ostripper said:I really like this amp, I was a little scared of the YAP..
I've heard the folded design before (roenders amp) but this
is symmetrical (I like).
About the 1530/ 201's are they vertical ? I have a pile
of IRFP 240's/9240's (free)
OS
K1503/J201 are verticals, so are 240/9240 but not good replacements; they are matched for capacitances rather than for transconductance as the K1530/J201. Which makes them good for switching rather than for linear applications. 240/9240 can be used though, but don't expect the same results.
That toroid will do
At 320W into 4 ohms, what is the kHz for 0.1% THD?
An externally hosted image should be here but it was not working when we last tested it.
Regards
Mike
Ok, I decided to bring the gate stoppers down to 150/100ohm (same as in YAP). This, and a slightly different compensation (a bit of lead compensation in the global loop, see the attachment) brought the ULG up to 2.2MHz with a slightly better phase margin. Distortions are now measuring down to 10ppm (0.001%)at 360W and 4 ohm load. Interesting enough, 3rd harmonic is now dominant, nothing over the 4th. Nice to see harmonics of around 200uV amplitude, buried in 50V of fundamental
That's pretty much it from a design perspective, now finalizing the other measurements.
That's pretty much it from a design perspective, now finalizing the other measurements.
An externally hosted image should be here but it was not working when we last tested it.
A Nigel Wright masterpiece:
http://www.diyaudio.com/forums/showthread.php?postid=1542476#post1542476
Expert in supply rail bypassing.
http://www.diyaudio.com/forums/showthread.php?postid=1542476#post1542476
Expert in supply rail bypassing.
Bipolar version
Ok, so I finished experimented a version with power bipolars (OnSemi NJL4281/NJL4302 ThermalTraks). Schematic is attached.
It requires (for the same performance) a higher ULG (3.8MHz)than the MOSFET version. I don't really understand why, the triplet bias doesn't seem to have any impact (simulations also show the same thing).
Anyway, it works perfectly fine. I'm though not sure which version I'm going adopt for the final build. Not having much experience with, I'm a little shy of power bipolars. Not sure if there is any significant (audible) performance difference between ThermaTraks and (e.g.) 2SC5200/2SA1930 and conventional thermal compensation. Obviously, this is much more complex than any stationary effects...
Is the close thermal tracking really an audible improvement? I've read the threads about ThermalTraks, I understand the phenomena, but from a practical perspective I think the situation is still unclear. That's because routing a decent high ULG two layer PCB board with ThermalTraks seems very difficult and the result may potentially degrade the performances more than the close thermal tracking would improve...
Anybody, any clues/suggestions?
Ok, so I finished experimented a version with power bipolars (OnSemi NJL4281/NJL4302 ThermalTraks). Schematic is attached.
It requires (for the same performance) a higher ULG (3.8MHz)than the MOSFET version. I don't really understand why, the triplet bias doesn't seem to have any impact (simulations also show the same thing).
Anyway, it works perfectly fine. I'm though not sure which version I'm going adopt for the final build. Not having much experience with, I'm a little shy of power bipolars. Not sure if there is any significant (audible) performance difference between ThermaTraks and (e.g.) 2SC5200/2SA1930 and conventional thermal compensation. Obviously, this is much more complex than any stationary effects...
Is the close thermal tracking really an audible improvement? I've read the threads about ThermalTraks, I understand the phenomena, but from a practical perspective I think the situation is still unclear. That's because routing a decent high ULG two layer PCB board with ThermalTraks seems very difficult and the result may potentially degrade the performances more than the close thermal tracking would improve...
Anybody, any clues/suggestions?
An externally hosted image should be here but it was not working when we last tested it.
I haven’t looked up the specs on the thermal-trak transistors, but are you sure you want 5 diodes plus a pot compensating 4 Vbes? Especially considering that two of those Vbes (Q91 & Q92) probably don’t even get that hot. Seems more like 3 diodes plus a pot would be closer to the mark?
I’m also curious about the pick-off points for of the compensation networks at Q91 and Q92’s emitters. Seems like the collectors of Q14 and Q15 would be the natural pick-off points. Have you built prototypes, or are things still at the sim level?
Why the change to the OPA637? The THS 4031 seemed to have some pretty decent specs.
I’m also curious about the pick-off points for of the compensation networks at Q91 and Q92’s emitters. Seems like the collectors of Q14 and Q15 would be the natural pick-off points. Have you built prototypes, or are things still at the sim level?
Why the change to the OPA637? The THS 4031 seemed to have some pretty decent specs.
Looks interesting.
The term "current feedback" is ambiguous - it usually means that the feedback is proportional to output current, but is also sometimes used to mean that feedback is to a low-impedance node, although neither is the case here. It would be better to described the output stage as having a common-base input.
Also... ULG? I don't recall having seen that acronym before. From the way you've used it I assume it means unity-gain frequency?
How about replacing Q79/80/81/82 with a pair of JFETs, and cascoding them with another pair of JFETs instead of Q13/10? It would greatly simplify biasing, reducing component count by 6.
The term "current feedback" is ambiguous - it usually means that the feedback is proportional to output current, but is also sometimes used to mean that feedback is to a low-impedance node, although neither is the case here. It would be better to described the output stage as having a common-base input.
Also... ULG? I don't recall having seen that acronym before. From the way you've used it I assume it means unity-gain frequency?
How about replacing Q79/80/81/82 with a pair of JFETs, and cascoding them with another pair of JFETs instead of Q13/10? It would greatly simplify biasing, reducing component count by 6.
kaos said:
1. Unfortunately not. The incorporated diodes are far from the Vbe both voltage drop and thermal coefficient. And all junctions are hot, being mounted on the same heatsink. I'll try other bias schemas as well, but this kind of issues (that also can be barely simulated due to the lack of appropriate models) is one of the reasons why I'm not a big fan of power bipolars.
2. Long story here, the basic idea is to avoid loading the VAS at high frequencies. There are pros and cons as usual, read about in the Negative Feedback thread.
3. Looking for high speed FET input opamp. THS4031 is decent but has input bias current that creates an output offset. It happens that I have some OPA637 and they are in DIP8, much more convenient for experimenting. BTW, THS4501 works great as well, but is SOT8.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.