Dear Bob,
SA2021 is a Class H design by (mostly) Dave Zan and me. This link shows the correct schematic https://diyaudio.com/forums/posts/7428118
The posted link from PB2 requires that you use the same post per page settings - otherwise you get the wrong post displayed.
The connector at the right side of the schematic feeds +/- 20V and +/- 72V
This saves a lot of dissipation power at normal to mid level listening.
On the bench the amp has been very well tested (with +/- 80V). A tower case is work in progress. Here the first hearing Test: https://diyaudio.com/forums/posts/7532135
I generally write "preliminary" to schematics as long as the amp doesn't do the every day work in my home cinema.
BR, Toni
Last edited:
@astx Hi Toni, I have a Sound Craftsman 5002 Class H amp that failed was repaired (not by me) and
failed again. The top rail is I believe 93V. Do you think that your design would work well for it?
Fairly sure that it was 250W/ch. I have too many projects and probably will not get to it, but
who knows? It used Toshiba outputs that are NLA now.
failed again. The top rail is I believe 93V. Do you think that your design would work well for it?
Fairly sure that it was 250W/ch. I have too many projects and probably will not get to it, but
who knows? It used Toshiba outputs that are NLA now.
Does anyone remember a Class H amp from the early 1980s that had +/-60VDC and +/-90VDC? It had, I think, 3 OPs running on 60V and a single device to pull it to 90V. It had a problem with 22K ~1/2W resistors that ran too hot and eventually went open. It was odd to have an amp failure be a resistor, although I think I did have to find OP's for one. I expect I will recognize the name if I hear/read it. It could have been a Canadian brand.
Hi Toni,
Thanks for that schematic correction - now it all makes sense. It is a very nice class H design. It looks like a lot of thought and attention to detail went into it. I like the idea of using power MOSFETs for the outer pairs.
Cheers,
Bob
The earliest ones I am even familiar with were the Proton/NAD and the Carver cubes. IIRC, the Proton used a single 30 amp darlington to pull the rail up. I remember because a buddy of mine had a Proton and he used it to rip the suspension out of his woofers playing 1812 overture. We had that thing apart investigating it’s inner workings. I had the unfortunate pleasure of working on a couple of the cubes - and they were real POS’s. If you thought PL’s were finicky, poorly made and designed….. got nothing on these.
They went mainstream with the QSC MX2000, which were awesome amplifiers. They uesd the famous inside-out common “emitter” amplifiers and parallel supply commutators instead of series so it was fundamentally different from everybody else’s. Then class H started to appear everywhere.
Thanks, I was surprised with the high 93V rails, maybe for very high short term peak power?
Seems too high for 250W/ch but if they wanted, real, short term peaks of 400W/ch that
would probably do it and make it sound much more powerful.
Seems too high for 250W/ch but if they wanted, real, short term peaks of 400W/ch that
would probably do it and make it sound much more powerful.
Bob, I have not seen the mechanical details of this amp so this might be a dumb suggestion, but could you cut a rectangular hole in the steel chassis and bolt the spreader directly to the heat sink? As you know, steel is a very poor thermal conductor. When I was working, we once used stainless steel instead of our usual aluminum for an airborne electronics housing to protect the components from a short-term severely hot environment!
Bruce
Hi Bruce,
Thanks for the suggestion. It is certainly not a dumb question for the PL700 with plastic vertical MOSFETs. Without thinking hard about it, I would tend to worry that it might make the rear part of the chassis too weak, since the spreader is about as wide as the heat sink fins. I had thought of mounting the MOSFETs on the outside of the heat sinks, where the TO-3 parts used to be, and cutting holes to allow the pins of the MOSFETs protrude through the heat sink into the interior, but that is kind of ugly and would then require things on both sides of the heat sink, making things quite permanent. My inclination is the start by doing thermal measurements between the header and the heat sink as I described above. I am indeed concerned about the lower thermal conductivity and also the thermal conductivity loss of the heat sink compound on both sides of the steel. The only saving grace might be the sheer amount of area of the contact surface across the steel, on the order of 12 square inches. Come to think of it, I'll look online to see the ratio of thermal conductivity of Aluminum is to that of steel. BTW, do you know of some extra-high thermal conductivity heat sink compound?
