them from any rig that had them.
2000 pf is really high for a screen bypass cap, there go the highs....

They really wanted that rig narrow I guess.

[AB2EZ]

For a capacitor or an inductor, the current is 90 degrees out of phase with the voltage*. The average power (averaged over each cycle of AC) flowing into the capacitor or the inductor is zero*. [This is also described in the AC power literature as a situation where the power factor is 0.]

During half of each cycle, power flows into the component, and is stored (not dissipated as heat) in the electric field of the capacitor or the magnetic field of the inductor. During the other half of each cycle, this stored energy is returned back to the source as power flowing out of the component.

*Ignoring dielectric losses in the capacitor, and resistive losses in the wires of the inductor.

Because of dielectric losses in the capacitor, the voltage across the capacitor will produce some in-phase current (modeled as a resistor in parallel with an ideal capacitor)... which will produce heating... possibly enough to destroy the capacitor in a dramatic way. Because of resistance in the wires of an inductor, including the skin effect, the current flowing through the inductor will produce some in-phase voltage drop across the inductor (modeled as a resistor in series with an ideal inductor)... which will produce heating of the inductor.

Stu

With all that energy dumping into bypass caps when do they start smoking?

I wondered the same thing.  I didn't remember to give 'em a feel after a long carrier at 8Khz.

Is the capacitive power factor the reason?   I'll bet if the capacitive reactance was pure resistance it would have smoked by now... 


Added a Hall effect shutdown circuit to the Dual Quads PDM rig today. Nice - if it goes above 800 ma, it shuts down as fast as  a Monty game on Broadway.


Still thinking about an NPL for Fabio II. I don't know if it's smart to be leaning on an NPL with that mod iron. Might be better to run a fat carrier and do 110% positive and call it a day.

T

[N2DTS]

The rig sounded good to me, but then most sound good to me.
I could not hear the arc's.

[W2VW]

A perfect capacitor would not have any problems. Real capacitors shunting several hundred watts of steady tone are going to show some distress.

[K1JJ]

Opcom,

Yes, sweep the rig and pull out the mod xfmr caps - and start reducing the others you mentioned.

I've had a VERY interesting conversation with Bruce / W2XR.  He feels that tetrodes do not do well with capacitive loads, while triodes do better.   I might run some THD tests as -is (tetrode connected) and switch to triode-connected (grid tied to screen and driven) for the same tests.

Anyone have an opinion of running 4X1s tetrode or triode config?  I'm hoping I will get more voltage swing to work the NPL and maybe see it sweep above 8KHz when triode-connected.

Check out these excerpts from Bruce. VERY interesting comments and think they can help others make a decision too. He gave his permission to post:

---------------------------------------

Bruce's Comments below:


I have been following this thread with considerable interest.

I think the bulk of the problem you are experiencing here may have to do with the tetrode modulator tubes driving a highly capacitive load impedance. Tetrodes do not swing voltage well across a capacitive load; triodes are much better in this regard. That is one key reason as to why I have only used triodes (833As) in my rig over the past 35 years. I have 1000 pf as the RF plate choke bypass, and 680 pf as the plate blocking cap, and my rig can easily modulate linearly to 12 KHz, and with no excessive modulator plate current, etc. I may well reduce the RF choke bypass to 500 pf, and see how much of an improvement results. I suspect the improvement will be minimal in my case.

Reducing the values of the RF decoupling/bypass and blocking capacitance in the tetrode-modulated HPA should result in a substantial improvement in the high frequency response, and a corresponding reduction in the HF distortion, as the modulator will not to be driven as hard to maintain a given percentage of modulation above the point where the capacitive load starts becoming difficult.

The other advantage I like of using triodes as modulators is the reduction in the possibility of over-voltaging the primary of the mod xfmr in the event of a short-term/transient unloaded condition. This is one reason why most triode-modulated BC rigs did not use protective spark gaps across the mod xfmr, whereas it was almost universal with tetrode or pentode-modulated transmitters.

I am of the opinion that one of the best choices of modulator tube for 1000 watt plate modulated service is the 3-500Z; zero bias, class AB1 operation (no driver xfmr or audio power required), and extremely linear operation. Jeff/NBC is of course building a rig around this tube. The downside is the availability of finding good, serviceable Eimacs nowadays.

