My take on a Single Ended amp

As a precursor to this discussion I should mention why I picked the bias point I did for the 12AT7 stage. The output stage requires 100V peak to peak drive, and I knew I had to use a 47K grid resistor on the output stage to stay within the output tube specs when using fixed bias. I wanted the 12AT7 stage to operate in a (visually) linear portion of the grid curves. The bias point picked was 3V cathode, 225V across the tube, and 4 mA quiescent current. Swinging 100V peak to peak from that Q point into a 47K load is running the tube near its limit on the negative side of the signal swing, leaving a mere 1 mA conduction headroom on the negative most excursion before the tube goes into cutoff. But I knew this and picked that bias point anyway because I favored linear swing over anything else.

When adding plate to plate feedback, the plate is loaded down even further at AC, so much so that keeping the bias point the same would reduce drive signal ability down to 80V peak to peak--not enough to drive the output stage to full power. To address this, the 12AT7's bias point would need to be moved. If it were moved up to 7.5mA quiescent current, -2V cathode, and 225V across the tube, that would do it, but it's swinging close to 0V grid now at the positive peak signal excursion, as well as moving the signal swing into a more non linear portion of the grid curves. Also running at that Q point puts the plate's static dissipation at 67% of design center max, so tube life might be shortened a bit as compared to the original bias point I selected.

The feedback calculation with plate to plate feedback is a little more complicated now so I didn't run through that math to determine how much feedback would be applied given a 100K feedback resistor. Very rough estimating says maybe 10 dB feedback would be applied.

So...I don't know but it seems the 12AT7 would not be the best choice in this amp for Schade style feedback. But, not having built it, I'm just theorizing.

I can see how this type of feedback would be useful with lesser requirements on the output stage, such as with EL84 or 7591 type output stages, and/or with a stouter frontend tube.

All that being said, it's still pretty easy to try it, and look only at small signal analysis. It would still be instructive to learn how this type of feedback affects HF performance.
 
Last edited:
Your schematic is very similar to the RH amp. On that schematic. On the 12at7 there is no grid stopper resistor and the grid leak resistor is increased to 1M. This enables the most signal to the grid. I have had no problem with this implementation as long as i use shielded cable from input to the tube grid. The cathode bias resistor is 360 ohms. with plate resistor of the 100k which is also the FB resistor. According to the designer input sensitivity is just over 1.0v. In this scheme the output tube is a kt88 with just over 350v Av.
In some different schematics a plate load resistor for the 12at7 is also used along with the plate to plate FB resistor.
In my amps since i am using fixed bias, i kept the grid to ground resistor of the kt120 at just over 50k. But, did you notice that the original GEC kt88 data sheet shows for fixed bias no more than 150k grid leak resistor. I always wondered why their tube could tolerate a higher grid resistor in fixed bias.
Another thing that i tried in my amps is to use a LED for the bias of the driver tube. That seems to be a good addition for the 6sn7 driver but i have not tried it for the 12at7. You can add an additional 100 ohm resistor for the 12at7 cathode with the the led to keep the led away from cut off conditions.
Anyway, you really need to listen to the amp just to get a taste of how the sound compares to other amps. I have found a couple of good test speakers for my workshop. One is a pair of NHT Zero something and the other pair is a Yamaha NS 10 monitors, which are pretty good sounding up close. While test numbers are an important guide they are certainly not making you any music. regards, Primo.
 
Schade type feedback small signal testing.

First test was direct plate to plate with 100K resistor. Feedback amount applied: 6 dB. But immediatly I saw rounding on the bottom of the sine wave more so than the top, even at a mere 1 watt output. So this approach was abandoned.

Next I tried screen to plate feedback with a 47K resistor.FB applied was 6 dB. Sine waves looked great (visually). An 8 KHz square wave at 1 watt output looks like this:

Screen to plate FB 6 dB.JPG

Frequency response:
+0/-0 dB to 14 KHz
+0/-1 dB to 30 KHz

Since the secondary is not involved, there were no peaks above the reference voltage delivered at 1 watt output into an 8 ohm dummy load at any frequency, and basically no transformer resonance noted at all.

