@wg_ski
If a PWM regulator can't be done then perhaps what the PL700 needs is a series connected
output stage. Now, there are only 5 outputs but there's also a TO3 driver, so move the driver
to a new driver board to make room for 6 outputs on the top and bottom and put them in
series. If the devices can take the full supply voltage then even if the division is off now and
then it probably wouldn't harm anything. I believe that Bryston was still using a series output
stage not too long ago - it can be made to work. It probably works better if the transistors are
kept out of saturation. This would probably be an excellent mod for the PL700 but I prefer
something more efficient.
What is the simplest Class G or H design around that works well?
If a PWM regulator can't be done then perhaps what the PL700 needs is a series connected
output stage. Now, there are only 5 outputs but there's also a TO3 driver, so move the driver
to a new driver board to make room for 6 outputs on the top and bottom and put them in
series. If the devices can take the full supply voltage then even if the division is off now and
then it probably wouldn't harm anything. I believe that Bryston was still using a series output
stage not too long ago - it can be made to work. It probably works better if the transistors are
kept out of saturation. This would probably be an excellent mod for the PL700 but I prefer
something more efficient.
What is the simplest Class G or H design around that works well?
The way to get voltage division to work well at high power is to use EF3. Yeah, two drivers but the can be MJE15032/3 and mount “somewhere” on the backplane and not in one of the TO-3 sockets. Thats’s right, Super Leach output stage. I’ve done it up to +/-127V (10 outputs). Those amps blow like hair dryers and make the plugs all melty at war volume. It would be practical to do the PL700 with 6. One would need to use MJ2119x.
The most straightforward and foolproof class H switch is the one QSC uses, with the floating 311 op amp and complementary pair driving a hexfet. Shoehorning them in to the existing layout and being able to maintain proper local supply bypassing would be problematic. If one redesigned the backplane to be on a PCB instead of PTP it could probably be done. Stacking boards is possible then.
Class G with a “Linear” set of upper transistors looks attractive until one realizes that class G or H do NOT reduce the peak power dissipation per transistor in the output stage, only the average. Cutting the number of active outputs to three won’t fly. Four might, but there aren’t eight sockets and no room for that many. You might as well redesign around flatpacks if you do that.
The most straightforward and foolproof class H switch is the one QSC uses, with the floating 311 op amp and complementary pair driving a hexfet. Shoehorning them in to the existing layout and being able to maintain proper local supply bypassing would be problematic. If one redesigned the backplane to be on a PCB instead of PTP it could probably be done. Stacking boards is possible then.
Class G with a “Linear” set of upper transistors looks attractive until one realizes that class G or H do NOT reduce the peak power dissipation per transistor in the output stage, only the average. Cutting the number of active outputs to three won’t fly. Four might, but there aren’t eight sockets and no room for that many. You might as well redesign around flatpacks if you do that.
Last edited:
Yamaha EEEngine. The implementation requires Sanken LAPT outputs. It combines a small parallel stage to handle the fast rising edges with a bigger set out outputs fed by a tracker to handle bulk currents.
With anything resembling the PL it would be problematic. Perhaps just do regular class H, but generate the mid rails with simple buck converters. They wouldn’t need to be co-located within the amplifier circuit, and would allow the use of the original transformer. Feeling adventurous? Go 3 or 4 step instead of two. That has the further advantage of reducing peak dissipations which were the PL’s main weakness to begin with. But that more switches that need to be shoehorned into the output stage itself.
You know, I was thinking, what about double die outputs but there are no modern ones
as far as I know in BJTs but there are the Exicons in Laterals. Also should not have SOA
issues. I've not looked at them in a long time have to look them up.
Not in the spirit of the original design but would make it reliable and probably sound good.
I'm finding 200V, 16A 250W are these the double die version?
If they really do not have SOA issues they'd probably work, also without cross conduction.
as far as I know in BJTs but there are the Exicons in Laterals. Also should not have SOA
issues. I've not looked at them in a long time have to look them up.
Not in the spirit of the original design but would make it reliable and probably sound good.
I'm finding 200V, 16A 250W are these the double die version?
If they really do not have SOA issues they'd probably work, also without cross conduction.
