So, just how bad *ARE* transformers driving class B grids???

[steve_qix]

THIS bad!

I've been fooling around with a Navy TCS-12 transmitter I've owned for almost 40 years.  Way back then, I had modified the modulator (a pair of 1625s) and rewired the tubes to triode connection.

Ok, anyway my son might bring this to the college ham station at UMaine so we figured we'd check it out and make ready.

So, I dug out an old solid state audio amp I had lying around, and hooked it up to a backwards connected 50 watt audio output transformer.  Oh, this is a VERY good transformer - an "ultra linear" (with screen taps) transformer out of an old heathkit audio amp that used KT88s.

Anyway, hooked all of this stuff up, transmitted into the dummy load and observed (and listened to) the modulation using the modulation monitor.  Definitely could hear some distortion, although the frequency response was pretty good.

Then, we put a triangle wave into audio driver.  Distortion was evident on the signal, and upon further checking, the signal at the tube grids was also distorted.... I assumed the audio driver was at fault, but lo and behold, it was the TRANSFORMER.

The picture below shows the audio at the transformer secondary (top trace), and the transformer primary (lower trace).

There is quite a bit of distortion at the secondary, but the input signal looks perfect.

Anyway, even a good transformer has trouble looking into the varying load of class B grids.  I tried loading the secondary with resistors.  It helped a little, but the distortion was still there.

The next step is a source-follower driver for the grids which will eliminate the transformer distortion.  Interesting, to be sure.

[N2DTS]

That does not look bad.
Any transformer will have problems with sharp angles in the waveform I would think.

[w1vtp]

Steve

Yes, the varying load of a class B amplifier does affect distortion as you point out.  One possible work around would be to calculate the peak voltage (or experiment with different settings) needed to drive the grids and using a multitap driver transformer such as a UTC 59AX to keep the source impedance as low as possible while providing sufficient peak voltage for the grids.  By doing this balancing act source impedance could be kept low as possible to minimize distortion due to the varying load presented to the driver amplifier.  Or one could consider AB1 and using a phase inverter circuit - but I imagine this would be beyond the scope of your present project

While, again, this might be a bit farfetched for your current project it might be worthy of consideration for that major hollow state xmtr by those who are planing such a project.

Al

[w4bfs]

hmmmm .... yes, I think I see what Al is speaking about and the positive peak running into the lowering grid impedance .... I calculated the minimum driving impedance for hi-mu triode connected 813 at 384 ohms minimum .... your 50W tranny is likely higher Z which would aggravate the problem ... of course a source follower at just a few ohms z is a better driver

the curious thing that surfaced from calculating the hi-mu connection is that everything from a 6aq5 up to the 813 wound up with similar minimum drive impedances of 400 to 1k ohm

the fet source follower is fb until the tube burps and then unless well protected is history

[steve_qix]

That does not look bad.
Any transformer will have problems with sharp angles in the waveform I would think.

It's really pretty bad     The transformer has no problem passing the frequencies involved (it was only an 800 cycle triangle), the problem comes about when the grid really starts conducting, and the transformer losses show up causing the non-linearity in the output waveform.

I figured I could get away without using a MOSFET driver, but obviously one is required if I want good performance.

It is possible (and practical) to design a bullet-proof MOSFET driver these days.  Even if the tubes shorts grid to plate, the MOSFET will be unaffected by using a few simple protection devices (resistors, zeners, etc. in the gate side and good 3kW transzorbs and diodes in the output) along with a fast blow fuse in series going to each grid.

Bipolars were not good, and early designs such were used in the Gates BC1 series were very unreliable.

Oh well - I did have hopes for the transformer!

Regards,  Steve

[Tom WA3KLR]

All above presents a good argument for pairing plate-modulated PA's with modulator tubes that can provide the necessary audio power in class AB1 only.

[w1vtp]

That does not look bad.
Any transformer will have problems with sharp angles in the waveform I would think.

<snip>

Oh well - I did have hopes for the transformer!

Regards,  Steve

Not with that trannie  

Your Mosfet solution is the elegant one.  While you're at it, I need a MOSFET driver for my Viking I  

Al

PS:  I just noticed the premise of your experiment: Was that a push pull AF transformer?  Don't you need to have a mod reactor with a blocking cap (aka the misnomer "modified Heising")? If not, might that not be a source of problems?