Since I'm here, thought I would measure full power output:
screen to plate FB 6 dB 15 watts output.JPG

And therein lies the issue. Uneven drive. This speaks of inadequate drive for this kind of feedback with the topology and bias point used. I think this is worth investigation further, but it's taking me in a different direction than I wanted to go for this build. Like you say, if the feedback resistor is the plate resistor it might change things. But that's analysis for another day.
 
Hi Kev, thanks for taking the measurements. It seems that you had some interesting results with the Schade fb resistor. You noted the without the output transformer in the loop revealed no resonances or bad behaviour. I think that also shows up in the sound quality of the rh amps which to me have a very natural life like sound.
One question i was wondering what you mean by "uneven drive". Which you said occurs under full power output. I am not familiar with that term or how much of it is encountered under normal listening conditions. Could you elaborate? regards, Primo
 
Uneven drive meaning that, with the amp as built on the test bench right now with the plate to plate or screen to plate feedback approach used, this type of feedback changes the AC load line of the 12AT7 stage so severely that the stage cannot swing the 100V peak to peak needed to cleanly drive the output stage. Specifically the 12AT7 goes into cutoff on the downward going portion of the input sine wave signal, while the upward going portion of the waveform is amplified correctly. Thus "uneven" drive on upper vs lower portions of the sine signal. And you can clearly see that in the above scope shot taken of the output signal at the OPT secondary--fat round troughs on the bottom, nicely formed peaks on the tops.

But if the 12AT7 stage were designed correctly to work with plate to plate feedback, this situation would not occur. So it's just a manifestation of poor design of this amp when used with the plate feedback approach. Obvously the RH amps work well and probably sound excellent, so it's clearly possible to do, just not with the design I've presented here.
 
Back to distortion measurements...

With the cathode resistors reversed in position like this,

upload_2018-9-8_0-58-1.png

distortion measurements were overall a little better at mid band but a little worse at the band extremes on both ends, but also applied feedback measured at 9.25 dB in this config instead of 10 dB as in the other cathode configuration. With respect to distortion, these two cathode configurations appear to be nearly identical, so I see no reason to deviate from my original design.
 
I usually like to run my feedback by the first configuration (the bypassed resistor next to the cathode), as it impresses just a little less DC on the output transformer secondary (the DC voltage at the bottom of that bypassed resistor is closer to ground, than at the cathode itself).

Yes, I know that it's only millivolts, once it goes through the feedback resistor itself- but anything to make it closer to ideal (pure AC with no DC offset) is an improvement, IMHO.

Regards,
Gordon.
 
Uneven drive meaning that, with the amp as built on the test bench right now with the plate to plate or screen to plate feedback approach used, this type of feedback changes the AC load line of the 12AT7 stage so severely that the stage cannot swing the 100V peak to peak needed to cleanly drive the output stage. Specifically the 12AT7 goes into cutoff on the downward going portion of the input sine wave signal, while the upward going portion of the waveform is amplified correctly. Thus "uneven" drive on upper vs lower portions of the sine signal. And you can clearly see that in the above scope shot taken of the output signal at the OPT secondary--fat round troughs on the bottom, nicely formed peaks on the tops.

But if the 12AT7 stage were designed correctly to work with plate to plate feedback, this situation would not occur. So it's just a manifestation of poor design of this amp when used with the plate feedback approach. Obvously the RH amps work well and probably sound excellent, so it's clearly possible to do, just not with the design I've presented here.

IIRC, most RH type amps are using an output tube that itself requires less drive than the KT88 or KT120 (i.e, 6L6, 6V6/6AQ5, or especially EL84). They probably never tax the 12AT7 to its cutoff limits.

