What is causing this audio highs attenuation/ distortion problem?

[WD5JKO]

  One thing about NFB seldom discussed is what happens when an amplifier with NFB is driven into overload. If there is 10 db NFB, then during the overload portion of the waveform, the amplifier goes open loop...meaning that the gain is increased 10 db during the overload. This squares off the overload more then if there had been zero NFB, and also..the worst thing is with R-C coupled stages within the NFB loop, there can be a sudden bias shift from grid current. Then after the bias spike, there is an R-C delay based upon the circuit R-C time constant. Imagine those 6550's in a Ranger with no driver transformer. Folks often have .1 uf coupling caps, and a 470K grid resistor...here the time constant is 47ms, and it might take 100ms or more to audibly recover from a single transient overload..this is VERY audible; much worse then the overload itself.

   Most of us running AM do not have precision level controls, and compressers almost always overshoot during the the "Attack Time". We need some sort of limiting following the compressor. When we inevitably have that occasional voice peak that overloads, the best way to limit the carnage that follows is to limit the NFB around the stage overloaded to 3-6 db, and to have a low impedance grid circuit to minimize any DC shift following the transient overload. An interstage transformer usually considered as a "has to go item" has the benefits of supporting class Ab2 operation, and a low time constant should the amplifier have that occasional overload.

Jim
Wd5JKO

 

[N2DTS]

That looks about 8kc wide, which should be good.

I always thought you could have a great sounding signal without the audio going 20kc wide.

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.

[N2DTS]

Very interesting stuff.
I always thought the cleanest audio amps (tube) ran AB1.

And AB1 can be very clean with some tube types.

With modern stuff, or special (very big) tube drivers, AB2 or class B could be ok.
The easy way is to drive tubes that take grid current with an 8 ohm to say 4000 ohm center tapped Hi FI transformer.

You can also drive AB1 tubes with a transformer (my 4x150 mod deck is set up that way), and there is no feedback.


I also always thought it was a waste of time to run 200% modulation.
Phase the audio right, run it up to 90% negative and something like 120% positive.
Anything else is just distortion, no?

Some people on the air do the big carrier and do not push the modulation.
I remember when FM broadcast started. We had WYSP in Philadelphia, who played great music, and the DJ's spoke softly and at reduced volume. It was wonderful.

I guess its good to want to be very clean and hi fi, but I think its all in the EQ and the voice.
You can have a great sounding signal with some distortion (plenty really) if the EQ is good or the voice is a radio voice.

One of my favorite signals was a guy over in PA, forget the call, but he was running an old Sure S40 mic into a stock 32V3.
Sounded like god.
 

[VE3AJM]

In my previous post in this thread I was asking a question regarding Bruces comments here as quoted by Tom JJ about the 3-500z and then later, about the 3-1000z:

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

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.

And later as I read on this:

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.      

I understand that the 3-500z is a zero bias class AB2 tube and the 3-1000z is zero bias class B triode. I have designed and built a class AB2 3-500z modulator for my TMC GPT-750.

Just to clarify things with where I was coming from with the question. I have enjoyed reading this thread as it has gone along of course. Thanks.

Al VE3AJM

[K1JJ]

Good news...
 
The hard work is paying off.

After a number of tests and changes, the key was to switch the NFB tap on the mod xfmr from the secondary winding, to the  primary. I had wanted to include the mod transformer in the NFB loop, but it was a mistake. (in addition to the mistake of using big bypass and coupling caps)

I ran some new audio sine sweeps. The audio now sweeps cleanly up to 10K at 100%++ modulation !  This is drawing the normal 400 ma – same current as the lower freqs.   It will actually go to 12K, but only hit 80% modulation cleanly before there is excess current and slight crossover distortion begins.  The triangle wave looks even better now at 5Khz.  
 
So, now Fabio II will sweep cleanly from about 20 Hz to 10K at 100%++  mod -      and cleanly from 10K to 12K at 80%.  Not bad.  Still working on refinements for more NFB and this may get better..  I BE HAPPY!!
 
