Hi,
It looks great and it's also nice to read it's a success. The circuit's interesting - not that I understand it all yet - but first off I do have a very simple question about the dcr value differences in the kt88 cathode winding when everything else is perfectly symmetrical. Are they design values or measurements of the actual transformer?
It looks great and it's also nice to read it's a success. The circuit's interesting - not that I understand it all yet - but first off I do have a very simple question about the dcr value differences in the kt88 cathode winding when everything else is perfectly symmetrical. Are they design values or measurements of the actual transformer?
Those are the measurements I made on the transformer itself.
This article may help if you are figuring out how the Unity Coupled circuit works: http://www.cieri.net/Documenti/Schemi/McIntosh%20-%20Description%20and%20analysis%20of%20a%20new%2050%20watt%20tube%20amplifier%20circuit.pdf
This article may help if you are figuring out how the Unity Coupled circuit works: http://www.cieri.net/Documenti/Schemi/McIntosh%20-%20Description%20and%20analysis%20of%20a%20new%2050%20watt%20tube%20amplifier%20circuit.pdf
Thanks for that. I looked but didn't find it. I did however find a plethora of other interesting stuff.
A couple of them are
https://www.youtube.com/watch?v=W58Rqx_4u2w
http://www.tubebooks.org/books/lockhart.pdf
A couple of them are
https://www.youtube.com/watch?v=W58Rqx_4u2w
http://www.tubebooks.org/books/lockhart.pdf
Thanks for pointing out that Youtube video, that was cool to watch.
Edit: And you'll notice that the problem that Lockhart was trying to solve with an additional winding in the transformer, I solved with the floating MOSFET screen regulator.
Edit: And you'll notice that the problem that Lockhart was trying to solve with an additional winding in the transformer, I solved with the floating MOSFET screen regulator.
Last edited:
No, that would have taken me a day or two's worth of cogitating to notice, thanks for bumping the understanding forward!
I remember when I was first introduced to the circuit. It takes a bit to wrap your head around. It solves some problems in some really elegant ways. Unfortunately, it is a half-cathode follower so it takes a very large voltage to drive it. That's the main problem it creates. So if you can overcome that problem you can reap some benefits.
The tubes are triodes at DC and pentodes at AC. I had to draw myself some pictures of anode and cathode waveforms at first to convince myself of this but it all works out in the end.
The tubes are triodes at DC and pentodes at AC. I had to draw myself some pictures of anode and cathode waveforms at first to convince myself of this but it all works out in the end.
Do Q2 and Q4 also work to balance current through the output tubes? Trying to understand the numbers in the output section I see a current offset just shy of 7mA.
No, Q2 and Q4 just drop the screen voltage to 250V(opposite tube plate voltage - 200V). That's lower than most KT88s are run at but this amp has a 6.4k load (higher than most designs) so it seemed appropriate to drop the screen voltage to that level to make it so that the load line doesn't cross way below the knee of the curves. Driving pentodes into saturation below the knee of the curves can be hard on the screen grid so I dropped the knee by lowering the screen voltage.
Where are you getting the current offset from? Are you trying to work out the bias conditions and factoring in the different DCRs?
First off, I just labeled those so that I could add a resistor to balance them out. So in the working amp they are the same on both sides. This is just a simplified schematic. The bias is applied to the gates of the mosfet followers by an auto-bias board that right now is set for 50mA per output tube. The actual measured voltage on the grid of the output tube was -33V I believe.
The numbers on the schematic were my attempt to guess at what the bias voltage would be. It is hard to work out ahead of time.
Normally in a Unity-Coupled amp you would determine DC bias conditions from the triode curves since the plate and screen are always the same voltage at DC. But I complicated things by tying the screen to the plate - 200V. You could try to figure out bias conditions from pentode curves with screen at 250V but I don't know of any that I have seen, only 200V or 300V in the GEC data sheet. So it is not straightforward. But anyway, the numbers in the schematic were a guess that turned out to be a bit wrong but still well within the capability of the auto-bias board.
Where are you getting the current offset from? Are you trying to work out the bias conditions and factoring in the different DCRs?
First off, I just labeled those so that I could add a resistor to balance them out. So in the working amp they are the same on both sides. This is just a simplified schematic. The bias is applied to the gates of the mosfet followers by an auto-bias board that right now is set for 50mA per output tube. The actual measured voltage on the grid of the output tube was -33V I believe.
The numbers on the schematic were my attempt to guess at what the bias voltage would be. It is hard to work out ahead of time.
Normally in a Unity-Coupled amp you would determine DC bias conditions from the triode curves since the plate and screen are always the same voltage at DC. But I complicated things by tying the screen to the plate - 200V. You could try to figure out bias conditions from pentode curves with screen at 250V but I don't know of any that I have seen, only 200V or 300V in the GEC data sheet. So it is not straightforward. But anyway, the numbers in the schematic were a guess that turned out to be a bit wrong but still well within the capability of the auto-bias board.
