Hello everybody,
I hope I did get the name of the topology right.
The story is as follows: I managed to get an excellent deal on two used 1600va 2x115V secondary toroidal transformers and right now I'm contemplating building a very large amplifier around these two.
Unfortunately there are no additional secondary taps available and thus the amplifier will have to be class B with no sliding rails.
There are complementary 350V bipolar transistors available from On-Semi, but the problem with these is that they're almost useless at these voltages due to SOAR limitations. I thought the next best thing would be to connect the output transistors in series just like in the "low-end zesilovac 1kW" from Alan Kraus. Czech magazine "amatorske radio" published this design back in 2000. I don't think it is worth the effort to build this design as both input stage and VAS are very simple and I assume that the performance won't be very good. In any case, I have a pdf available if anyone is interested.
I did like the output stage however and I tried designing something similar using more modern plastic-cased transistors. However the problem with this type of design is that I cannot get it running stable. I've run many simulations over the past weeks and I think the instability issues can be really narrowed down to the output stage itself.
I've attached a stripped-down schematic which shows just the output stage with 'ideal' VAS and bias generator. The circuit appears to be very sensitive to changes in R1 and there is significant gain above 1MHz or so. This could explain why even minor changes in the VAS stage make a mess of everything.
I've read critical comments about power cascades or totem pole arrangement output stages and I'm not so sure what to do with this information. I haven't seen this kind of arrangement anywhere except with the Alan Kraus design and the Leach Superamp and those two were using older and slower power transistors.
Right now I'm not sure if a complementary bipolar power stage is possible for such high supply voltages. I would appreciate any input or perhaps schematics for similar designs.
I hope I did get the name of the topology right.
The story is as follows: I managed to get an excellent deal on two used 1600va 2x115V secondary toroidal transformers and right now I'm contemplating building a very large amplifier around these two.
Unfortunately there are no additional secondary taps available and thus the amplifier will have to be class B with no sliding rails.
There are complementary 350V bipolar transistors available from On-Semi, but the problem with these is that they're almost useless at these voltages due to SOAR limitations. I thought the next best thing would be to connect the output transistors in series just like in the "low-end zesilovac 1kW" from Alan Kraus. Czech magazine "amatorske radio" published this design back in 2000. I don't think it is worth the effort to build this design as both input stage and VAS are very simple and I assume that the performance won't be very good. In any case, I have a pdf available if anyone is interested.
I did like the output stage however and I tried designing something similar using more modern plastic-cased transistors. However the problem with this type of design is that I cannot get it running stable. I've run many simulations over the past weeks and I think the instability issues can be really narrowed down to the output stage itself.
I've attached a stripped-down schematic which shows just the output stage with 'ideal' VAS and bias generator. The circuit appears to be very sensitive to changes in R1 and there is significant gain above 1MHz or so. This could explain why even minor changes in the VAS stage make a mess of everything.
I've read critical comments about power cascades or totem pole arrangement output stages and I'm not so sure what to do with this information. I haven't seen this kind of arrangement anywhere except with the Alan Kraus design and the Leach Superamp and those two were using older and slower power transistors.
Right now I'm not sure if a complementary bipolar power stage is possible for such high supply voltages. I would appreciate any input or perhaps schematics for similar designs.
Attachments
I have found two areas that can cause instability in these kind of amps.
1/ The VAS stage capacitor needs to be just right. Usually 100pf
2/ The resistor between driver and output transistors needs to be right too. I put 10R in series with each output transistor base.
1/ The VAS stage capacitor needs to be just right. Usually 100pf
2/ The resistor between driver and output transistors needs to be right too. I put 10R in series with each output transistor base.
You do have a point there. Stepping the output transistor base resistors does make a significant difference here. Didn't expect the difference to be that large to be honest.
Attachments
One more consideration, if I may - with such a big number of the output pairs (10 pairs in each "barrel"), considering the sum of the base currents and total capacitance - I would either use something like njw0281/0302 (or even 3281/1302) for the drivers, or split the output pairs into two groups, 5 pairs each, having the separate drivers for each group.