Looks like tapping the NFB off the secondary was causing all kinds of problems with the high end.    Funny how one mistake like this can cause so much effort to find...  
 
I still have only 7dB of NFB and working on increasing that now.   I actually ran out of NFB pot adjustment and need more NFB signal to work with.  My goal is to run it up to 15 dB of NFB and then back it off to about 9 dB, if possible without instability.  With the mod transformer phase shift no longer in the loop, this  may be possible.

T

[N2DTS]

I do not understand.
How are you doing the feedback now?

What happens if you do not use any feedback?

[K1JJ]

I do not understand.
How are you doing the feedback now?

What happens if you do not use any feedback?

Hi Brett,

The negative feedback is derived from one 4-1000A modulator plate cap to a 1.3 Megohm  resistor ladder that feeds the first stage of the solid state audio driver, 180 degrees out of phase.

Before, I had the NFB tap on the secondary of the mod xfmr. This added some phase shift problems. Though, if the feedback ladder was compensated for the mod xfmr phase shift, it would probably work FB.

If I disconnect the NFB, the audio is not as clean sounding. I did not do detailed sweep tests with or without NFB yet, but suspect it would show a cleaner output, especially in the highs where I've had trouble before.

The NFB also lowers the tetrode's source impedance approaching a triode, if the NFB is high enough. This may help when modulating a capacitive load, though I was able to reduce this C to about 1/4 of original via the cap reductions mentioned earlier in the thread.

I am presently at a stable 11dB of NFB  -  and just ordered some 47K, 3  watt, 750V resistors to build a new NFB ladder that will be safer and allow more NFB. I'd like to see 15dB, and then backed off to about 9dB, maybe more, depending on stability.  The audio now is as clean as any plate modulated transmitter I have ever built. Very pleased with the crisp highs and triangle tests at key audio freqs.

T  

[WD5JKO]

Tom,

   "I am presently at a stable 11dB of NFB"

This is a good thread and an excellent opportunity to overload the modulator on a voice peak to see how things behave as you vary the NFB up and down. This means watching the modulated envelope on the scope, listening to recovered audio, and looking for bias shift in the FET audio driver stages.

Some folks would just say, don't overload the modulator, ever. These are sometimes the same folks that run wide open, with no limiters, or clippers. Most compressors overshoot during the attack time too. A whole lot of us punch the audio such that we overload at least a little bit on every syllable spoken...I contend that a lot of NFB used around the stage being overloaded can lead to bigger problems like transient intermodulation distortion when coming out of the overload. You might be surprised with the results. There should be a happy point where the NFB cleans things up a bit, and still allow the occasional overload on a voice peak without having an ill effect.. In some of my gear, that level of NFB is around 6 db.

Jim
Wd5JKO

[N2DTS]

I would think there would be problems taking feedback off of only one modulator tube....
That leaves the other one out of the loop, no?

[K1JJ]

Hi Jim,

Thanks for the heads up.  You may have saved me from blowing the mod xfmr eventually. I am testing at only 2KV to minimize errors, but still.

Think I will leave things as-is other than building a better NFB ladder. I think I have plenty of headroom right now to back the NFB off 6 dB and end up with 6 to 8dB actual.  I will run some tests to find the sweet spot you mentioned from overload.

BTW, a good friend of mine saw your post and sent me an email that agrees wholeheartedly with you.
Thanks again, OM.



My friend's comments on excessive NFB:

Jim, WD5JKO is right about the NFB amount.  6 - 8 dB max before you get into TIM issues which will blow up mod iron pretty quickly!  This is a transient response due to the NFB overshooting it's correction when the waveform abruptly changes shape and frequency as happens with voice or especially music.  Looking into an inductor like a transformer makes it worse.
 
I've been able to wrap 12+ dB of NFB around a transformer, tapping off the primary as you are doing now and have had severe transient problems even though sine waves and frequency response were excellent.  Like all things in this world, there is a trade off and the one here is between steady state response when you run sine waves vs. speech which has lots of opportunity for transient generation with NFB.  
 