Your rig sounds absolutely superb, and the audio quality is the benchmark to which I strive for in my plate-modulated rig. I'd really enjoy trading notes with you sometime about your audio processing, etc. I'm back on the radio, typically Saturday afternoons, on 3705 KHz.

All the Best & 73,

Bruce, W2XR-----------------------

Interesting on your take of tetrodes and capacitance. I never knew that. It makes me wonder if I shud put the 4X1's into triode. (grid/ screens tied together) I ran it that way before but the Tron said it wud sound better in tetrode, so I did. I personally never liked the sound of tetrodes for any kind of audio amplification application, when compared to low-mu/low plate resistance triodes, such as the 833A, but your rig does spectacular! This may well be an example where the technical solution may not be optimum, at least in my opinion, but never knock success. When you designed this rig around tetrode modulator tubes, you did it right; the screen supply of any tetrode or pentode linear amplifier must absolutely be very tightly electronically regulated, to minimize the distortion inherent in this type of tube. By their very nature, tetrodes and pentodes have a much higher plate resistance when compared to triodes, and this in part accounts for the difficulty they have linearly swinging and developing voltage across a substantially capacitive load.
>
> I know that 813s triode-connected work absolutely superb. No question; they were triode-connected, this does reduce the Rp of the tube in this type of operation. That is the only way I personally would run a tetrode as an AF power amplifier; triode connected. But personally, I still think a true (not triode-connected tetrode or pentode, also referred to as pseudo-triode operation) directly-heated triode (DHT) with a low Rp is the best, most versatile, and linear power amplifying device you could wish for. 810s are not may favorite either; they may be a DHT, but their Rp is a little too high for my taste. The same is true for many of the Eimac triode transmitting tubes, including the 3-500Z.

One other really nice thing about triodes in modulator service is the ease of varying the modulator and HPA B+ to lower the RF output. With tetrodes, you hit a brick wall due to the screen voltage, and if you lower the AF output by virtue of reducing the B+ too much, the distortion increases radically. That is one reason why some tetrode BC rigs resorted to stupid (but necessary) means to reduce the RF output (i.e. 1000 watts to 250 watts, etc.) by putting a high-power audio attenuator network between the modulator tubes and the HPA load. With triodes in the modulator, you simply change the tap on the plate xfmr (or variac the AC input down), and all is well.

Did you ever get GFZ's driver going there?  Not yet, but FC and I are in touch about this. I still have to take the scope to one of the push-pull grids of the 833A modulator deck, and measure the positive voltage swing for Frank. My gut feeling is the existing GFZ design won't provide enough linear voltage swing. I also have the 4x 807 DC-coupled cathode driver unit from a Gates BC-1G I can use too, to remove the existing driver xfmr and massive push-pull 845 audio driver I currently use. The p-p 845 DHT triodes, 6 dB of NFB,  and the 23 lb Peter Dahl replacement for the original Gates BC-1F driver xfmr still sounds extremely good and clean to me. The issue is occasional overshoot with this set-up, and my desire to provide even tighter control of the modulation with my Orban Optimod 9100A.

I'm inspired to give the triode connected 4X1s a try.    If I had a pair of 3-1000Zs I use them in a heartbeat, but I don't.

Now that would be a great tube to try in this application, but I am not sure as to what the Rp of a class AB1 3-1000Z may be.       But the 4X1 is so rugged even *I* have never blown one...:-)


I will run some THD and IMD tests on my SDR receiver as-is. Then configure them to triode and test again.  I've used triode-connected several times in the past and liked them.

I'd really like to see the test results of the two different configurations. The load line of the triode-connected tetrode vs. that of a true triode will be different, and the true triode will prove to be more linear. I have no idea as to what the Rp of the triode-connected 4-1000A will be, and I'm sure that Eimac does not publish this information in the tube data sheet, nor would I know how to calculate this. Perhaps this calculation is available in the Radiotron Designers Handbook. Ideally, you need to know the tube's Rp so you can properly select the correct impedance ratio of the modulation xfmr.

I'd like to start a thread that discusses triodes, triode-connected tetrodes and tetrodes on the BB.  Kinda as a learning experience for those who will eventually build some of these rigs.  Is it OK if I cut and paste only the technical paragraphs of your last two PMs and post within my thread?  I will credit you as the author of course.    It might get a few more guys to contribute their own similar experiences... By all means, and feel free to do this.