This may also increase the "speed" of the OPS, positively influencing the overall stability.
This may also increase the "speed" of the OPS, positively influencing the overall stability.
Something is off with the biasing.... getting a lot of DC on the output.
Try using two 1.2V sources in series instead of the 2.4V, and connect your signal to the center of them.
The signal source acts like a DC short and is keeping the top of the 2.4V source near ground potential.
Try using two 1.2V sources in series instead of the 2.4V, and connect your signal to the center of them.
The signal source acts like a DC short and is keeping the top of the 2.4V source near ground potential.
Corrected that.
@vzaichenko:
True. I'll consider that later on. Right now the drivers are very likely outside the SOAR for Tc=100°
@vzaichenko:
True. I'll consider that later on. Right now the drivers are very likely outside the SOAR for Tc=100°
115V+115V makes 322V total so over 100Vrms.
100V in 8 Ohms is 1,250 Watts. That appears to be a heavy load on 1,600VA transformers. We usually like 2X or 2500VA. Granted that there is no speaker which will comfortably take all of a 1250W amp long enough to warm-up the transformer, so this may be OK.
My rule of comfort for plastic transistors is at-most 50W output per pair. 40W may be safer. 1250/50 suggests 25 pair, 32 pair may be safer. You have 20 pair. I'd want more, but it should probably survive.
If changing the base resistor *radically* changes the frequency response (as it seems to do), then the amp is VERY load-sensitive and will probably go crazy on a real loudspeaker load.
Your peak output current is 20 Amps. While a triple Darlington has huge current gain, with base shunt resistors and all it may need 2mA peak into the array. This is a heavy load on any VAS stage.
The less-insane path may be to make single +160V DC to power cap-coupled amplifiers, four channels of 300 Watts. Or combine to a Bridge for 1200W (though with +80V DC on the speaker terminals!).
100V in 8 Ohms is 1,250 Watts. That appears to be a heavy load on 1,600VA transformers. We usually like 2X or 2500VA. Granted that there is no speaker which will comfortably take all of a 1250W amp long enough to warm-up the transformer, so this may be OK.
My rule of comfort for plastic transistors is at-most 50W output per pair. 40W may be safer. 1250/50 suggests 25 pair, 32 pair may be safer. You have 20 pair. I'd want more, but it should probably survive.
If changing the base resistor *radically* changes the frequency response (as it seems to do), then the amp is VERY load-sensitive and will probably go crazy on a real loudspeaker load.
Your peak output current is 20 Amps. While a triple Darlington has huge current gain, with base shunt resistors and all it may need 2mA peak into the array. This is a heavy load on any VAS stage.
The less-insane path may be to make single +160V DC to power cap-coupled amplifiers, four channels of 300 Watts. Or combine to a Bridge for 1200W (though with +80V DC on the speaker terminals!).
Even if you manage to get simulations stable, the real thing will be difficult to stabilize. Parasitic capacitances and inductance which do not show up in your model start to get significant when you parallel that many devices. Don't try to get a crazy amount of open loop gain. Use local feedback in each stage like Leach did, then the overall loop has a chance of being stable. Local instabilities in the OPS will have to be dealt with separately. I only built a 5 pair per barrel and found that base stoppers on everything and zobels for the upper banks were necessary. As well as tinkering with the values of the base to base resistors on the outputs and drivers. Any capacitance to ground at the node between master and slave banks creates a parasitic colpitts oscillator (negative R at the base). You have to de-Q it. Output stages are much more unstable at high vce, regardless of whether stacked on not. When testing on a dim bulb or variac it may be perfectly stable, but fly when given full voltage. If you could do class H it would mitigate that - because vce is never crazy high if the transistor is conducting. Would it be possible to add a pair of 55-0-55 secondaries at half the VA rating? If the window is large enough you can get more copper in. Another possibility is to use the Yamaha EEEngine output stage. It has sliding rails but doesn't require extra taps because it is switch mode (and relatively simple with no critical dead time requirements). Of course you could just buy a 1000 va 55-0-55 and use it with the 115 (two different transformers). Then you could build a pair of amps which would be easier to design/build. You could clone the amp circuit of the Crest Pro 9200 which is proven and runs off +/- 80 and 160.