I'd stick with the 6 dB and be done.  Much safer for reliable performance and it cleans up 95% of the audio.  No one will hear that last little bit and in fact too much NFB will sound awful!  If the mod transformer survives!  BC rigs rarely used NFB and dealt with the resulting 3 - 5 % distortion.




Brett, one tap off the plate cap is all we need for full NFB due to the auto-transformer effect working against the center tap.  It was explained to me by a friend:


"You do get a complete signal from each plate on the primary of the mod transformer, one is out of phase with the other but totally complete.  To prove this to yourself, just look at the divided sample from the modulator plate on your Oscope.  If you have a dual trace scope, take the same divided down sample from the other plate and put it on the other channel and you will see the same complete audio waveform exactly out of phase.  If there is some small distortion on one half of a sine wave, for example, you will see it on both traces from each plate so you are losing NOTHING when coupling from one plate."

[K1JJ]

   Most of us running AM do not have precision level controls, and compressers almost always overshoot during the the "Attack Time". We need some sort of limiting following the compressor. When we inevitably have that occasional voice peak that overloads, the best way to limit the carnage that follows is to limit the NFB around the stage overloaded to 3-6 db, and to have a low impedance grid circuit to minimize any DC shift following the transient overload. An interstage transformer usually considered as a "has to go item" has the benefits of supporting class Ab2 operation, and a low time constant should the amplifier have that occasional overload.
Jim
Wd5JKO

Jim, I just got a chance to really absorb your post back at the top of the page.

I can see now how dangerous for the mod iron it is to run > 6dB NFB and not to have a limiter after the compressor, etc.

I'm moving on to building up a small negative peak limiter using an op-amp and diode. This works well for me in the past.  But what about the limiter for the positive peaks?  I've been axing around for opinions and they vary widely.  I already have a Berhinger 6-band processor and it's in the cellar... don't want all the bells and whistle and menus.

Right now I'm using just an RE-20 mic > 528E voice processor  > 31 band EQ.  Before I search around the web, is there a simple circuit, like the op-amp NPL, that will limit positive peaks effectively?   I don't want a multi-band processor nor a Volumax with transformers...

Any ideas?

BTW, you mentioned the benefits of having a low impedance grid circuit... do you think the GFZ SS driver circuit qualifies in this area or do I need to take some more precautions with the grids?   I can post the schematic if you need.


T

[Tom WA3KLR]

Unfortunately I have no time to spend to write this better,  My two cents, written while the previous two posts were being written:

The two modulator plate signals on the primary should basically contain the same signal, one is 180 degrees from the other, can’t use both at the same time.  BUT the modulation transformer combines the output of both tubes into one output signal, essentially averaging the transfer function of the low end of the tube curve and the high end of the tube curve overlaid together, creating lower distortion, without feedback.  This is why push-pull sounds better than single-ended, if properly implemented. 

The bad thing about the modulation transformer secondary is the leakage inductance working against the sum total of the r.f bypasses and pi-network capacitances to ground.  The phase shift from this L-C is the killer in trying to push the frequency response out on the roll-off point when doing global negative feedback, can’t fight Mother Nature. 

Hi–fi amps that have good performance at 15 kHz is due to the basic frequency response of the amp is out at least 100 kHz. so feedback/phase shift at the audio frequency is handled well.

[K1JJ]

Just bought a 1982 vintage CRL PMC-300A modulation limiter on eBay.   Jeff/ W2NBC has one and thinks highly of it.  It's a simple straight forward circuit according to the manual below.  

It has an adjustable negative peak limiter, positive limiting, density and a few knobs that makes Jeff always sound like a million bux.  That's good enough for me... :-)


T

http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CCYQFjAA&url=ftp%3A%2F%2Fftp.orban.com%2FCRL%2FPMC300A%2520Manual.pdf&ei=0YEOU9bcIebMsQSus4Ao&usg=AFQjCNEaQQHlRVZs7f7Zj41q7vGPe_uaTA&bvm=bv.61965928,d.cWc

[Steve - K4HX]

Been using a 300 limiter since the 1990s. It is FB.