[N2DTS]

I have tetrodes and triodes as modulators, and can hear and measure very little difference between them.
But my stuff is not in the same class as most, garbage audio equipment, crappy microphone, etc.


But I remember my push pull 812a rig modulated by 811a's was a smooth operating and sounding rig.

I have a bunch of 813's, maybe a pair set up as triode modulators is in my future.
Is there any info about drive requirements and power output?

I have four 100TH's in a modulator, but they take a lot of drive and want very high voltages (to lower the drive needed).

Something like 900 watts of audio at 3000 volts though...

I think its very odd that despite going through everything mentioned in this thread, my 32V3 has the widest response with voice. All the rigs use the same audio drive setup...
I can get them up to 2750 volts before I run out of bias voltage.

[VE3AJM]

Check out these excerpts from Bruce. VERY interesting comments and think they can help others make a decision too. He gave his permission to post:

---------------------------------------

Bruce's Comments below:

I am of the opinion that one of the best choices of modulator tube for 1000 watt plate modulated service is the 3-500Z; zero bias, class AB1 operation (no driver xfmr or audio power required), and extremely linear operation. Jeff/NBC is of course building a rig around this tube. The downside is the availability of finding good, serviceable Eimacs nowadays.

I've never seen any tube data for the 3-500z running in class AB1 modulator service. Does anyone have any information on this.

Al VE3AJM

[W2VW]

A 3-500Z would put out very little in ab-1.

[K1JJ]

AB2  -   At 3KV with 244 mA grid current a pair  will put out 1400 w of audio:

http://www.g8wrb.org/data/Eimac/3-500Z.pdf


But AB1.... as Dave mentioned, AB1 is a very stingy mode for output power.  That's the reason I started with the 4X1s as modulators thinking in AB1 they would have the headroom.  But even they are close to maximum because the grid current wants to flow early.

Though, I think maybe the 3-500Zs in AB2 might be a good choice cuz they were designed to run in GG linear service (with grid current) and are pretty clean tubes that way.  Though, I've never seen a tube in linear service that didn't degrade in THD/ IMD when grid current started flowing. It's just the nature of the beast.


T

[Opcom]

I never run caps across the mod transformer, and always removed them from any rig that had them.
2000 pf is really high for a screen bypass cap, there go the highs....

They really wanted that rig narrow I guess.

It was built in the days of SSB promotion where "narrow" was the happy fun goal.
The spectrum received on an SDR set on 16KHz BW is attached. I was only recently aware of this.

[W2VW]

But AB1.... as Dave mentioned, AB1 is a very stingy mode for output power.  


T

I think Al AJM was thinking about tetrodes. AB-1 plays a lot different with those : )

[K1JJ]

Here's some more information gleaned from a sharp amplifier engineer regarding triodes / tetrodes,  negative feedback and more.  I axed about switching my 4X1 tetrodes over to triode-connected config, etc.   I know some of you will find this info as valuable as I did. I have decided to stay with what I have - AB1 tetrode modulator operation with NFB.

-----------------------

His comments:

[axed about triode-connected compared to my current tetrode config?]
AB2 or B will require a really good low impedance driver and with the same 10 dB of NFB, probably be worse in distortion.   The gain in output power from the rig with AB2 or B is not much practically speaking since you already have 2X in plate dissipation in the modulator as in the final.  
 
[Asked about running bigger audio]  BTW, I have come to the conclusion that asymmetrical audio has no advantage over symmetrical audio and will always sound more distorted with most receivers. 110% positive is fine with negatives going to maybe 95% at the most.  The gain in loudness in going to 150% is under a dB and that's assuming a perfect detector in the receiver.  That's a lot of audio and increased distortion for little return.  I.E.: not worth the effort.  All BC stations these days including just about all SW broadcast now use symmetrical modulation. The gain in loudness and intelligibility comes from multiband audio processing while maintaining low distortion.
 
Run the rig with the maximum carrier capable of just 100% positive peaks with the available audio.  The PEP will be the same as with reduced carrier and 100+% audio.  The big carrier approach will sound cleaner and give you a better "quieting" effect.  Plus it moves the S meter slightly higher.
 