Cheers,
Bob
Thanks for the suggestion. It is certainly not a dumb question for the PL700 with plastic vertical MOSFETs. Without thinking hard about it, I would tend to worry that it might make the rear part of the chassis too weak, since the spreader is about as wide as the heat sink fins. I had thought of mounting the MOSFETs on the outside of the heat sinks, where the TO-3 parts used to be, and cutting holes to allow the pins of the MOSFETs protrude through the heat sink into the interior, but that is kind of ugly and would then require things on both sides of the heat sink, making things quite permanent. My inclination is the start by doing thermal measurements between the header and the heat sink as I described above. I am indeed concerned about the lower thermal conductivity and also the thermal conductivity loss of the heat sink compound on both sides of the steel. The only saving grace might be the sheer amount of area of the contact surface across the steel, on the order of 12 square inches. Come to think of it, I'll look online to see the ratio of thermal conductivity of Aluminum is to that of steel. BTW, do you know of some extra-high thermal conductivity heat sink compound?
Cheers,
Bob
From what I remember of the stock heat sinks, and given the high bias you plan to use
I don't think that it will work. I did read that you plan to test it, curious about what you find.
I would replace the heat sinks with something much higher performance.
A fan might make the old work but I expect that a fan will be needed even with better
heat sinks.
I don't think that it will work. I did read that you plan to test it, curious about what you find.
I would replace the heat sinks with something much higher performance.
A fan might make the old work but I expect that a fan will be needed even with better
heat sinks.
Higher-than necessary rail voltages worry me a bit. The Phase linears are like that. Carver was an advocate of having high dynamic headroom in an amplifier, which meant a fairly soft power supply whose voltage at idle was significantly larger than that necessary to achieve the rated continuous power. Unfortunately, to best capitalize on that one should have generous reservoir capacitors, which he did not.
One passing thought I had was to deliberately lose some rail voltage with some quasi regulation thereof. I know that regulation can be costly dissipation-wise, but maybe not too bad if not dropping a lot of voltage, maybe like dropping 10-15V off the rails using vertical power MOSFETs. If doing so, the question arises as to whether the series MOSFET pass transistor is best put before or after the bridge rectifier. If before the bridge, the series device might tend to increase the rectification conduction angle, with some possible benefit. Peak currents into the reservoir capacitors might be significantly reduced on average. I note that when the load is light and the MOSFETs must drop a lot, their current is low. When the load is heavy and the average current is high, the voltage drop across them is low. Just thinking out loud here. It may be a crazy idea. Finding space for whatever heat sinking for the series MOSFETs is needed might be a challenge.
Cheers,
Bob
You know, I had a crazy idea for an adaptive regulator, where when current was below some threshold
it drops say 30V, then simply based on current, 3-5A snaps to full conduction. And the regulator should
be PWM for low dissipation. There was actually a TRIAC based soft start on here that comes to mind but
it would probably be too crude and slow. It would probably have to come after the huge PSU caps for
fast response time. Thinking about it now, that Sound Craftsman has a Triac in it, don't remember where.
This idea is not to regulate but lessen the Pdiss of the outputs, then snap to the full rail voltage when
required for peaks. Save on Pdiss but don't give up headroom. And much bigger caps.
it drops say 30V, then simply based on current, 3-5A snaps to full conduction. And the regulator should
be PWM for low dissipation. There was actually a TRIAC based soft start on here that comes to mind but
it would probably be too crude and slow. It would probably have to come after the huge PSU caps for
fast response time. Thinking about it now, that Sound Craftsman has a Triac in it, don't remember where.
This idea is not to regulate but lessen the Pdiss of the outputs, then snap to the full rail voltage when
required for peaks. Save on Pdiss but don't give up headroom. And much bigger caps.