[WD5JKO]

Tom,

  It sounds like you have a good handle on things. I'm a novice when it comes to audio processing. For the DIY guy though this article is very informative, and inspiring:

http://tonnesoftware.com/appnotes/speech/speechamp.html

James Tonne is an OM now, and has been refining the concepts of speech processing for a long time. He originally had a QST article on this in 1956:

"Compression and Clipping", James L. Tonne, W5SUC, QST, September, 1956

I'd love to see that old tube type compressor. Surely by now the QST Copyrights have expired.

   On a whim I bought the K7DYY processor for my amplified D-104 UG-8 base. I had high hopes, and expectations. These were all met, but I had to do some serious work to it before I was ready to put it on the air. In my case, the MC-320 element was too hot, and the processor chip clipped on audio peaks when close talking, and also during the attack times. So I had to do my own research, make some changes, and in the end, I love my K7DYY setup. To be fair, it seems that many that use this board are using the Ebay MC-320 replacement elements, and these might not be as hot as a good old Astatic MC-320.  If anyone is interested in what I did, please PM me.

Jim
Wd5JKO

[K1JJ]

Been using a 300 limiter since the 1990s. It is FB.

All these years and you never told me your best secret!  (or I didn't pay attention is more like it)


Jim, I just ran some overload tests.   I did try the NFB at various settings, like 1 to 11dB and could not see any signs of overload transients under heavy voice modulation.  The SS driver stayed stable at the grids too.

Could you describe in more detail what you are looking for and how to make it happen?  Do I need to actually slam the waveform into a severe flat top condition and look for some kind of poor recovery?  I don't have an NPL on there yet, so maybe it's not a good idea to be that aggressive until the PMC-300 arrives....

In the meantime, I set the NFB at 8 dB and will leave it there for now.


Tnx for the processing link above.

T

[N2DTS]

I ordered one also!
My first piece of real audio gear!

I have a vx2000, but the compressor in it has a lot of overshoot.

[K1JJ]

I ordered one also!
My first piece of real audio gear!
I have a vx2000, but the compressor in it has a lot of overshoot.

Cool.  Did you buy the second remaining one from the engineer estate sale?  I got the first one. They looked like NOS, never used.

I read the manual last night and see there is a lot of calibration and adjustment time. Maybe a lot of that can be bypassed.  Jeff / W2NBC says he will guide me through the initial set up and get right to the meat of it.  
He says the NPL works well and the asymmetry positive peaks control will let us dial in say, -95 negative and 150% positive - or anything in between.  (Assuming the transmitter will do it in the first place, of course)

My goal is to run a fat carrier and adjust the audio for about -95% negative and let the positive peaks go naturally where they will....maybe 110% - 120%. Run the modulators as cleanly as possible in AB1.  This limiter is exactly what I need here to keep the mod transformer and other parts safe from negative peaks and positive peak spikes. (and hopefully reduce NFB transients as discussed)


Here's a few comments to get you excited:

" I will say that the CRL is virtually transparent with the amount of negative peak limiting in direct relationship to the amount of drive applied and the recessed "output" pot . The RATIO between positive peaks and negative clipping is adjusted with the "asymmetry" recessed pot.. there is no actual positive CLIPPING .. its a all pretty seamless.. I can , for instance get as much as 150% positive and 90 % negative....   I run it using mostly natural asymmetry and about 130 % with 95 neg... you can crank it all up but does NOT swell the background or get noisy at all..

The cool thing about this "processor" is that it is a superb NPL but with the ability with the "asymmetry" control to limit positive peaks as well. There is also a built-in low pass filter and with simple resistor changes limits at 6, 8, or 10k .. No other frequency limitations or phase rotation, etc. They are pretty cheap on Ebay, and not a bad box to have around.."

