NFB Question

[W8UJX]

I am trying to continue my education on improving the quality of the audio on my Globe King 500A modulator.  The circuits have no NFB to improve response and reduce distortion.

I want to experiment with NFB but I am not sure how much to add or where to install the loop.  I would think the FB should be fed back to the cathode of the 6SJ7, the first audio amp. 

There are three points I could source the FB.  (1) The plate of the 6L6 which drives the grids of the 811A’s through a transformer.  (2) The plate of an 811A. (3) The top of the modulation transformer.

How much feedback should be applied or would it depend on the beginning and end points of the loop?

Thanks,
Jerry, W8UJX

[Opcom]

The top of the mod xfmr is OK, but there may be phase shifts in it that would cause instability especally if several transformers are in the chain.
It is as well to take the voltage for feedback from the plates of the modulator tubes and send it back to just after the next transformer back, so only tubes and RC coupling circuits are in the loop. That is what is done in some BC transmitters. Start small.. 3dB or maybe 6dB of gain reduction (which you have to be able to make up somewhere). It should not take much to start making a difference. Unless someone has mods already worked out, it's an experiment anyway, so have fun with it and post the results!

[KL7OF]

There is a speech amp circuit in the late 40's early 50's arrl handbook that uses 6sj7 to 6sn7 transformer coupled to PP 6L6's...The interstage tx has a split secondary and the neg feedback is tapped from that split secondary...I played with that circuit and tried a couple different interstage  txfmrs..(a stancor ...and the interstage tx from a military tx)When viewed on a scope....the audio was exactly 180 deg out of phase with the audio ahead of the interstage....I was never able to see audio much over 90 deg out of phase when I tapped plates or cathodes for my source of neg feedback.....  The tap used a cap for coupling  and a resistor to set the feedback voltage..The values I used varied from the published values..I had motorboating problems in the audio when it was over coupled and when the feedback was 90 deg out or less...180 deg out of phase feed back works the best for me...I have a couple of these speech amps in use..

[W8UJX]

Thanks OPCOM.

Let me be sure I understand you comments.  The modulator tubes are a pair of 811A’s.  The tube lineup is a 6SJ7 R/C coupled to a 6C5 which is R/C coupled to a 6C5 which is R/C coupled to a 6L6 transformer coupled to the 811A’s

I understand you to say take the feedback from the 811A plate and feed it back to an 811A grid which is after the next transformer back.

This leaves out all the R/C coupled stages before the 811A’s.  If I make sure the R/C coupling is designed to pass a wide frequency range with (hopefully) minimum phase shift I would think it would be better to include the first audio tubes in the loop.

But at this point I’m a dum dum and looking for guidance.

Thanks,
Jerry, W8UJX 

[Opcom]

Thanks OPCOM.

Let me be sure I understand you comments.  The modulator tubes are a pair of 811A’s.  The tube lineup is a 6SJ7 R/C coupled to a 6C5 which is R/C coupled to a 6C5 which is R/C coupled to a 6L6 transformer coupled to the 811A’s

I understand you to say take the feedback from the 811A plate and feed it back to an 811A grid which is after the next transformer back.

This leaves out all the R/C coupled stages before the 811A’s.  If I make sure the R/C coupling is designed to pass a wide frequency range with (hopefully) minimum phase shift I would think it would be better to include the first audio tubes in the loop.

But at this point I’m a dum dum and looking for guidance.

Thanks,
Jerry, W8UJX  

No, in that case, you may have to have the one transformer on the loop.

It would not make much sense to try to apply NFB only to the 811 stage, and difficult because it is drawing grid current too. To make it simple, the stage you apply feedback of this sort to, should not be drawing grid current. So, the 6L6 and 6C5's are good candidates.

One thing to be careful of is that whatever you do, the polarity or phase (as mentioned above) is real important. One 811 plate cap or the other, in your case.

One, the wrong one, will cause unwanted positive feedback and a howl in the modulator (bad for the mod transformer - try it with a 5K resistor on the secondary or something for safety of the transformer), the other 811 plate cap, the right one, will produce the desired effect of negative feedback, which should only reduce the gain. This is determined usually be experiment.

Try going back from one of the 811 plates to the second 6C5 cathode. The things that concern me are the coupling modules between the 6C5's and between the 6C5 and 6L6. I have no idea what amount of phase shift would be in those, if they are used for limiting the audio response, it could vay greatly over a wide range of frequencies and cause instability. To reduce the issue, i chose to go to only the second 6C5, so only one of those things is involved.

In the schematic, I try to show a general way to approach it first. Others may disagree. Before trying it, ask someone who has worked on that set to help, maybe, because i have no experience on that unit, only generally with some high powered audio amplifiers and modulators, but all are different.

some notes for the drawing:
1. The purple + and - symbols indicate the polarity of the signal.

2. The red dots indicate 'like' polarity of the signals at the transformers. These could easily be reverse (see above about feedback). The manual does not say. You may have to scope this out. If it is 'wrong', then swapping the 811 plate caps should sort it out. OR, swapping the 811 grid leads. (not both).

