thermal compensation

[ilimzn]

Try a Vgs multiplier instead or as Chocoholic mentioned,an emitter resistor in the Vbe multiplier.
For the Vgs multiplier you need a MOSFET with reasonably high Gm, so use a higher current one, IRF520, 530 or even 540. Capacitances will not be a problem here as you need to bypass it with a cap anyway (Vgs multipliers are much worse voltage sources than Vbe multipliers due to lower gain).
Since the typical threshold voltage of the IRFP240/9240 is about 3V for some tens of mA of current at 25C, and you have two Vgs to contend with, the Vbe multiplier will need to multiply the typical Vbe by at least 10, so the thermal gradient of Vbe of 2mV/K on the BJT will also be multiplied 10 times, to 20mV/K. which is about 2-3x too much for a VMOS. You will need to experiment with an emitter resistor in the Vbe multiplier. This will depend largely on the current through it, The source resistors, and to a smaller degree on the bias current chosen.
Also, if your supply rails are below 45V or so, you may want to condiser using IRFP240 and IRFP9140 as a complementary pair (both without the N at the end!) - they tend to be more complementary than 240/9240 in nearly everything save the maximum Vds in your design this might reduce distortion or you may be able to reduce idle current.

[anatech]

Hi All,
You really do not want to copy a Counterpoint in any way. They are not very good. You also do not want to use IRF mosfets as output devices as in a Counterpoint. The sound quality is very poor. It escapes me how they sold these things to begin with.

A BJT with a degenerated emitter works okay for bias with mosfet outputs. The later SA-100's ran at +- 50 VDC for rails.

I'll say it again, if you copy a Counterpoint, you end up with Counterpoint problems. They are many.

-Chris

[djdamix]

Here is the amp as it is now (the resistor in serie with the VBE x is not on the schema, but I've trying it and it seems to be a good solution !)

Any comments or improvement is of course welcome ;-)

Thanks.
David

[ChocoHolic]

Do you really need M4 & M6 as current gain stage?
If you are using MosFets I would try deal without such a
"darlington" arrangement.
May be you would need to enforce your VAS, but your VAS BJT still have margin. If we assume max. 140V Vce then you can allow 40mA.
My proposal would be to reduces R8 down to 33 Ohms.
But this would require a much higher current setting in you differential input stage.
BTW already now (about 7mA idle current in the VAS ? ) your differential input stage is running at a low idle current. The current into the base of Q21 is not neglectable....

For the input stage I would propose the follwing:
clamp the bases of the differential input to about +/- 15V.
(Two reverse-serial-connected zeners from basis to ground will do it.
==> four 15V- zener diodes)
Additional you can connect 33V zeners in series to the collectors of Q23 und Q24. (==> two further 33V zeners).
If you add the clamping and the collector zeners, then the voltage
load for Q23 and Q24 will be reduced so that you can use BC546 small signal transistors with high gain.....

Just my two cents....

Have fun
Markus

[AndrewT]

Hi Markus,
I cannot understand your suggestions for using the zeners. Could you post a schematic?
Djdamix,
the ltp appears to have 1mA in the tail (r30 560r), the load resistor (r5 560r) also has about 1mA. This leaves the other half of your LTP (q24) cutoff or very low current.
Try increasing the tail current to about 2mA to balance the LTP. This will increase the first stage gain but you can reduce it by introducing emitter degeneration resistors for q23&24.

[ChocoHolic]

Sorry for poor qual of pic.
Little bit in hurry.

Just grey theory...
-Zeners: 4x 15V, 2x 33V
-R8: 33 ( Use HEAT SINKS for Q21 and Q2 ! )
-R30: 150
-R5: 270
-R1&R32: to be set new
-100nF across Vbe-mult
-Degen.-res. in Vbe-mult

[ilimzn]

Originally posted by anatech
Hi All,
You really do not want to copy a Counterpoint in any way. They are not very good... The sound quality is very poor. It escapes me how they sold these things to begin with... I'll say it again, if you copy a Counterpoint, you end up with Counterpoint problems. They are many. -Chris

I can only agree with teh above!
SA-100 looks like a high-school science fair project which was backed up by several 10000$ worth (and I am trying to be kind here). The 'Advanced dual output power transformer' is an outright scam, the outputs are connected inside it!!!

As for:
"You also do not want to use IRF mosfets as output devices as in a Counterpoint."

Actually you can use IRFP (not IRF) with fine sound quality IF you know how to pair them. I'll just say this: IRFPXXX and 9XXX are NOT the best pairing (by far) but you can find at least 2 pairs (for rails up to +-45 and up to +-90) that will work great. The Counterpoint does not use the correct ones (though it's a moot point considering they are not made any more), but that's the least of it's problems. The MOSFETs are the least responsible for it's questionable (to say the least) sound...

[anatech]

Hi ilimzn,
Actually ... it's all bad, but, replace the outputs with a bipolar circuit and the sound has improved by a large amount. The IRF fets have an abrupt gate charge change. That is hard to drive.

The entire thing is a bad dream.

I started by trying to eliminate the design errors that made them blow up and sound terrible. One fix lead directly to another until there wasn't much that remained the same. Sounds pretty good now (compared to the original). Still room for improvement but they are on the right track.

I use a stock unit to compare against a modified one against the next step (that's three SA-100's). This way I can keep improvements on track sonicaly and with measurements. That's why I say, "don't even go there" when someone asks about a Counterpoint amp.

-Chris

[amplifierguru]

Hmmm....

I'm planning to use 3-4prs of IRFP240's in an Nch only nested feedback 150W monoblock - as soon as I sort my test eqpt.
Just got to build a low THD wein bridge osc and canyon filter.

They're the best value power around and with a high CMRR FET frontend and nested high PSRR design I'll get my target 0.002% at 20KHz at lower cost than ANY alternative to run on a simple unreg PS with at least 2dB headroom.

Watch this space.

[ilimzn]

amplifierguru, I agree - I've been using several pairs of IRFPs and they are nearly unbeatable re value for money.
The interesting thing about them is that they are not merely repackaged dies from something like 540, 640 (which I would not use for power amps unless I was making a workhorse so price was an issue and not best performance) etc, but truly independent designs. I was surprised that IR does quite good modeling of these for the linear region, sim results tally with measurements - I consistently get at least 2x lower distortion with IRFPs compared to IRFs in the same circuit (bias has been readjusted of course for minimum distortion on both).
But then I don't pair the IRFPs IR would recomend - i look for as close as possible capacitance matching and transfer characteristics.