[Opcom]

Using the 'special class b' connection?

What is the peak grid (G2) current for each 1625 tube during this issue? That may be revealing when looked at compared to the DCR of the transformer used for drive.

What does the current waveform from the solid state amp look like in comparison?

When you push the special class B connection hard, and possibly bring the plate voltage down to where the screen can really conduct, then you get the thing you see. You could use a transformer with lower winding resistance. Fine audio iron may have more resistance than is good for driving high current grids. I don't know if this makes sense.

[N2DTS]

I have both AB1 and class B modulators, and its hard to tell the difference in the mod monitor, let alone at the far end.

Could anyone actually HEAR that distortion?

All above presents a good argument for pairing plate-modulated PA's with modulator tubes that can provide the necessary audio power in class AB1 only.

[steve_qix]

I have both AB1 and class B modulators, and its hard to tell the difference in the mod monitor, let alone at the far end.

Could anyone actually HEAR that distortion?

All above presents a good argument for pairing plate-modulated PA's with modulator tubes that can provide the necessary audio power in class AB1 only.

Non-linearity distortion in a modulator can definitely be heard if the receiver is good enough.  I notice it more if someone is really CLEAN, because it is not so common.  

Anyway, the ultimate answer to the problem is to just build a source-follower audio driver and that will straighten out the drive problem..

[KI4YAN]

That is TOTALLY not fair to the modulator.

You're using a transformer designed to present a relatively constant 3500R or so load to a pair of KT88's.

It's working into a load that swings from high impedance to 380R...it certainly is going to distort!

Use a proper, gapped, driver transformer and watch that clean up considerably. Make sure your bias supply can source 2X the calculated required current, and that will also help considerably.

[K1JJ]

I continue to use Frank / WA1GFZ's  MOSFET audio driver for the 4-1000A modulators in my homebrew plate modulated rig.  It has provisions for NFB.  Real pleased with it.  


The prototype:

http://amfone.net/Amforum/index.php?topic=23632.0


Steve, didn't you design up a direct coupled SS driver a few years back?  

T

[IN3IEX]

I think that classes with positive control grid voltage can be driven only by dc coupled cathode followers. Alternatively just stay AB1 by increasing screen grid voltage of the audio finals and giving more negative bias to control grids. Triode connection may be the origin of the problem, have you tested ultralinear connection? G2 driven by a voltage divider between anode voltage and a suitable V+?

[WD5JKO]

   I once used a Citation V audio output transformer (rated 40W down to 7 hz) as a reverse connected modulation transformer for a Viking I transmitter. Of course I had to use a choke, and cap with it to keep the DC current off that transformer. The driver was a big Crown SS amp, an old M600 (600W RMS). The transformer was transparent...what goes in comes out. Triangle / and Square waves were very good. The difference is that high end transformer makers of the day used special winding techniques to minimize turn to turn capacitance, and had methodology on how to minimize leakage inductance between primary and secondary. So at the mid band frequencies, these acted much like a direct coupled amplifier. Transformers such as this one are expensive to build, and I doubt few makers (including Hammond) offer such a product today.
  
   I have built several SS drivers over the years. One thing I learned is that an amplifier that can pass any waveform without distortion from say DC to 10 Khz may have low Total Harmonic Distortion, but may also have high Intermodulation distortion. With complex waveforms such as speech, some of these amplifiers can sound awful. This gets into the Solid state versus tube arguments where a tube amp with zero feedback and 5% THD can sound "better" than a Solid State amp with 50 db negative feedback to achieve .001% THD.

   One thing for sure, if you made a modulator that was flat from DC to 10 Khz, and used a microphone such as a condenser microphone that is flat over a wide audio range, then without any equalization (a form of distortion), the audio reports you will get from folks with a receiver IF bandwidth  of 6 Khz will say your audio is muddy, and with too much bass.

   My point is that in this thread we are demonizing a transformer in the audio path. I contend that many setups with any form of equalization will also fail the sine/square wave test even if the audio driver behaves absolutely perfect.

