JustMe2023
Member
What is your theory?
Dynaco Stereo 35 replacement output transformers
- Thread starter guitargrump
- Start date
What is your theory?
the regulation in the tube is worse with individual resistors. With a shared per channel resistor, you get more of an average voltage across it since as one tube draws more current, the other draws less. With a single resistor you don't get that. Shared across the amp you get the other channel to help regulate it, though that doesn't really do much if both are passing the same signal.
if you want a study in how it was meant to work in stock form, this gets into it pretty well
New Modifications for the Dynaco SCA35 and ST-35 (tronola.com)
also includes info on how to overcome the limitations of cathode bias in the first place. Its not quite so bad in tubes that don't have a big change from idle to full output, but with a 6bq5 family there is a pretty wide swing in current from idle to full power, which creates a large shift in bias as a result. In a Class A stage, the current shift is very low and cathode bias works very well there.
just as an aside, it seems a lot of attention is paid to the DC operating conditions of tubes but a whole lot less is paid to the AC operating conditions. Idle current is important and all that but the AC conditions are what we are actually listening to. If things fall apart when playing music, all the ideal DC conditions in the world don't make up for that.
if you want a study in how it was meant to work in stock form, this gets into it pretty well
New Modifications for the Dynaco SCA35 and ST-35 (tronola.com)
also includes info on how to overcome the limitations of cathode bias in the first place. Its not quite so bad in tubes that don't have a big change from idle to full output, but with a 6bq5 family there is a pretty wide swing in current from idle to full power, which creates a large shift in bias as a result. In a Class A stage, the current shift is very low and cathode bias works very well there.
just as an aside, it seems a lot of attention is paid to the DC operating conditions of tubes but a whole lot less is paid to the AC operating conditions. Idle current is important and all that but the AC conditions are what we are actually listening to. If things fall apart when playing music, all the ideal DC conditions in the world don't make up for that.
JustMe2023
Member
The dissipation per tube in the current ST-35 is approximately 13 watts per tube. Is that not class A? Change in power draw for idle state vs. full power output should be minimal.
JustMe2023
Member
This thread has gone in an unexpected direction. Oops.
Current version B+ about 380-400, ultralinear configuration, stock ~8k OPT. About 13W dissipation per tube.
Please explain how this circuit (the original ST-35 version) is not class A or very close to it:
The individual bias mod in question uses four resistors and four trimmer pots so nothing is shared, not per channel and not per quad. The original quad shared R13 and C8D in the amplifier schematic above are replaced. Under something close to class A operation, this arrangement seems to be nearly ideal unless I am missing something.
Current version B+ about 380-400, ultralinear configuration, stock ~8k OPT. About 13W dissipation per tube.
Please explain how this circuit (the original ST-35 version) is not class A or very close to it:
The individual bias mod in question uses four resistors and four trimmer pots so nothing is shared, not per channel and not per quad. The original quad shared R13 and C8D in the amplifier schematic above are replaced. Under something close to class A operation, this arrangement seems to be nearly ideal unless I am missing something.
Attachments
Last edited:
I know that its not shared. I was comparing that to the stock configuration where it is shared across both channels. The measurable performance of the amplifier will be worse with individual resistors than it would be with the common for both channels resistor, or even a one per channel resistor.
you can prove its not class A if you stick a voltmeter across the cathode resistor and compare the voltage at idle to the voltage as power output increases. You'll see a very large swing in voltage which indicates a shift in current. If it was class A you would see very little change. The high idle dissipation is a thing you typically get with cathode bias in an effort to keep it from grossly under-biasing itself and creating crossover distortion as power increases. It still happens, but if you idled it at a lower current its even worse.
you can prove its not class A if you stick a voltmeter across the cathode resistor and compare the voltage at idle to the voltage as power output increases. You'll see a very large swing in voltage which indicates a shift in current. If it was class A you would see very little change. The high idle dissipation is a thing you typically get with cathode bias in an effort to keep it from grossly under-biasing itself and creating crossover distortion as power increases. It still happens, but if you idled it at a lower current its even worse.
https://www.vtadiy.com/loadline-calculators/loadline-calculator/
Try the sim to get an idea about how this all plays together. There is no way this can stay in class A to a very high output power from the high B+ voltage. Look at the data sheets for 6BQ5 single ended suggested operating points. About the best you can hope for in PP class A is to double the plate load impedance and double the expected output power at a similar B+. As the power supply voltage climbs the transition to class B comes sooner if you stay withing dissipation ratings.
Try the sim to get an idea about how this all plays together. There is no way this can stay in class A to a very high output power from the high B+ voltage. Look at the data sheets for 6BQ5 single ended suggested operating points. About the best you can hope for in PP class A is to double the plate load impedance and double the expected output power at a similar B+. As the power supply voltage climbs the transition to class B comes sooner if you stay withing dissipation ratings.
Last edited:
JustMe2023
Member
So the tradeoff is,
Possibly getting a few extra clean watts, assuming tubes are quad matched under actual operating conditons and not just a tube tester, and assuming they stay matched that way over time,
vs.
Losing those potential few extra clean watts but being able to tweak each tube over its life and also being able to use a choke in the available chassis space.
Seems to be six of one and a half dozen of the other I guess.
I would think that the enhancement provided by using a choke, individually adjustable tube bias, and maybe larger PS caps (that are very easy to do), would be a fine approach even if it might not allow the absolute maximum possible power output.
I suppose 4 regulators would be ideal if someone was building an amp from scratch, but I'm not, and at that point it's really not an ST-35 anymore anyway.
Possibly getting a few extra clean watts, assuming tubes are quad matched under actual operating conditons and not just a tube tester, and assuming they stay matched that way over time,
vs.