Last edited:
The only things I would keep about “the original design” is the general topology of the front end (of the series 2), the TO-3’s mounted on those external heat sinks (and the overall form factor of the amp), and the cool LED meters. The only way to deal with the SOA issues is to series output devices or Class-H it. The only way to deal with the bulk heat is to class H it. You can’t class H the original QC triple. Making something out of flatpack outputs is entirely new - might as well just put it in a modern 2U case. It’s better optimized that way anyway.
What I kind of want is something that looks like a PL700 S2 but can turn speakers to charcoal before the amp even thinks about crying uncle. I’d be running the compression drivers off an existing 400 S2, with DC protect relays, of course. Those “BGW’s” I’m working on would feed subs. Why? A re-creation of a system I was using back in 85 (and too poor to own then). Stuff I own now blows it away, but I’ve been having those Benjamin Sisko moments for almost 40 years now.
What I kind of want is something that looks like a PL700 S2 but can turn speakers to charcoal before the amp even thinks about crying uncle. I’d be running the compression drivers off an existing 400 S2, with DC protect relays, of course. Those “BGW’s” I’m working on would feed subs. Why? A re-creation of a system I was using back in 85 (and too poor to own then). Stuff I own now blows it away, but I’ve been having those Benjamin Sisko moments for almost 40 years now.
Yes, the challenge is to make it reliable in the original package and if possible more efficient,
if not then mandatory fan cooling.
Okay, thinking more about it, if a digital amp amplifies DC, even at unity gain, then it can work
as a sort of "tracking regulator" to provide 1/2 or even 1/4 of Vout across the output devices.
Maybe one on each power rail. Of course why not just go digital, but this way we keep the
analog nature of the original PL700.
If you remember, I think it was, Class AA in the AES. Run a small "clean", perhaps Class A amp
and modulate the power supply with a less good "dirty" amp. I could think of the small amp
as "error correcting" the bigger amp but it was not described that way.
Can't find the Class AA article right now but it is in my files.
I also like the double die Exicon idea as long as they are reliable with such high rails, no idea.
if not then mandatory fan cooling.
Okay, thinking more about it, if a digital amp amplifies DC, even at unity gain, then it can work
as a sort of "tracking regulator" to provide 1/2 or even 1/4 of Vout across the output devices.
Maybe one on each power rail. Of course why not just go digital, but this way we keep the
analog nature of the original PL700.
If you remember, I think it was, Class AA in the AES. Run a small "clean", perhaps Class A amp
and modulate the power supply with a less good "dirty" amp. I could think of the small amp
as "error correcting" the bigger amp but it was not described that way.
Can't find the Class AA article right now but it is in my files.
I also like the double die Exicon idea as long as they are reliable with such high rails, no idea.
The Yamaha trackers keep about 15 volts across the outputs. The switching frequency of the trackers is some 40 KHz. The trackers are self-oscillating buck converters. It also has a single output transistor per rail in parallel with the whole works, to fill in any short transient that the trackers can’t track and cause the main outputs to drop out. I have my doubts about that part of it working well with 4 MHz power transistors. And even the rest of it will probably have difficulty in a PTP layout. You really need a compact PCB where everything is electrically small to get the level of performance that they do. Class H switching works fine with EF3’s - even when it is a rail full of 4MHz TO-3’s.
I suppose one could build a Yamaha clone that physically fits in the PL 700 chassis, using a bunch of 2SC5200’s mounted to a flat aluminum plate, then bolt that to the inside of the PL heat sinks. Dummy transistors could even be installed. You’d want to find dead PL909’s of course to make it look original.
I suppose one could build a Yamaha clone that physically fits in the PL 700 chassis, using a bunch of 2SC5200’s mounted to a flat aluminum plate, then bolt that to the inside of the PL heat sinks. Dummy transistors could even be installed. You’d want to find dead PL909’s of course to make it look original.
The Hafler DH500 is a good comp, also with high rails at +/-90V and it uses only 3 pairs of 125W outputs,
seems to me that 5 pairs of standard Exicons and certainly 5 pairs of the dual die parts would easily run
in the PL700. The DH 500 is rated to drive 2 ohm loads and with 5 pairs I'd expect it to be easy for the PL700.