With the KT88/6550 or KT120, things just have to be beefed up in the driver stage. Same situation as when Dave G. converted the Eico HF60 to KT88s- he had to add a negative B- supply to the phase inverter cathode to provide enough headroom for the extra drive needed...

Regards,
Gordon.
 
IIRC, most RH type amps are using an output tube that itself requires less drive than the KT88 or KT120 (i.e, 6L6, 6V6/6AQ5, or especially EL84). They probably never tax the 12AT7 to its cutoff limits.

With the KT88/6550 or KT120, things just have to be beefed up in the driver stage. Same situation as when Dave G. converted the Eico HF60 to KT88s- he had to add a negative B- supply to the phase inverter cathode to provide enough headroom for the extra drive needed...

Regards,
Gordon.
In the RH "universal" amp which uses a kt88 or 6l6gc, el34, the 12at7 uses a combined plate load and fb resistor or 100K . I have made several of this type and that seems to work and sound good too.
 
Putting all my research and bench development together now, here's the proposed build schematic (see attached).

A couple of notes:
  1. The power supply specifies two 0.5 Henry 300 mA chokes. I specified the Triad chokes because they are small and light, don't cost a lot, and will fit under a 2" high chassis. Cascading two L/C filter sections together gets ripple down to 1 mV at idle while adding minimal series resistance to the power supply.
  2. The negative bias supply is heavily filtered to keep ripple noise down. Ripple is about 25 uV. Even heavily filtered this way, from a cold start, the bias supply will come up to full voltage in just over one second, well ahead of the output stage starting to conduct.
  3. All power supply capacitors must be rated for 500V DC, since the unloaded B+ voltage will rise to 490V for 10 to 15 seconds before the output stage warms up.
  4. The negative bias supply has a safety feature built in that will force bias voltage applied to the output stage to go more negative to protect the output tube, should the the wiper of the adjust pot disconnect from its track.
  5. Using a fixed bias approach provides an inexpensive method using cheap low wattage parts to tune the output stage for maximum performance regardless of the exact plate voltage used. However, a distortion or spectrum analyzer is required to fully take advantage of this feature.
I received the chassis today that this amp will be built on. It's nice thick aluminum, with African Padauk side accents. Should be a beautiful amp. But I've got to put this project on hold now for a few months while I work on other projects. I'll get back to building/testing this amp probably by December.
 

Attachments

  • KT120 SEUL draft.pdf
    140 KB · Views: 117
Last edited:
Great thread, gentlemen. Even for someone as ignorant of valve electronics as I am....very instructive. I daresay this is AK at its best.
 
In the RH "universal" amp which uses a kt88 or 6l6gc, el34, the 12at7 uses a combined plate load and fb resistor or 100K . I have made several of this type and that seems to work and sound good too.

I would expect things to be 100% OK with the 6L6 or EL34- but, maybe a bit on the edge with the KT88.

Even so, it would only become audible near cilpping- the consequence would be a loss of a little bit of effective headroom, before the amp started to sound "nasty" in clipping. It's not going to be a dramatic effect at lower volume levels. One may not knowingly perceive it, unless specifically looking and listening for it.

Regards,
Gordon.
 
Putting all my research and bench development together now, here's the proposed build schematic (see attached).