3. The added network at the top is a very general way to try to put 1/3,100th of the 811 signal voltage (0.032% of it) to the cathode of the 6C5.
 - That is, 4,700,000 divided by 1500 is 3,100. more or less.
 - The network is part of a voltage divider from the 811 plate to GND, with a tap at the 6C5 cathode, 1500 Ohms up from GND.
 - (I ASSUME the value of that resistor is 1500 Ohms because the GK500B has a better shcematic and it is 1500 on that. On the bad schematic for the "A", the value can't be read.)

So if the peak to peak voltage at the 811 plate is 1500V, then 0.48V peak to peak will be pushed onto the 6C5 cathode to 'fight' the signal which exists at the 6C5 grid. That is the negative feedback. It should be less than 6dB with this circuit but i can't guarantee that. Start small; it might not need a lot anyway.

a.) The question I would ask, before doing this, is "how much voltage is on the 6C5 grid at 100% modulation?" Whatever it is, the voltage applied to the cathode determines the dB of feedback, more or less.

b.)if the 6C5 grid voltage is 4V peak to peak, and there is 1V applied to the cathode from the feedback, then it is 25%. If the 6C5 runs 2V p-p at 100% mod., then the same feedback is 50%. and so on. Hope there is enough extra gain to make it up. It has to be made up or the modulation % won't hit 100.

4. The network has a 4.7M resistor. A 2.2M will make the feedback stronger. 2,200,000 divided by 1500 is 1466. A 2.2M resistor would put (using this logic, 1500V/1400= about 1V) 1 volt of feedback on that cathode. Note small resistors are not rated for 2-3KV so use a string of 1/2W resistors. Most 1/2W resistors are rated about 500V.

5. The network has a 22pF capacitor. This increases the feedback at the highest audio frequencies and usually helps stability, but might not be necessary. The 2KV voltage rating is because that resistor previously mentioned will have the full peak to peak 811 plate swing on it.

6. the network has a 0.05uF 3KV cap. This is to block the 1000V DC voltage of the 811's plate supply from the cathode of the 6C5. Otherwise it could raise the cathode voltage and cause distortion.

7. the 25uF cap across the 6C5 cathode resistor has to go, because it would absorb the little 1V signal we might try to feed in there with this scheme. If RF is an issue, put a 0.001uF or 500pF cap there instead.

8. Again I have to say I don't know this transmitter, and am not responsible for any results from this general discussion of what I might try to do, myself, as a starting point. If you can find more knowledge or an example of it being done, do so. And don't get shocked by the 1000V or whatever else is in there.

Others on this topic ought to comment on this.

PJ

[W8UJX]

Thanks for the info on the NFB circuit.  I have an idea that may allow me to eliminate the transformer between the 6L6 and the 811A's.  The modulator has four 8-pin octal sockets.

How about using two 6L6's and R/C coupling them to the grids of the 811A's.
That leaves two sockets to build the circuits to drive the two 6L6's.  I would think that would allow the use of more NFB from the top of the mod xfmr back to the first or second speech amp.

I plan to use an outboard preamp for the D104 so I am not worried about having enough overall audio gain.

I would start with 5K ohm plate load resistors for the 6L6's.  I wonder what value of coupling cap and grid resistor to use for the 811A's?  I have upped the B+ on the 811A's to 1500VDC so I would bias the 811A's with -4.5VDC through resistors.

If you think this circuit is worth the effort to experiment with I could build it up on an outboard chassis to minimize the mods to the King's circuits.

Jerry, W8UJX

[Opcom]

811's still require grid current, very hard to get near enough even with elaborate with RC coupling methods, and with the regular or well-known RC coupling method, you will end up with the coupling caps charged by the diode action of the grid current with no simple way to discharge them, likely until the grids are biased so negative the tubes are cut off and no grid current can flow, therefore no drive.

Some people use a cathode follower to drive the grids but that complicates the bias and it is more appropriate for bigger stuff. Others have used MOSFETs as DC amplifiers for this same purpose and it is compact but the same complexity issues remain.

The one transformer between the 6L6 and 811's probably won't make trouble. Trouble comes in when there are:
1.) Too many transformers inside the feedback loop
2.) Transformers having restricted frequency ranges (causing phase shift within the voice range)
3.) Undersized or cheaply made transformers
4.) R-C coupling circuits with excessive phase shift across the voice band
These four factors multiply against each other to make things worse. It's not possible to say what you got there in terms of 2 or 3, it's an experiment most of the time.

For the above reasons, feedback can be generously applied all the way around the many stages of a tube hi-fi amp because it is designed to have a much wider frequency range than it is used for. Those amplifiers use as few audio transformers as possible (usually only an output transformer), the transformers are designed for smallest phase shift and uniform amplitude response across a very wide frequency range and are not cheaply made and often oversized, and the RC networks have large coupling caps to reduce phase shift at all but the lowest frequencies, below where the amplifier is rated to operate. The preceding explains why many hams will built as high-fidelity a modulator as possible. It's not to have a wide signal on the air but to have the lowest distortion and highest stability under most conditions.