   I once built a audio driver with a pair of 6B4's driving a quad of 808's class B. The driver transformer was from a Globe king 500 (these are a P-P input tranny originally used on the GK 275/400). The steep step down ratio combined with a low plate resistance driver tube did pretty well. After adding about 6 db NFB (balanced Push Pull NFB to the 12BH7 driver cathodes), the results were really good. This driver really had good load / no load regulation. The 808's were biased at about -40v (med Mu tube), so the transition from no grid current to grid current is in an area where any driver loading would result in a highly visible "kink" in the driving waveform. There was no kink, or slope change as grid current commenced.

    With an imperfect driver transformer to class B grids, a step down ratio will do better then a step up ratio.

Jim
Wd5JKO

[steve_qix]

Lots of good discussions....

But, let's bring it back to the original purpose - which was simply to illustrate that a reverse coupled audio output transformer driving class B grids is a bad idea - and a lot of people do this!

It would be an interesting experiment to test an actual driver transformer, but I suspect the results would not be much better, and at that point we would be adding another variable to the mix and that would be the driver itself and what is happening there - apart from the transformer.

As far as successfully passing an 800 cycle triangle wave, really, a modulator on its own and apart from equalization etc. should be able to do this.  If it can't, it's no big deal but it does point to distortion.  This is the test I use for every modulator I build or use.

Anyway, I'm probably going to build up a source follower driver, which should be a WHOLE lot smaller than what I'm using now - and will certainly work better.

The experiment was purely an exercise in laziness - I had the stuff lying around, hooked it up and gave it a try 

Regards,  Steve

[VE3AJM]

Has there ever been any discussion on here regarding the guys running the class E rigs on 3875 every evening and how broad/bad that their signals are, let alone some of their operating practices/power used?

I've noted bandwidths in excess of 30kc+ with quite a bit of peak distortion and audio spurs/products up and down the band. There are no audio transformers used with those transmitters.

Al VE3AJM

[AB2EZ]

As a (perhaps) minor clarification regarding AC coupling:

I use a pair of source followers... between the electronic phase splitter and the grids of the 6550s... in my Ranger. This phase splitter substitutes for the Ranger's stock audio interstage coupling transformer.

I AC couple the source followers to the grids of the 6550s using 10uF capacitors. The grid-to-ground (or bias) resistors are each 10k ohms. Therefore (in the absence of grid current) the RC time constant is 0.1 seconds ... i.e. a 1.6 Hz low frequency 3dB cutoff.

http://mysite.verizon.net/sdp2/id13.html  

[Note: in this schematic, I show 100uF coupling capacitors... which worked fine... but I subsequently changed them to 10uF to take into account that the Ranger's 300V (nominal) power supply... that delivers power to the phase splitter circuit... has a small step down (decrease) in voltage when the Ranger is switched from standby to operate. This causes the grid-to-cathode bias on the 6550s to become slightly more negative during the time it takes for the coupling capacitors to discharge to compensate]

With normal voice operation, the grids of the tubes are driven positive relative to the cathodes only on voice peaks. These voice peaks may cause a small amount of charge (change in voltage) to accumulate on each capacitor... but it will leak off with a time constant of 0.1 seconds.

Therefore, with this long time constant in the RC coupling, the effect of using AC coupling (rather than DC coupling) is not noticeable... either subjectively (in terms of how the audio sounds) or objectively (in terms of the observed waveforms in normal voice operation).

The effect of the AC coupling would only be observable with a periodic input signal (like a sine wave or a triangle wave) that is large enough to cause grid current on each cycle of the periodic audio waveform. The grid current (which, of course, only flows into the grid) pulses on each positive peak would cause charge to accumulate on the AC coupling capacitors. This would result in a corresponding decrease (more negative) in the grid bias... which would reduce the average grid current to a small enough value to allow the 10k ohm grid resistors to deliver it.

Therefore, the long-time-constant AC coupling acts like an slow acting, automatic bias-adjustment control when the input audio signal is periodic; but has no noticeable effect on normal voice audio.

Using this approach, one avoids the complications associated with DC coupling.

Stu

[steve_qix]

Has there ever been any discussion on here regarding the guys running the class E rigs on 3875 every evening and how broad/bad that their signals are, let alone some of their operating practices/power used?