Yup, exactly. Just using 'the law' to try to understand what's going on.
Thanks for the detail. I don't know much about using multi-grid tubes yet but there's lots written that says it's best to have screen voltage stiffly regulated . . . . and here we are with a successful circuit where the screen swings with the plate, and perhaps even influences the cathode current (not sure how big R3 and R5 are.) There's more to ask but let me read the 50 Watt article first.
You do neat stuff and this one is an interesting study. Thanks again for posting it.
The screen swings with the opposite tube's plate, which follows the AC swing of the cathode, so the screen to cathode voltage stays constant. So the tubes operate as pentodes, but without a separate screen supply. It is really tricky what those guys did here when they invented this type of transformer.
R3 and R5 are 500k. I just put them there so that when the tubes cut off and current stops flowing into the screen, there would be some trickle current going through it to give the source follower some minimum load. I didn't know if bad things would happen if current stopped flowing through the MOSFET. They may not be necessary.
Well, I intended to but thought it best to hold off until I understood the signal circuit a little better. This is an involved circuit!
. . . . but you've answered the first question - how it sounds - so the next one is, does it actually eliminate the notch distortion completely?
From what I understand, notch distortion is only a problem when you have a really cold bias in a class AB amp, near class B operation. With such a cold bias, you also have distortion added by the nonlinearities in the output tubes (crossover distortion). So even if you eliminate notch distortion, you wouldn't necessarily want bias that cold anyway.
I have been running the KT88s at 50mA and 450V(22.5W), which is coldish but not all that cold. I looked at McIntosh schematics and it seemed that they ran their amps at similar operating points.
Now from what I have seen, you can easily run a normal push-pull amp at a similar operating point and normal push-pull output transformers will produce no notch distortion at those levels. So while Unity Coupled transformers have a theoretical resistance to notch distortion, McIntosh amps didn't seem to run the output tubes at some really low quiescent current, like say 10mA or less, and take advantage of that.. I'm guessing that distortion performance was better at more current so that's what they went with. Running KT88s at 22.5W would probably give excellent life anyway.
It would be interesting to keep backing off the current and see where notch distortion appears and compare with a conventional amp of similar power, but I don't think I'll do that for a while.
I'm hoping to get some distortion measurements done this weekend on the current settings.
Last edited:
I guess I should also state that it is my understanding that crossover and notch distortion are two separate phenomena. Most people use the terms as synonyms and it is my understanding from reading van der Veen's book and other writers that crossover distortion is simply distortion that results from the nonlinearities of the output tubes at low currents. Notch distortion is caused by part of the winding suddenly going open due to a tube cutting off suddenly. I have seen scope shots of both and they are very different. You will see numerous examples on the internet of scope traces of crossover distortion, but pictures of notch distortion look much different. I'm having trouble finding an example right now on the internet because there are so many people who post pictures of crossover distortion and call it notch distortion...
A Unity Coupled amp will indeed have much less crossover distortion but that is a result of the heavy local feedback inherent to the circuit, not a result of the tight coupling that reduces the potential for notch distortion.
Be aware that I think there are a lot of people out on the internet that don't understand things this way.
A Unity Coupled amp will indeed have much less crossover distortion but that is a result of the heavy local feedback inherent to the circuit, not a result of the tight coupling that reduces the potential for notch distortion.
Be aware that I think there are a lot of people out on the internet that don't understand things this way.
It appears that when the paper I linked to above by McIntosh and Gow refers to "notch distortion" it is referring to a phenomenon thoroughly described in a 1936 paper by Pen-Tung Sah entitled, "Quasi Transients in Class B Audio-Frequency Push-Pull Amplifiers". I can't find a copy on the internet to link to at the moment but that is the paper to look at if you want the gory details.
I don't claim to understand much in that paper but I do understand that it is describing something completely different than crossover distortion.
I don't claim to understand much in that paper but I do understand that it is describing something completely different than crossover distortion.
How do you like the Tent Labs autobias circuit? Does the Tent circuit allow a full output sine wave for an extended period without doing "weird" things to the bias? I realize that is NOT a normal operating scenario but am curious.
I'm not sure what it's drift rate is. I haven't attempted to measure. I had been working on this amp for like six years and finally got it all put together. I'm really enjoying listening to it now and am not anxious to take it back out to the shop.😀 I know that the circuit is designed to clip peaks from the AB waveform but I am assuming that there would be some drift as well. It would be interesting to quantify.
I will say that I notice that in practice I don't notice it start to sound bad after cranking up the music to ear-bleeding levels for extended periods of time.
I'm hoping to take some distortion measurements this weekend and I will let the amp dwell a bit and see if distortion gets worse with time, which might be an indication that bias is getting colder and inducing crossover distortion.
I did watch the speed of bias adjustment on power up and it seems to change at a rate of about 1V/sec in case you are wondering.