Losing those potential few extra clean watts but being able to tweak each tube over its life and also being able to use a choke in the available chassis space.
Seems to be six of one and a half dozen of the other I guess.
I would think that the enhancement provided by using a choke, individually adjustable tube bias, and maybe larger PS caps (that are very easy to do), would be a fine approach even if it might not allow the absolute maximum possible power output.
I suppose 4 regulators would be ideal if someone was building an amp from scratch, but I'm not, and at that point it's really not an ST-35 anymore anyway.
Last edited:
The fundamental cause for this is that increased supply voltage necessitates increasing the bias voltage to maintain the same idle current. Bias voltage must then be further increased to reduce the current below the previous value because higher B+ causes dissipation to increase [ I x E ]. This all means the cathode bias resistor must be made a significantly higher value. This, in turn, means increases in current at higher output power will cause bias to shift more quickly toward Class B.
Jack
JustMe2023
Member
To open another can of worms ... why not beef up the caps and drop B+ by 20 volts. 
One day I do hope to bring this thread back to the original topic of transformer availability somehow.
One day I do hope to bring this thread back to the original topic of transformer availability somehow.
Pio1980
AK Member
Class 'A' keeps the output tube always operating in the linear portion of conduction, idling at the linear midpoint with no signal input, rather than doing a linear conduction "hand-off" from one tube to another in a push-pull pairing as is the case with push-pull operation. The tube pairs only conduct just enough at crossover to keep them in the bottom of their linear operating range @ hand-off, where one is smoothly cut off and the other smoothly takes over. Bias of the pair sets the linearity of the pair at the hand-off point.
Last edited:
So what happens to the distortion created in each tube as it approaches cutoff? Regardless of exactly how it's biased, each will pass through this very non-linear region in AB1.
Jack
Pio1980
AK Member
Good question above my pay grade, maybe something with cancellation in the pair & feedback?
Mac did significant cancellation distortion reduction with their signature unity coupled output circuit.
Well, I'm pretty sure this is the reason AB1 small-signal distortion performance improves when idle current is increased. Greater idle current moves signals further away from the non-linear region. Conversely, as bias current is reduced, ever-smaller signals will enter this area. It's interesting that triodes and pentodes react so differently in this regard. If the screen voltage is fixed, a pentode is controlled primarily by its control grid. It experiences a "hard landing" when the negative-going portion of signals push current toward zero. But triodes are also controlled by their anode voltage. When the grid swings negative, reducing tube current, the anode swings positive, which increases it. In fact, when the load is transformer coupled, anode voltage can swing well above the static B+ voltage provided by the power supply. The overall effect is a "soft landing" when the signal pushes conduction toward cutoff. It follows that a higher load impedance (which allows the anode to swing more vigorously) will produce less distortion in a triode. I believe this is all borne out by the curves.
Jack
Jack
JustMe2023
Member
Please do a favor for the less technical people here and put in the values there for an actual ST-35 operating at about 14 watts per channel, and then explain it, if you have a few minutes. Thanks.
Pio1980
AK Member
How should that work in true ultralinear mode?
In the context of my earlier post, UL falls in-between. Screen voltage rises and falls as a percentage of anode variations, and this moderates the pentode's "trajectory" when moving toward cutoff. In the end (and in consideration of transformer characteristics), I suspect the most natural-sounding amplification must come from Class A push-pull triode amplifiers that operate with minimal NFB and which are limited to the area above the non-linear region. This topology also has certain advantages regarding power supply coloration. None of the vintage OPTs in my collection lend themselves to this mode, otherwise it would probably be my amplifier of choice.
Jack
Pio1980
AK Member
Makes sense, UL has a certain elegance for pentode/beam tetrodes.
I'll give a try. Here's a screen capture of the predicted load line for 380V, 14W dissipation (which is over the databook limit). I've used 14W to calculate the bias current but this doesn't really include the screen current. I just used 14W and 380V. But you can clearly see the predicted break in the load line where the system would transition from class A to class B. The power output prediction is off as you are not going to get 27W from a pair of 6BQ5 at this operating point. Pure class A would require a lower B+ and higher current to set the tubes at the center of the load line. You can guesstimate pretty closely by taking the SE operating point from a datasheet and doubling the power prediction for PP using the same op point and doubling the anode load impedance.
The second file shows a theoretical class A alignment where I've lowered the B+, increased the bias current and increased the OPT primary impedance to 10k. The loadline is straight and extends to the 0V grid line with the transition to class B beyond 0 volts. Keeping the 8K primary the class B transition happens just before getting to 0V on the grid so nearly class A.
Don't know if this is a great, or useful, explanation.
the only trick here, and I don't know how to model it, is the shifting bias point that happens as the output power increases in a cathode bias arrangement. I suspect that would show up as a narrower class A region. At a minimum it seems like it would have to make that load line less line-like.
everything is a trade off though. Class A seems like it should be the superior method in terms of fidelity. Its bad in terms of power efficiency though. Straight B operation is a winner in terms of efficiency, but tends to suck in terms of fidelity. AB is essentially a compromise between the two.
and yes I know class C exists but I don't think that really ever gets used for audio. Way too much distortion, though it seems to work OK for radio transmitters.
everything is a trade off though. Class A seems like it should be the superior method in terms of fidelity. Its bad in terms of power efficiency though. Straight B operation is a winner in terms of efficiency, but tends to suck in terms of fidelity. AB is essentially a compromise between the two.
and yes I know class C exists but I don't think that really ever gets used for audio. Way too much distortion, though it seems to work OK for radio transmitters.