One reason for going all the way back to the cathode of a little tube, where small voltages can influence the signal, is to avoid a large and power-dissipating feedback network. This said, you could try to encompass the whole thing with negative feedback but most of the distortion in ham transmitters occurs in the modulator driver and power stage, so there is less point in doing so and possibly upsetting an even more sensitive stage.

Studying enough hi-fi amps and other high quality amps with multiple transformer coupled stages may turn up another scheme, where feedback is applied around each stage separately. There is no reason to do that in this case and it adds much complexity. I think an Eico HF-89? 100W amp has that arrangement.

One way to be rid of the transformer is to change the modulator power stage to something like 4-65's, but that is way lot of trouble, requiring bias and screen supply. Plus, they make only 145W at 1500V in class AB1. I would try something along the schematic suggestion before going to that much trouble. Simple things at least for me have worked better for ham use.

Please keep in mind that even if your distortion is less than a few percent, nearly all receivers have several percent distortion to add to it in the detector, and a decent hi-fi speaker still has 2-5% distortion just in case the listener has tried to hse something better than a 4" speaker in a tin box. Plus QRM. At some point, you are only pleasing yourself. That's fine too.

Lastly, if the modulation transformer and the modulator power supply are not up to it, which is mostly the case with ham gear and seldom the case with converted AM broadcast gear, you may never eliminate all of the visible (scope-visible) distortion at high modulation levels. After pulling my hair out so there isn't much left, I'm satisfied to have any unavoidable clipping or nastiness in the modulator come on softly and gradually, giving an effect like compression and rounded corners instead of sudden clipping and splattery square waves when the modulator is pushed to 100%. This is my opinion, others may differ.

[W8UJX]

Ok.  Thanks for the info.  You saved me the time of trying that experiment.  I'll just play with the method you suggested.

Thanks,
Jerry, W8UJX

[W3GMS]

Jerry,
You have been given some very good advice and I  don't want to muddy up the waters but I have a suggestion. 

When I  go through a transmitter audio system with the goal of improving the audio, NFB is the last part I do.  Prior to doing the NFB, I rework all the stages for the lowest distortion and get the frequency response the best it can be.  Only after that do I apply some NFB and usually it does not take to much.  Doing it the other way is kind of backwards.  I am sure you have a scope and a good audio generation so start with that and go stage by stage optimizing whats in the rig.  Gain distribution is very important and you want to make sure you don't run out of headroom in an early stage driving a later stage.  If your not going to get rid of the driver transformer in favor of a FET source follower, then make sure you use the best driver transformer you can find!  I have never gone through a Globe King but the same basic rules apply as any other rig.     

Its a fun exercise and well worth the effort. You can take it as far as you want from just making it sound better to really going over the top.  First decide what your design goal is.   
73,
Joe, W3GMS     

[VE3AJM]

I agree completely with what Joe GMS has written. The negative feedback additions would only be added after having dealt with the GK500 speech amp and driver circuits, and looking at them a lot more closely.
Have you swept the audio to determine what the audio response is now as things stand? What are you ultimately trying to achieve with the audio. Do you want the changes to be all self-contained in the transmitter or some additions outboard etc.

Looking at the audio circuit for the transmitter, the use of the couplets in the low level audio cries out for their replacement with suitable discrete components. Also, the use in the GK500 of a single ended 6L6 pentode connected, driving class B modulator 811 grids? with no negative feedback around the 6L6 stage. This is a huge no no in the original design of the audio driver. Negative feedback around this stage would help smooth out the response and lower the apparent plate impedance. The 6L6 would act closer to a triode. I have used plate to plate negative feedback in receivers with the typical single ended 6AQ5 in the output with a small audio output transformer, and used a 1meg resistor/5meg pot from its plate back to the previous class A audio/driver stage. It really helped smooth out the audio in the receiver. You would have to crank the gain up of course. Perhaps change out the 6L6 afterwards with a 6550.

There are other ways of dealing with the 6L6 driver stage as have been pointed out previously. There is already a PP driver transformer there in the GK500. You could have use of that in a PP audio driver circuit. Change the 6C5 over to a dual triode tube that would plug right in, and rewire that stage as a phase inverter for the new PP stage.

One could eliminate the use of the driver transformer by using a direct coupled cathode follower audio driver. You would need a gain/voltage there prior to any cathode follower stage to drive class B grids. I believe that the Clegg Zeus modulator uses a cathode follower audio driver to drive class B 811a grids. The circuit has a lot of audio restrictive components there as I recall. I think the audio driver tube was a dual triode 6BX7 or 6BL7.
And there are the direct coupled FET designs that some of the other guys have mentioned on here too of course.

Al VE3AJM

[WD5JKO]

Guys,

Refer to this recent thread as there is a similar discussion with good info relevant to the Globe King 500 audio:

http://amfone.net/Amforum/index.php?topic=34456.msg266296#msg266296

Also, take note that the GK 500 modulator inter-stage transformer is a push pull input device, but Leo Meyerson for whatever did away with the push pull 6F6 driver (triode connected), and instead used a single 6L6. That centertap is still available should you decide on a push pull driver.

Jim
WD5JKO

[Opcom]

I agree with the previous posts advising doing clean-up before modifying. Thank goodness others have added their opinions!