A couple of notes:
  1. The power supply specifies two 0.5 Henry 300 mA chokes. I specified the Triad chokes because they are small and light, don't cost a lot, and will fit under a 2" high chassis. Cascading two L/C filter sections together gets ripple down to 1 mV at idle while adding minimal series resistance to the power supply.
  2. The negative bias supply is heavily filtered to keep ripple noise down. Ripple is about 25 uV. Even heavily filtered this way, from a cold start, the bias supply will come up to full voltage in just over one second, well ahead of the output stage starting to conduct.
  3. All power supply capacitors must be rated for 500V DC, since the unloaded B+ voltage will rise to 490V for 10 to 15 seconds before the output stage warms up.
  4. The negative bias supply has a safety feature built in that will force bias voltage applied to the output stage to go more negative to protect the output tube, should the the wiper of the adjust pot disconnect from its track.
  5. Using a fixed bias approach provides an inexpensive method using cheap low wattage parts to tune the output stage for maximum performance regardless of the exact plate voltage used. However, a distortion or spectrum analyzer is required to fully take advantage of this feature.
I received the chassis today that this amp will be built on. It's nice thick aluminum, with African Padauk side accents. Should be a beautiful amp. But I've got to put this project on hold now for a few months while I work on other projects. I'll get back to building/testing this amp probably by December.
Nice schematic, although i am left wondering how this arrangement will sound. And to me the second best part besides listening to an amp of your own design, is putting it together.
 
For sure this design is more controversial and is not your "mother's" typical SE amp. It's not using choke input in the power supply, it's not using cathode bias, and it's using global feeback encompassing the output transformer!! All things to be shunned by conventional wisdom. No matter to me. Shedding all preconceived biases and trying something new (to me) is no problem for me at all. If it is great I will have an awesome new amp to listen to. If it's not great I will have a bunch of experience gained and will try again with a different design. :)
 
Primo--no I haven't listened to the amp at all yet. I want to try that Schade style feedback at some point. Maybe I could use a 6FQ7 in place of a 6SN7 on the test jig. I'm stoked right now though that the Edcor OPTs seem to be really well wound. Not at all like my Edcor PP output transformer experiences...
Hi Kev, that would be most interesting. With your test rig it would just require moving the connections to accommodate the 6fq7/6cg7. Plus the requisite resistor and cap changes. The 6fq7 would certainly push more current than the at7. Cheers,
 
I don't want to clutter up this thread with further discussion of alternate typologies. But since I'm done with my jig now, and since the amp wired on it currently is really close to the RH universal, It would be almost trivial to cut over to it and compare. I'll do that on this thread and let the discussion go where it may...
 
Last edited:
The chassis for this amp finally came. It's beautiful polished aluminum with Padauk side wood accents. Problem is this chassis is extremely thick walled aluminum--so thick that it's hard to drill large diameter holes, and thick enough that the 2 Inch high chassis won't fit a 2" bias transformer underside with the bottom plate on. So I'll have to finagle something--probably cut out a rectangular hole in the bottom plate so the transformer can poke through.

I ordered the power transformer around the first of December so I expect it will arrive within the next two to three weeks. In the mean time I've done about as much chassis drilling as I can.

IMG_3264.JPG


The chassis is 5/16" thick aluminum. This is too thick for amateur drilling, and too thick to easily drill with a step bit, and too thick to bottom mount tube sockets. :(

I've also painted the butt-ugly Edcor blue end bells on the output transformers to something more presentable. It takes a good long while to paint these things. You've got to pretty much strip down the paint to bare metal, then spray on two coats of primer, two coats of paint, and three coats of clear coat. (I use automotive paint products for this in spray cans you get at the color match section of the auto parts store). About as labor intensive as can be. And when the PT arrives, I have the pleasure of doing it again.

:whip:


IMG_3274.JPG
 
Man, that is a beautiful chassis. I guess the workability issues are the price you pay!
 
You've got to pretty much strip down the paint to bare metal, then spray on two coats of primer, two coats of paint, and three coats of clear coat. (I use automotive paint products for this in spray cans you get at the color match section of the auto parts store). About as labor intensive as can be. And when the PT arrives, I have the pleasure of doing it again.

Edcor sells raw, unpainted end bells -- perhaps next time you can specify the raw end bells when you place the order? FWIW, if it were me, I would simply take them to a powder coat shop; it's something I got used to doing back in a previous life when I restored air-cooled VWs as a hobby ... there's a lot of engine shrouding as part of the cooling system in those puppies!
 
Back
Top Bottom