a wonderfull sziklai

[catalin]

Originally Posted by dadod

Your FFT shows nothing below 76dB, just short input capacitance for that simulation and you will get it below that.

Slew Rate measurement.

Dadod ,if thd shows :

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 2.859e+00 1.000e+00 -2.49° 0.00°
2 4.000e+04 8.816e-05 3.083e-05 -177.94° -175.46°
3 6.000e+04 2.915e-05 1.019e-05 -85.93° -83.44°
4 8.000e+04 2.117e-05 7.403e-06 -161.72° -159.24°
5 1.000e+05 2.111e-06 7.384e-07 -31.80° -29.31°
6 1.200e+05 1.623e-05 5.676e-06 -160.35° -157.86°
7 1.400e+05 2.498e-05 8.737e-06 77.81° 80.29°
8 1.600e+05 1.248e-05 4.364e-06 -146.11° -143.62°
9 1.800e+05 4.474e-05 1.565e-05 79.16° 81.65°
10 2.000e+05 1.197e-05 4.187e-06 -140.47° -137.99°
Total Harmonic Distortion: 0.003873%

Then it doesn't matter if you have Harmonic 2 greater than Harmonic 3 or H2>H3 because you can't hear nothing under a value of 1000 greater ! For ex 3% !Try on your amp by lowering the vas load with a simple resistor and "hear" the thd .

Also by removing input cap you don't go under 73 db .
If you want to go deeper you need to change the window function to blackman or what do you need :


But remember ,at this value of thd it does not matter how the harmonics are shown,They are 1000 times lower that humans can hear ...
It matters more the phase deviation in closed loop which can be 20 degrees for ex .

The .ac and sin analysis .Like I said please be sure that the idle current is 15-20mA maximum for 2sc5200/1943 .Together with 2sc5171 we will have maximum 30 mA in 0.22 ohms

And again for dadod :

[dadod]

Here is what I've got with your simulation. I changed drivers and output transistors as I do not have the models for the ones you use. How do you tame output triple with no compensation whats ever? When i put 220pF between base and collector of the predrivers oscilation stops.
dado

[dadod]

Sorry I did not check the bias current after models change and it was more then 2 Amps, but how the models could be so different. By the way I used Cordell's model.
dado

[catalin]

Hello Dadod ,It seems that one of your model is not good.In my sim and more important than this,in real life on scope the amps is very stable .To design a good sziklai output you mast have bjt power+drivers with minimum cob .
I will post models which I have ,also from Cordell site .

[Jay]

Have you built the P3A with top quality transistors? You have to have a benchmark for anything you want to design

[catalin]

It was my first or second diy ..
One of the good signs of better sound is the value of the miller cap.
lower the miller cap ,better the sound..

[Jay]

Good to hear that. Did you compare side by side? I don't like the idea of using cascode and current mirror simply to achieve lower THD. I intended to build one of your amp, the one without cascode (V4.4), but first I simulated it and I found that P3A is far superior on paper so I cancelled the project.

[catalin]

Jay ,I don't remember to have a schematic with cascode ...
Always pay attention to the simulation ,idle currents and models .

[dadod]

This is what I've got when the bias was set to 22mA.
I think that your BD139 model is not good, I used Cordell's model and had to adjust the bias.
dado

[dadod]

And this is for 20kHz.