I own a DH500 and it has never given me any trouble.
seems to me that 5 pairs of standard Exicons and certainly 5 pairs of the dual die parts would easily run
in the PL700. The DH 500 is rated to drive 2 ohm loads and with 5 pairs I'd expect it to be easy for the PL700.
I own a DH500 and it has never given me any trouble.
But the TO-3 Exicons are gone. Still making TO-247. Back to that particular problem again.
Unless you want to take apart two perfectly good DH500’s. I’m crazy but not that crazy.
Unless you want to take apart two perfectly good DH500’s. I’m crazy but not that crazy.
I was not aware of that. Use TO-247 s and put a cover over them?
https://apexfasteners.com/fasteners...t-board-hardware/power-transistor-covers-to-3
Goal would be to make it reliable first.
https://apexfasteners.com/fasteners...t-board-hardware/power-transistor-covers-to-3
Goal would be to make it reliable first.
I built a power amplifier in 1980 using the exact Otala topology, albeit with different semiconductors and compensation part values. With 2 pairs of MJ802/MJ4502 output transistors and a power transformer that I custom rewound, it produces 50W into 8 ohms and 75W into 4 ohms. Achieving unconditional stability with its 3 gain stages and lead/lag compensation was laborious but ultimately successful. Lacking any tools like Spice, I manually generated gain and phase plots and optimized them after the addition of each stage. The final key to success was the Miller Input Compensation. There are those who extol the MIC approach but they appear to be in the minority. Bob Cordell does not recommend MIC for audio amplifiers. Having built two power amplifiers using it, I can say that it may be made to work very well. What has worked for me is to configure the amplifier with a differential (balanced) input and drive it with controlled source impedances. The excess noise of MIC is out of band and is not really an issue.
My amplifier produces lovely square waves even into complex loads and has never had a stability problem that I am aware of during its 44 years of service. I still use it to drive remote speakers.
Excellent! I always wondered if it could be improved. Do you have a schematic handy
if you don't mind sharing it?
if you don't mind sharing it?
Sadly, I have only this embarrassing "hen scratching." In 1980 I had an excellent memory and only needed to look at the Otala schematic and I would remember the changes [who needs a stinking schematic! 🙂]. In 1990 I reverse engineered the amplifier but never got around to drawing a pretty schematic. One may still need to reference the Otala schematic to fully understand it. I will try to answer answer any questions that you might have.
I've had the same experience, and I am very unhappy about it. I was hoping that ADI would fix this in an updated software release, but I have seen no evidence of that. Probably not as well-maintained since Mike Engelhardt left. There are thousands of LTspice sim files out there that likely suffer from this. For this reason, I still usually use an earlier version of LTspice IV for many of my sims.
Cheers,
Bob
@Bob Cordell Weird, it won't let me quote your post. Edit: Now the quote option shows up.
Here's the solution to fix small fonts from post #28 thanks to rongon.
This works, the icon launches the program with this command, add the - options as shown:
Make an LTSpice icon on your desktop, then edit the properties with these dash options:
"C:\Program Files\LTC\LTspiceXVII\XVIIx64.exe" -FixUpSchematicFonts -FixUpSymbolFonts
Open old .asc files by dropping them onto this icon, you might have to save it for the fix,
then reopen.
Here's the solution to fix small fonts from post #28 thanks to rongon.
This works, the icon launches the program with this command, add the - options as shown:
Make an LTSpice icon on your desktop, then edit the properties with these dash options:
"C:\Program Files\LTC\LTspiceXVII\XVIIx64.exe" -FixUpSchematicFonts -FixUpSymbolFonts
Open old .asc files by dropping them onto this icon, you might have to save it for the fix,
then reopen.
Last edited:
I don't recall saying I did not like Miller Input Compensation, but don't recall the context you are referring to. I actually used MIC in my "MOSFET Power Amplifier with Error Correction" (available JAES paper on my website at cordellaudio.com) way back in 1984. The version I used took the output of the VAS back to the input of the input stage. This required some lag-lead compensation of that loop (as described in my paper), but that compensation did not diminish performance much, since it was in a fast loop that did not include the output stage.
Cheers,
Bob