a wonderfull sziklai

[ostripper]

I actually built the sziklai OP stage and a little problem..
I will show it on a sim , but it does the same on the real thing.


[note "wiggle" on SW]

Self and others warned of this with the sziklai, as another
FB loop has its advantages and drawbacks (oscillation).

The fix was simple enough , a local shunt C (220p) on
each driver....


[fixed SW at 20k]
This might not be a serious problem as I could not hear any
difference in sound and my zobel network did not get hot..
but 2 caps are cheap and do the job..


[the "fix"]

Anyway, good work,catalin. I fully in tend to offer my amp
in both "flavors"(EF and sziklai) as the latter has a different
sound than the EF.

OS

[catalin]

Hello .I really dont'n know how do you get at these waveforms.
Do you simulate my amp or yours ?On my oscilocope the waveform (practical and theoretical) are perfect !
The success of the sziklay is to put the prefinall stage transistor with little c-b capacitance(10-20pf);the mje are not good .
If you simulate your amp I think that it hasn't phase reserve. put an pole and an zero on the diferential stage.
if you put 220pF there the slew rate will get low.So why do you design an sziklay if you get low slew rate...
Those waveforms where do you get it ??

[ostripper]

By catalin - put an pole and an zero on the diferential stage.

That is a good idea.. R/C at differential, deal with problem
"before the fact"

the mje are not good

We use what is available...
My EF shows no such effect (sim or real) so it must be
the OP.

if you put 220pF there the slew rate will get low.So why do you design an sziklay if you get low slew rate...

I used 220p just to make the point. On both the sim and
real I will "fine tune" 10-47p (or maybe none if diff. trick
works.. )The waveforms are on LT4 ...easier to show than
webcamming my CRO..
OS

[catalin]

At the sziklai stage, at higher frequencies, the output returns at input with 360 degrees.The "road" is by the c-b capacitance of the final bipolar and the cap of the prefinall stage.Then it a must to put in the prefinall stage bipolars with low capacitance like 2sc5171..etc. In that case(with mj) it is normal what is happening.
With those 220p you get the signal who returns at the base of prefinall stage at ground .

[ostripper]

So what you are basically saying is that without a faster
(2sc5171) lower cob driver the benefit of your OP stage
will never be realised..

I will model with the faster devices and see if I still still experience errata..
OS

[catalin]

Exactly !
Even cob for the final stage .... 2sc5200
For the high freq...there is a "road" :from the collector 2sc5200 to the base of the 2sa1943...so it will create oscilaltions if the cob's are big...like at the slower bipolar(mj)

[catalin]

The characteristics :
Thd = 0.001(2) at 1khz with 5 harmonics .
Slew rate measured by scope 70V/uS positive edge , 100 negative.
Bandwidth -3dB: 5Hz -- 340Khz .
Open loop Gain: 120 dB.
Closed loop Gain = 33 .
0 db frequency without input filter >14 Mhz.
Gain bandwidth product: 34e+6 .
Open loop gain at 20 khz : 65dB
Ic0 ! 25-40mA => Vr(output)~ 6-8 mV .
Below i tested this amp with real mark audio analizer with alc889(192 khz 32 bits) and the tda7294 ,Remember that all results are the sum of the [alc889A + one of 2 amp's] and the distortion and noise of alc889 are amplified by the amp.It may be inaccurate but it make diference between those .The load during test was a 4 ohms speakers(2 way).

[catalin]

trise measure with resistence load ~70V/uS

[catalin]

Test tda7294 + alc889a

[ostripper]

Even with the MJL's the harmonics (THD)were the same (H2/5,
no H3(see below). My slew wasn't as impressive , but that was because
of the slow drivers.

So your findings confirm mine as I compared sziklai to
EF (equal H2/3/5) both in spice and CRO (FFT, no fancy analyzer).

OS