Reached this point in your personal line of investigation of amplification techniques, I recommend learning switching regulators, it can change your view about class AB output stages.
The high transformer voltage is a good starting point for the usage of switching regulators in the rails. This can simplify output stage considerably, just a few pairs of power bipolars, for same power as a few dozens without switching regulator. You can build more amplifiers with same amount of parts.
The switching regulators themselves are halves of a class D amplifier, but these do not need to be very accurate, so this is a platform for learning class D and SMPS.
There are known brands and products using this approach, but the point of DIY is doing it in a different flavor, maybe a more nutritive one, and documenting/sharing the process.
The high transformer voltage is a good starting point for the usage of switching regulators in the rails. This can simplify output stage considerably, just a few pairs of power bipolars, for same power as a few dozens without switching regulator. You can build more amplifiers with same amount of parts.
The switching regulators themselves are halves of a class D amplifier, but these do not need to be very accurate, so this is a platform for learning class D and SMPS.
There are known brands and products using this approach, but the point of DIY is doing it in a different flavor, maybe a more nutritive one, and documenting/sharing the process.
Even larger transformers would be even better probably, but you have to keep in mind that "schuko" outlets here will provide 16A@230V at most. 16A line safety switches are common. (10A and 13A can be found in older households as well.) This equals to about 23A max current or so for one hour until the breaker drops. At least that's what the datasheet of a common household line safety switch suggests ;-)
I might change my opinion if I manage to find a third transformer though. If I think about it having a third transformer would even add an option for running the amplifier from three-phase outlets...
@wg_ski: I'm not sure if adding a 2x55V transformer is the most cost-effective idea, but having the secondaries reworked could be an option if all else fails. No idea about the involved costs though.
I do like the Yamaha EEEngine concept however and right now I think this is the way to go in my case.
@Eva: I didn't think about using switch-mode supply regulators until now, but I will be writing my master's thesis next year and this might be actually an interesting topic.
If you are considering switching regulators, you can pretty much have any rail voltage you want, up to 160 volts. Even the multiple taps required for class H. But the Yamaha circuit or some other variant of class TD is actually simpler at that point. You could develop your own tracking preregulator rather than just clone something. Or just step 160 down to 80 fully regulated, and build a killer class AB that doubles its output power at 4 ohms and doubles again at 2.
Of course there is always the option of doing straight class D. At that voltage/power level it would be a long development process because you would need a from the ground up all discrete solution. No IRAUDAMP reference circuit to work from, no cute little all in one driver chip. Don't expect that to work the first time you turn it on. Unless you intentionally design for a low (say 25khz) switching frequency, a lot of dead time (like that built into standard SMPS controller chips that you would use for your buck converters) and make it subwoofer-only. That may work without a lot of fuss.
Of course there is always the option of doing straight class D. At that voltage/power level it would be a long development process because you would need a from the ground up all discrete solution. No IRAUDAMP reference circuit to work from, no cute little all in one driver chip. Don't expect that to work the first time you turn it on. Unless you intentionally design for a low (say 25khz) switching frequency, a lot of dead time (like that built into standard SMPS controller chips that you would use for your buck converters) and make it subwoofer-only. That may work without a lot of fuss.
Dennis Feucht has written a classic analysis of Emitter Follower stability, well worth the effort to find on the 'net.
Make each transistor ideal, infinite Ft, zero capacitance and see which is the problem as you systematically move towards reality.
You have Q61 and Q62 for the "outside" half but no equivalent on the "inside", is that deliberate?
Best wishes
David
Thank you. Do have a schematic for this PCB layout as well?
@Dave Zan: Thank you for the suggestion with Dennis Feucht. I found some pdf online, but I'll try to track down his book as well.
Well, sort of. One of the earlier iterations of this circuit had the voltage dividers connected to the output and these transistors seemed necessary for clean output wave forms with large voltage swings.
- Status
- Not open for further replies.