Old and surprisingly good amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
This 1975 design of a quasi complimentary power amp suffered somewhat
from the poor properties of the 2N3055: very poor ib/ic characteristics,
low ft ( 15 kHz) , high leakage currents.
A big surprise: replacing the 2N3055 with 2SC2922 makes an amp featuring
second harmonic at -60 dB and higher harmonics not measurable at 80% max power ( here 25 Vp in 5 ohms) . Fantastic triode- like sound. Very easy to built.
Temperature bias compensation is dynamic Q6 needs not ( and should not!) be in thermal contact with heatsink.
 

Attachments

  • Equin.jpg
    Equin.jpg
    78.9 KB · Views: 770
What's wrong about R4's location?

I don't like the relative sizes of C6 & C4.
I would experiment with C4 moving towards 2200uF and reducing C6 towards 470nF.

I would also look at modifying the quasi so that the drivers each see the same loadings. Baxandall diode & JLH capacitor.

None of these mods may give any improvement, the designer may well have got the sound just right with the choices he made.
 
Hi

Why do you call the compensation dynamic? To me, it looks like a pretty standard compensation, albeit with a slope correction.
BTW, R4 seems misplaced.

R13 seems to be a special compensation function - perhaps therefore the actually not right estimate for a dynamic compensation ?
I am looking for a basic article (Vbe multiplier) arround this actually not necessary R13 - who know that paper?

in this case also this threads are of interest:
http://www.diyaudio.com/forums/solid-state/165165-edwin-20-watt-elektor-1970-may.html
http://www.diyaudio.com/forums/solid-state/113943-quad-303-triple-cascade.html
http://www.diyaudio.com/forums/solid-state/165130-curious-extra-knob-trans-amp-6.html (post #41 and #52)
http://www.diyaudio.com/forums/soli...opologies-100ma-idle-sound-closest-class.html
 
You are not wrong.
R3 is in series with the emitter, exactly like R8.

R4 does a different job. I don't see it as being wrong located. There is no missing emitter resistor for it to be moved to.
The voltage drop that you mentioned across R4 is only 80mV to 90mV. That just reduces Vce of the lower driver by 80mV to 90mV. I don't see that as an issue.

A bigger issue may be the completely different loads seen by the emitters of both drivers. That's why Shaw, Baxandall & JLH changed the load seen by the lower driver's emitter.
 
The design IS Quad303 triple cascade. The dynamic compensation of junction temperature of the power BJTs is via current feedback with R5, R10.
I have simulated that with the temp statement of LTSPice and it works well
up to 100 ° junction temperature of Q1 Q2 Q3 Q8. The qc of Q1 Q2 goes from 50 mA at 30° to 90 mA at 100° that is excellent. There is no time delay as it is in the case of the Vbe multiplier thermally coupled with heatsink .
The performance of this very simple design is amazing when Q1 Q2 are (as tested) 2N2922 which feature good ib/ic linearity up to ic 6 amps and ft 40 mHz.
The compensation C3 can be reduced to 15 pF and C2 to 10pF thus that distortion at 20 kHz is only second harmonic.
Btw making the triple cascade complimentary does not improve the performance, h2 becomes smaller but odd harmonics appear.
 
You are not wrong.
R3 is in series with the emitter, exactly like R8.

R4 does a different job. I don't see it as being wrong located.

What is R4's job then?

The dynamic compensation of junction temperature of the power BJTs is via current feedback with R5, R10.
I have simulated that with the temp statement of LTSPice and it works well
This is not "compensation", just degeneration, and anyway, in such a "super-CFP" configuration, the Vbe to compensate is that of the drivers, not the output devices, and that's what the Vbe multiplier does.
The advantage with triples is the low power dissipation/variation in the drivers.
 
Hi, WuYit, what is "minor circuit clean-up"?

I see a half CFA plus bootstrap, I never tried this one.
The best output stages I ever heard and built is EF facing CFA (bumping) in floating driver supply. Its sound is stuning me and my friends.
I never heard any expensive triode amps.
 
The design IS Quad303 triple cascade. The dynamic compensation of junction temperature of the power BJTs is via current feedback with R5, R10.
I have simulated that with the temp statement of LTSPice and it works well
up to 100 ° junction temperature of Q1 Q2 Q3 Q8. The qc of Q1 Q2 goes from 50 mA at 30° to 90 mA at 100° that is excellent. There is no time delay as it is in the case of the Vbe multiplier thermally coupled with heatsink .
The performance of this very simple design is amazing when Q1 Q2 are (as tested) 2N2922 which feature good ib/ic linearity up to ic 6 amps and ft 40 mHz.
The compensation C3 can be reduced to 15 pF and C2 to 10pF thus that distortion at 20 kHz is only second harmonic.
Btw making the triple cascade complimentary does not improve the performance, h2 becomes smaller but odd harmonics appear.
What is R4's job then?
This is not "compensation", just degeneration, and anyway, in such a "super-CFP" configuration, the Vbe to compensate is that of the drivers, not the output devices, and that's what the Vbe multiplier does.
The advantage with triples is the low power dissipation/variation in the drivers.
In an old article from ELEKTOR (edition December 1972 EQUA project) this is detailled explain, also various differences (disadvantages) to the Quad 303 triplet buffer.
http://www.hifi-forum.de/viewthread-71-8390.html
also this patent could be of interest in this case
http://www.freepatentsonline.com/5726602.pdf

BTW - until this day I haven't find an effective solution to remove the tendency to oscillate in that PP buffer - even not with help of CAD. This is cause the differences between the upper and lower half and I am not able to find a right compensation network.
 
Last edited:
Another significant advantage of this design is low reactve distortion. The thd is hardly ever measured on complex / reactive loads although every speaker is a complex load.
The good performance is possibly due to current feedback in the triple and to the input BJT. On the other hand CMRR is poor . There are BJTs with even better ib/ic linearity than 2SC2922 and BD139/140
 
Hi,
there are problems with the triple. When composed of class AB biased bipolar transistors, it´s prone to oscillation and to produce unpleasant crossover distortion. The complementary feedback pair is actually great, but it works best in class A (with feedback through just two devices!)
On the plus side: single-ended input (nice "triode-like" sound, but higher sensitivity to ripple) and current feedback (wide bandwidth, good stability and low reactive distortion).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.