[Talking about source impedance of tetrodes with NFB vs: triodes] Yes, the 4X1's in tetrode mode with NFB will get close to the 3X1's and if you can run enough NFB, maybe pass them!  and yes, does lower the source impedance driving the final but that may not make much of a difference modulating a class C final since it's resistive and constant in impedance.
 
[Why is my mod xfmr / audio highs topping out at 7500 Hz?]  Well, the mod transformer is a big player here.  The leakage inductance will come into play and be your limiting factor.  Do you know what it is in that iron or do you know how to measure it?


[How to measure L leakage?] Easy.  Disconnect all leads going to the transformer. Short the primary and measure the inductance on the secondary.  It will be in the mH range.  That represents the loss due to incomplete coupling and results in an effective series inductance with the primary and secondary windings.  This "stray" inductance acts as a seris L with external reactances and typically limits big mod iron in the 6 to 12 KHz range. It is the main reason for phase shift at the high frequencies that limits the amount of NFB you can wrap around it.
 
Iron designed for triodes and other low impedance tubes will be a bit better, thus the anecdotal knowledge that triodes work better into squirrely loads.  
 
Bottom line is that you are probably extracting the last ounce of performance from that transformer!

Usually 6 to maybe 10 dB of NFB is OK but at some point you get into transient instability which what sounds like is happening to you.  You want to stay at least 6 dB away from that since that will be the beginning of IMD.  BTW, how does IMD look, two tones, say 300 Hz and 1350 or some such non even or odd multiple of two frequencies?  Try 50 Hz and 4 KHz if you are bold and have a spare transformer...{:>)
 
[Notice to stay 6 dB away from the instability point and get 6dB of working NFB means the circuit must be able to handle 12 dB of NFB without taking off - not an easy task]

[Opcom]

The only warning about the 3-500Z is that when used as zero bias grid driven audio amplifier it has a grid impedance to the driver that varies from about 2000 Ohms to about 500 Ohms over the half cycle. For this reason, a cathode coupled driver or another driver with good regulation might be best.
However, the drive part of it is not in any way causing a frequency response issue and the distortion looks low because the traditional "speech amp" driver has been super-sized and swamped.

This weekend's a Hamfest. Later I can work on the TX with the great cap-removal suggestions given.

[K1JJ]

More great info from Bruce / W2XR regarding NFB:

------------------

Indeed, the judicious application of NFB will definitely reduce the source impedance of the tetrode (or any other tube), but the issue is this: too much feedback can ruin the purity of the audio as well, and not only from the loop bordering on the point of going into oscillation with respect to phase. A very good rule of thumb to consider when using NFB is that true directly heated triodes (not triode-connected tetrodes or pentodes) need only about 6 dB of voltage feedback to substantially improve the bandwidth, amplitude flatness, and to reduce the distortion components. With tetrodes (and even triode-connected tetrodes) you will typically require about 14 dB of NFB to achieve the same results. This is another good reason to consider the use of real triodes in the audio path of the rig, including the high-power modulator stage.

Too much feedback can result in slew-induced (SID) and transient IMD (TIM), both of which are well known to experienced analog audio amplifier designers, but probably little known elsewhere. It can also substantially increase the spectra of the higher-order odd-order distortion products, and this is the most dissonant sounding distortion of all, on par with IMD. I personally prefer as little feedback as possible to do the intended job; for example in my homebrew directly heated WE-300B class A power amplifiers for my all-homebrew stereo system, I use zero NFB, including in the phono-playback preamp. No voltage feedback anywhere! The RIAA EQ is accomplished with a passive split RC network, and not with an RC feedback loop.

For real Hi-Fi audio, (and not amateur radio purist AM transmission, where these things don't apply), feedback also kills the dynamics of the sound (think of an audio compression loop), and to my ears at least, imparts a "lisping" quality to the audio. This "lisping" may be due to time domain issues, as the feedback is always attempting to correct something that has already occurred.

[KA2DZT]

Never could get the NFB concept,  I always thought the FB was behind the signal or in effect occurs after the signal.  Anyway, thought it just may be something I'm missing.

Very surprised to read the last line in Bruce's comment, "as the feedback is always attempting to correct something that has already occurred".

Fred