I've noted bandwidths in excess of 30kc+ with quite a bit of peak distortion and audio spurs/products up and down the band. There are no audio transformers used with those transmitters.

Al VE3AJM

This is strictly operator related and the transmitter is not at fault.  Any transmitter, no matter how good it is - if operated incorrectly or otherwise pushed into distortion will generate undesirable artifacts.

I've heard wide and distorted flex 5000s on the air, but it would be rather unproductive if I stated that flex 5000s are unduly wide because I've heard some that are.  I've heard wide DX100s, Rangers, Johnson 500s, homebrew 4-400 rigs, etc. etc. etc. - you get the idea.

The problem is unrelated to the underlying technology and I think we all know this to be the case.  

The [intended] purpose of the thread was/is to relate findings about a technical feature which a lot of people have used or are using now  

[VE3AJM]

Perhaps then, what should be done is to start a thread on how bad some class E transmitter operators are then. These guys can't be giving the class E transmitter community or AM in general, a very good name with their antics. This commercial is on the air every night for hours.

There is a strange smugness/unaccountability there, that they can do no wrong, especially when they are told on the air that their transmitters are putting out spurious emissions and are excessively broad in their audio bandwidth and using 150% modulation etc., which does not by and large exist in the AM community that I am in contact with.

Al VE3AJM

[N2DTS]

That is somewhat of a problem with very powerful very high fidelity transmitters.
Unless you limit (or at least try and limit) the bandwidth, it gets very wide.
With iron and glass, its not easy to get the response way out there other then using broadcast equipment.

We should have a thread about simple effective ways to limit bandwidth for times when that is called for.

My back to back EQ units do not seem to do the job very well, with 36 db cut of the highs....

I have heard some software audio processors can be set to limit the bandwidth, but other than that, it seems old expensive broadcast audio processors are the only good way to go.
Or is there another solution?

[AB2EZ]

Yes

If those guys... who are running 1kW at carrier and 150% on modulation peaks (6.25 kW PEP)... with 20kHz bandwidth audio chains and lots of compression (but insufficient, or no post-compression bandwidth limiting)... only had some good technical information at their disposal, I am sure that they would immediately take steps to limit the interference they cause to others up and down the band.

Stu

[KA8WTK]

Back to the original subject .....
 
 I was looking at the class B modulator information in Orr's Radio Handbook (18th edition). In his write up it shows a 20K resisitor in series with each of the grids, screens connected direct. He says that this is to balance the current drawn by the grids.
  Are you connecting the screens and grid together like you would with an 813? 813's would not need the resistor accoring to the text. You might try inserting the 20K resisitors in line with the grids, as Orr suggests, and see if that helps.

Bill KA8WTK

[w1vtp]

That is somewhat of a problem with very powerful very high fidelity transmitters.
Unless you limit (or at least try and limit) the bandwidth, it gets very wide.
With iron and glass, its not easy to get the response way out there other then using broadcast equipment.

We should have a thread about simple effective ways to limit bandwidth for times when that is called for.

My back to back EQ units do not seem to do the job very well, with 36 db cut of the highs....

I have heard some software audio processors can be set to limit the bandwidth, but other than that, it seems old expensive broadcast audio processors are the only good way to go.
Or is there another solution?

Yeah Brett - with the HIFI AM station comes the responsibility to monitor one's BW.  I run class E here and have taken the time to look at my sideband products and they roll off as they should with a FC of ≈ 5 - 6 KC.  

I agree that it would be productive to have a thread on proper BW management of one's AM signal.  One real good way is to monitor one's modulation level.  Levels of positive peaks of 150% modulation does NOT necessarily mean excessive bandwidth.  I have been monitoring my spectral occupation for a long time and have never seen a relationship between positive peaks of 120 - 150% and excessive BW.  Running processing improperly with the view of increasing one's "modulation density" can be a problem but does NOT mean that BW is necessarily going to be excessive. It will result in higher average of sideband power going out to the outer limits of channel occupation.  If one is going to run HIFI AM, a good modulation monitor is an important part of the station and should be a "must have."

I do not agree with a broad brush approach that postulates that all class E operators are occupying too much bandwidth. Nor to I agree with generalizations that serve to polarize the AM community. What we need is a gentleman's approach to our actions on the air and if we are going to operate a station that uses modern technology we should also secure the means to monitor our spectral purity.