[N2DTS]

I do not remember where I got it from (info at home), but there were a lot for sale.
I did not get the one on ebay because they did not say they tested it.
$240.00 for the one I got.

I just want to be able to limit the negative modulation to about 90% without overshoot.
The passband filter will be nice also, I think I will use 8Kc.

I read the manual, the only odd thing was the bit about using the mod monitor positive mod alarm light to set things up.
Lucky I have a mod monitor with the alarm lights....

[K1JJ]

They also talk about "old" AM transmitters that use plate modulation.  They have a "phase correction" for both the highs or lows.  I've seen this before and wonder if it is any value if the rig is already running well.

Did you see their examples of what a desired RF modulated waveform looks like? You're supposed to adjust the phase controls to mimic it. Problem is it looks like a flat topped / cross over distortion wonder.


Tip of the day  

For those of us using RF drivers that have tube finals:  (riceboxes or anything now in linear mode to drive your class C final)  I found my FT-102 did not like to key a carrier for more than a minute or so, otherwise it started to act funky from the heat.  I went in and adjusted the PA bias so that the final now idles below cut-off. BIG difference!  Now the tubes run cool and can generate a carrier indefinitely.  I am looking into adding a grid leak resistor to the bias to put it even farther into class C.  I use this FT-102 just as a driver for the class C Fabio II and the PDM Dual Quads rig, so it can stay heavily biased.  As long as there is enough internal drive in the FT-102 to overcome the additional class C bias, it will work FB as a dead carrier driver.  (It is no longer linear and not suited for ssb when in class C)

Update: In my FT-102, by padding R-109, a range resistor for the  fixed bias trimmer adjustment with a 15K resistor, I was able to get the bias to adjust far into class C.  The trimmer still has the range to pull it back into class AB2 if desired.  So your rig (whatever it is) may not require grid leak - rather just a little more adjustment range via fixed bias to become class C efficient as a driver.
T

[N2DTS]

Yes, I read that, and figure maybe with thing adjusted for calibration like they say it will look like that?
Sure would not want to SOUND like that!

[N2DTS]

I did some tests on 3 rigs tonight.
The 2x 4x150a rig did -30 db 2nd harmonic audio tone at 99% modulation, as did the 3x 4D32 rig.
811a modulators, common supply, has some carrier shift under modulation.

The 2x 813 rig modulated by 4x150a's did 40 db.

All the transmitters did better with more grid drive then I normally use, and the 813's did better with less screen current.

The 4x150 and 4D32 rigs started rolling off at 40 Hz, went 10% modulation less at 4000 Hz, but stayed flat to 18,500 Hz (high as I went) at that 10% down. Why it goes down 10% at 4000 Hz is a question.

The frequency response tests I did at 60% modulation, so at about 4000 Hz, the modulation dropped to 50% from a 1004 Hz reference level.

I think the 813/4x150a rig did well because of the separate power supplies and there being no carrier shift at all.
The modulator is also AB1.

[K1JJ]

-30 to -40 2nd HD is pretty good.


Try loading the final heavier and see if your THD improves. Decrease the C2 ant loading cap, which will mean a bigger carrier and require more audio. (more audio may erase the THD gains, however)   

I have found that less C2 (heavier loading) helped my THD get better as a result. There is a sweet spot.  A couple other guys have told me the same. You wud think it was the opposite, but heavier loading seems to help clean up a class C plate modulated final.  It helps with linear amps too, of course.

T

[N2DTS]

What about loading it up more but with less plate voltage?
I have the ability to adjust all the voltages, and I adjusted the modulator plate voltage without any change.

[K1JJ]

Generally, I like to run as much plate voltage as practical.  Most of my rigs really come to life once they hit a certain voltage level. This assumes the biasing, loading, etc., is right.  In general, higher voltage with recommended current seems a good  plan.

But the best way to find out is to run your various tone tests and try every change you can...  tank Q, voltages, loading, drive, etc. You will find the sweet spots that can sometimes eek out another -10dB of cleaner THD.

T