Some of this annoying problems has prompted me to completely switch to ngspice. For a power user (as me?) the lack of scripting options was a major reason to use ngspice instead of LTSpice. "SA2021-NCH" is my first amplifier which was completely designed using ngspice (Version 40). The scripting allowed me to test all parameters in one brute force run: FFT, output power, dissipation, THD, Gain margin, Phase Margin, dissipation of used resistors, SOA and many more parameters.
The gained free time as a side effect I have used ... doing other crazy things...😉
Have fun, Toni
I always wondered if and of the CPU compounds would be good, but they might be electrically conductive, not sure.
Conductive compounds are what you WANT if you are forced to live with a heat spreader. You have to insulate NPN from PNP (or + from -), so you use insulators and non conductive compound on each transistor. For the heat spreader itself to heat sink use arctic silver or something similar.
Hi Pete,
I looked into the thermal issues in the PL700 regarding passing the heat from the heat spreaders through the metal chassis to the heat sink in order to use plastic power MOSFETs. I still intend to do an experiment passing heat and measuring the temperature rise on the heat spreader and the heat sink to evaluate the steel interface, but I usually like to do some rough calculations before an experiment, even is the calculation is just a SWAG.
I looked up the thermal conductivity of steel and it is only about 45 watts per Kelvin per meter, where the meter is the ratio of contact area to distance the heat must travel, i.e. the thickness of the steel. By comparison, for Aluminum it is 203 watts/K-m, more than 4 times as great. However, in my planned arrangement the area is large and the thickness is small. The 6.75" by 1.5" heat spreader has area of 0.0065 square meters. The 16-AWG chassis steel has a thickness of 1.5 mm. The ratio is 4.3 meters. When the 45 w/K-m is multiplied by 4.3 meters, we have thermal conductivity of 194 watts/K, or thermal resistance of 0.0052 K/watt. This number seems surprisingly low, and seems too good to be true. It is either due to the large ratio of contact area to steel thickness or an error on my part 🙂.
Even if the above is true, there is another potential fly in the ointment - the black finish on both sides of the steel chassis, presumably some kind of paint. The thermal conductivity of paint is very low, at about 0.2 w/K-m. However, the thickness is small, typically at 30-100 microns. At 50 microns, the thermal resistance of the paint on each side would come to about 0.039 K/w per side, again, if I got it right. The total paint thermal resistance for two sides is 0.078 K/w, much larger than the 0.0052 K/w for the steel. But the total is still less than 0.1 K/w, still maybe too good to be true. If the presence of the paint is a big problem, I can sand it down in the contact area.
I also looked at CPU thermal grease online. There is a big range of thermal conductivity, with one really impressive one claiming 12.8 w/K-m.
Cheers,
Bob
It's been 3-5 years since I looked up testing on CPU thermal compounds and I chose MX-4 or MX-6, just
noticed that they claim it is not electrically conductive, should I use it on mica insulators?
https://www.amazon.com/ARCTIC-ACTCP00080A-MX-6-4-g/dp/B09VDL3CW6
https://www.electronicshub.org/arct...vity (10.5,drying and non-bleeding properties.
I'd be careful about who's compound you use because who knows how durable they are, some computer
guys change the thermal compound every 3-5 years.
Nice job on calculations but I was more worried about the main heatsinks being so small, we will
see what you find.
noticed that they claim it is not electrically conductive, should I use it on mica insulators?
https://www.amazon.com/ARCTIC-ACTCP00080A-MX-6-4-g/dp/B09VDL3CW6
https://www.electronicshub.org/arct...vity (10.5,drying and non-bleeding properties.
I'd be careful about who's compound you use because who knows how durable they are, some computer
guys change the thermal compound every 3-5 years.
Nice job on calculations but I was more worried about the main heatsinks being so small, we will
see what you find.
Hi Bob,
You might want to check on what fillers are used in the compounds. I’m thinking there could be a dissimilar metals and corrosion issue. Perhaps the grease they are suspended in prevents this from becoming a real problem. Just a thought.
Bruce