The Holy Grail Follower Output Stage

[lhquam]

The exercise is the overall amplifier performance. But here are some measurements of the output stage without feedback.

I do not have a distortion-free source for +/-20V to drive the OS other than the opamp.
Here are some spectra with no global feedback, the feedback loop is only around the opamp with the same closed-loop gain.
Spectra 1 and 2 are for the amp output 1kHz into 8R at 1W and 16W.
Spectra 3 and 4 are for the opamp output under the same conditions.

The spectra without global feedback are essentially identical to these shown with 12dB global feedback but multiplied be a factor of 4 (12dB). The opamp output has a minor contribution to the spectra.

 

[EUVL]

So we can deduce that the OPS itself has H2 of about -75 dB, and H3 of -100dB.
Is that correct ?

Not too different from those measurements of twitchie with the standard M2OPS with the same devices.
(Post #31)


Patrick

 

[generg]

And C2 is really 15000uF = 15mF?

awful big!

 

[lhquam]

C2 is 1.5mF (1500uF). And might be larger than necessary.

 

[gary s]

Of note is the fact that the output devices are not available, The FQA28N15 is end of life and Mouser will get last shipment in March 2024. The FQA36P15 is nil stock and Mouser does not show delivery till July 2025!

Have you thought about current availability of the outputs - what else could be used?

 

[EUVL]

Find someone else to share, or keep lifetime spares :
https://diyaudiostore.com/collections/active-devices/products/f4-ba3-transistor-kit

You will NOT get any Harris PMOS anymore, as someone cleaned off the entire stock of IRFP9140/9240 at Rochester.
So buy when you can.


Patrick

 

[TungstenAudio]

There are a couple INXS parts still in production that have been mentioned as alternatives. Finding good N & P channel pairs is tricky these days.

The IXTQ26P20P is another P channel part to consider. It has a lower input gate capacitance while keeping the low reverse transfer capacitance of the IXTQ36P15P.

 

[lhquam]

So we can deduce that the OPS itself has H2 of about -75 dB, and H3 of -100dB.
Is that correct ?

Not too different from those measurements of twitchie with the standard M2OPS with the same devices.
(Post #31)


Patrick

For 1 Watt 1kHz into 8R
H2 is about 76dB below H1 => 0.016%
H3 is 26dB below H2 => 0.0008%

These numbers are somewhat better than those in post #31, but H2/H3 is about the same.

Even at full output of 25W with 12dB global feedback, H2 is only about 0.012%. Can you hear that?

 

[lhquam]

Of note is the fact that the output devices are not available, The FQA28N15 is end of life and Mouser will get last shipment in March 2024. The FQA36P15 is nil stock and Mouser does not show delivery till July 2025!

Have you thought about current availability of the outputs - what else could be used?

That is interesting and unfortunate. I recall checking a week or so ago and they were still in stock.

Maybe someone is cleaning out the inventory of all the FETs useful for linear audio power amplifiers to resell at high prices.

 

[needtubes]

I have on order some IXYS IXTQ36N30P and IXTQ36P15P FETs that I will also try.

I have used these in a similar application with good results.

 

[yoaudio]

Opa445, IXFK220N20X3 and IXTK120P20T with voltage over 30V would be interesting.

 

[lhquam]

Why not IXFN140N30P and !XTN40P50P?

 

[EUVL]

I prefer FQA10N80 and FQA9P25.
The former is available at ~ 2USD a piece at Rochester or DK Market Place.
The latter is still available, though end of life.
So perfect case for a self organised GB.
Distortion can be seen at post #31.


Patrick

 

[yoaudio]

The IXFK220N20X3 and IXTK120P20T are $22.27 less per channel.

 

[TungstenAudio]

I see a potential commonality between the output devices in this and the Pass F5m thread:
https://www.diyaudio.com/community/threads/pass-f5m.406472/

Both output stages need N and P channel pairs that are fairly well matched. I have also noticed that the F5m outputs are biased at higher current, about 1.4A. I suspect that might be beneficial to the stage in question here as well.

 

[Zen Mod]

everything is same, except .....

-biasing, and
-common source vs. common drain, and
-Square Law

 

[lhquam]

The Bias Circuit:

The bias circuit has a three resistor network connecting the two optocoupler transistor outputs as shown in the 2nd image of post #1.

Here are some measurement of the bias circuit behavior at different power output levels for some different resistor values for the resistors R11 and R12. In all cases R10=4.7

R11=R12	0W	1W	4W	9W	16W	25W
Infinity	1.30A	1.40A	1.55A	1.70A	1.85A	2.01A
18K	1.29A	1.28A	1.27A	1.34A	1.44A	1.54A
16K	1.30A	1.29A	1.27A	1.31A	1.38A	1.47A
12K	1.30A	1,28A	1.24A	1.18A	1.18A	1.20A

The R11=R12=infinity case is similar to the XA25 bias circuit. With R11=R12 in the range between 12K and 15K the bias current is nearly constant with changes to the output power level, as predicted by the simulations.

 

[Zen Mod]

you didn't say how stable is Iq with change of rails ( mains fluctuation, +/-10%)

 

[Zen Mod]

hypothetically - considering that you're using switcher PSU, sort of stable - disregarding mains fluctuation

 

[lhquam]

I did run simulations of bias vs. rail voltage. Iq/Vrail was about 22mA/V. Not great.

Since the MeanWell LRS-150F SMPS has adjustable output voltage, I will attempt to make the measurement. The adjustment pot of one of the supplies is difficult to access when everything is together and powered up.