The original thread as it started was a good approach to low distortion using high level plate modulation.  I would like to see this thread get back on track to a fruitful discussion of this topic.  So far, I have gleaned that low source impedance is important if we are going to be using AB2 / B class modulation or that we need to go towards AB1 that uses mainly a voltage source as a driver.  I would also like to see the thread touch on the benefits of negative inverse feedback as a means to keep distortion products to a minimum.

Steve's method of using a triangular source to have a good idea whether or not distortion products are being generated as the results of this typical change of load using class B in modulators or other circuit design changes, for that matter, is a good one.  What combination could we use utilizing all the good design methods where that triangle coming out of the modulators would be the same as the input of the modulator and what details would benefit the AM community in future homebrewing a low distortion (and hence, narrower bandwidth) modulator?

Al

[AB2EZ]

Steve

Getting back to the main topic (as appropriately suggested): looking at the pictures in your original post...

The output voltage waveform on the secondary side appears to be tracking the primary side voltage waveform more closely on negative peaks of the triangle wave... v. positive peaks of the triangle wave.

Since the push-pull modulator output circuit is symmetrical, I wonder if at least some of what you are seeing is a result of something other than just grid current flowing into the push-pull modulator output tubes

P.S. On further thought, perhaps you are measuring the transformer output voltage from one side (one grid) to center tap (ground)... in which case, the output waveform would not be symmetrical with respect to the baseline.

Stu

[steve_qix]

I've seen (and tried) the resistor-in-series-with-the-grid model for some triode connected tubes in the past, and at least from the measurements I've made, it didn't work all that well, and was in fact detrimental to a linear modulation characteristic.

This is definitely germane to the topic.  I had a Viking II here for a while, and designed a direct coupled, solid state driver for the rig.  The Viking uses 807s which are pretty much electrically equivalent to the 1625s that are in the TCS (which started this whole discussion off in the beginning).

What I found after much experimenting, is that an approximately 60 to 80 volt (DC) difference between the control grid and the screen was optimum.  I used a string of zener diodes bypassed by a suitably large capacitor to achieve the difference.  Unfortunately I took the setup apart and didn't document exactly what I did, else I could give more exact numbers.

Both grids were then driven with audio, and from the same audio source (the direct coupled driver, in this case) with the DC potential between the grids maintained by running a small amount of DC through the zener string using resistors to the plus and minus power supplies.

I could achieve near broadcast quality from the Viking (using the stock mod iron) with this setup (and negative feedback around the whole thing).

But, I did start off with the resistor-in-series-with-the-grid setup, and the resultant audio was "grainy" (for lack of a better description).

Stu - yes, the waveform distortion is present on both sides of the push-pull, and it is virtually identical on each side.  The distortion only shows up where the grids are conducting heavily.  Otherwise, the triangle is clean.  As you pointed out, it is NOT showing up on both sides of the waveform when 1/2 of the push-pull is observed, even though the distortion is in fact happening on both sides (out of phase, of course).  This indicates that the transformer losses are specific to each side of the push-pull individually, and the waveform distortion is not being coupled between each end of the push pull output.

Al, thank you for the wise comments.  I can't agree more with your assertion that division within the AM community is very bad, and attempting to single out a particular AM technology is highly counterproductive.


Now, a disclaimer would probably be appropriate here.  I am probably what one could call obsessive about audio quality, and this is NOT necessarily a good thing !!  For instance, the TCS with the distortion, etc. sounds perfectly acceptable, and in fact with tests on the air, and in QSOs where I was specifically asking how the rig sounded, and expecting critical reports, only a couple of people noticed the distortion at all. However, it is absolutely evident listening off the modulation monitor.

Back about 25 years ago, I had a saying attached to my email footer - "Good is the Enemy of Better".  Well, one of my friends teased me about it and made HIS footer "Better is the Enemy of Good Enough".  You know what?  He was probably right !!  The older I get the more  this seems to be true, although it is very hard to let go of the "illusion of perfection", particularly when one writes software for a living!!!!!!!!!!!!!!!!

Happy New Year to everyone.    Regards,  Steve