A few observations w.r.t. a discrete output-driver

[clintrubber]

I've moved my ramblings w.r.t. the output-stage of the Altec 1589B here, might be better suited and unclutter the thread in which I originally added it at the same time
( http://www.groupdiy.com/index.php?topic=23570 )

So FWIW let's have a look at the 1589B output-driver. Nothing special, but a few observations/things w.r.t. the circuit came up, maybe sombody has some ideas or cares to comment?

Here's the output-amp that drives the output-transformer:

(BTW/FWIW, complete circuit is here: www.exemplaryrecords.com/altec1589bmanual.pdf )

#1
Closed-loop gain looks to be determined by R120 (33k) & R118 (100) resulting in a theoretical closed-loop gain of some 50dB (without output-TX).
But instead of some dominant pole to define the hi-freq rolloff there are R117 & C106 to lift it even further (lifting theor. closed-loop gain by a few dB beyond say 16kHz). Most likely to compensate for hi-freq losses elsewhere, but without any additional cap to bring things down again it looks a bit dangerous... time to do a freq sweep when the unit arrives. Maybe in reality there's simply a small cap across R120.

#2
NPN Q103 acting as a diff-amp on its own (in front of the 'VAS' Q106) ? (Voltage Amplification Stage in D.Self-speak)
Never seen a topology like that there before, or I must be looking it at it from the wrong side...
If Q103-E was connected to Q106-C it might have been a bit more familiar (resulting in a CFP/Sziklai) but would still be a weird VAS, so not sure Q103 would actually help things then.

Who knows it's not really connected like that, but since I haven't received mine yet I can't check yet.

#3
CR106 might provide some form of safety, but haven't given it much more thought.
Or maybe it's for some nicer startup-sequence, quickly charging C108 - seems more likely...


Thanks / bye,

Peter

 

[Samuel Groner]

1: At this noise gain there is little need for additional compensation beside the collector capacity of Q106. C106/R117 increase noise gain (if suitably chosen) and hence enhance stability rather than decrease it.

2: Never seen such an arrangement before either. Looks like a common-base stage. What's sure is that the collector currents of the input pair aren't especially well balanced due to Vbe of Q103.

Samuel

 

[bcarso]

That arrangement with Q103 is a lousy transition stage, essentially using the transistor as a diff amp (Note that it operates at about [edit] 16uA based on the 47k and base-emitter load of Q106). I say this having used it in ancient days myself, before I knew better. For one thing it is easy to induce Q103 beta-degrading reverse BE breakdown in overload; for another, as Samuel says it's not very well-balanced. A first-order adjustment for input pair current balance could be made by making R113 a bit smaller, but there's a large tempco. A reverse-protection diode could be installed across BE of Q103.

I doubt the amp will tolerate a lead cap of any significance across R120. As Samuel says the R117-C106 network is there to raise noise gain and steer away from the rocks of instability.

Overall, not a good design. Poor control over open-loop gain/phase. A lot of open-loop nonlinearity. It cries out for a better transition stage and something to stabilize transconductance of Q106. A little emitter degen in the input pair might be a good idea too.

But hey, nobody's perfect. We do know a lot more about making amplifiers now.

 

[clintrubber]

Thanks both for responding.

I don't see the reason for the weird Q103 either; wouldn't just a five-BJT suffice when we're talking simple DOA's anyway?
Maybe they couldn't get away with a simpler five-BJT DOA-circuit because of the high closed-loop gain (~50dB) they want to get with this output-driver ?
As drawn though, this other 'primitive' five-BJT DOA seems to do fairly high closed-loop gains as well (don't know if additional components are used there to bring the resulting gain down): http://www.avensonaudio.com/tech/Quadeight/mm100schematicsm1.jpg

It will be interesting to check for any benefit from that Q103: as I understand it now so far, it increases open-loop gain so given a certain closed-loop gain it has more gain to spare - but at the same time it's adding that additional gain in an 'unhealty' fashion...


Looking at the total 1589B-preamp (which is say a 1588B-plugin preamp and the above
output-driver circuit) it might be 'better' to get more gain from the first section
and modify the above output-driver for less noise-gain and skip Q103 & R115.
The 1588B consists of a mic-TX and a 3-BJT amplifier configured for 33dB of gain in total,
but as reported it can do some 10..20dB more, by which amount the gain of the modified output-driver
could be lowered.

That's just an idea though based on looking at this thing from the 'tech-side'; it may
partly remove the specific appeal this thing has for certain applications.

So the plan is to leave things as is and simply add two other output-drivers:
the 1589B is a two-in one-out mixer, so adding even a single 5532 can already make this
unit configurable as a dual channel stereo-preamp and it'll keep the 'flawed' output-circuit intact
for 'original use' as well.

(Need to prevent crosstalk between the two channels at the summing point)


Thanks / regards,

Peter

 

[bcarso]

It does have the "feel" of a highly bench-tweaked design (after all, I assume done when simulators were hard to come by and harder to use) so I probably wouldn't monkey with it too much. The protect diode for Q103 might be worthwhile if you think you're going to overload it often.

One thing it does do is nearly rail-to-rail (or rail-to-ground) output swing. It will also saturate Q106 and have a little delay recovering, which could produce some strangeness.

 

[clintrubber]

[quote author="bcarso"]It does have the "feel" of a highly bench-tweaked design (after all, I assume done when simulators were hard to come by and harder to use) so I probably wouldn't monkey with it too much. The protect diode for Q103 might be worthwhile if you think you're going to overload it often.[/quote]

Thanks, adding that protection-diode sounds like a good suggestion. Won't do other changes to this section apart from some caps and maybe later on some MF-resistors.

Regards,

Peter

 

[clintrubber]

[quote author="clintrubber"]I got the 1589B-unit in today, seems 8 years younger than the schematic as in the pdf ('81 vs '73), so I will check if/how the circuit is changed.[/quote]
FWIW, the later version that I got in has a 470pF cap across the B-E junction of Q103.

 

[PRR]

IIRC(?): Phase Linear used that general plan, with the wacky Q103 deal, in their early amps.

Probably comes from instrumentation amp tinkerers. Unlike most other simple schemes, it keeps the Q101 Q102 currents very nearly equal without hand-trimming or many extra parts. This reduces DC offset (who cares here?) and near-nulls 2nd harmonic nonlinearity. Of course it adds its own nonlinearity, but maybe not too much.

 

[clintrubber]

[quote author="PRR"]Probably comes from instrumentation amp tinkerers. Unlike most other simple schemes, it keeps the Q101 Q102 currents very nearly equal without hand-trimming or many extra parts.[/quote]
Interesting to hear that.
Could understand it if it was because of forcing about equal Vce for both diff-pair BJTs (apart from the 1 Vbe difference that is), but I'm not sure it's indeed because of that mechanism.

This reduces DC offset (who cares here?) and near-nulls 2nd harmonic nonlinearity. Of course it adds its own nonlinearity, but maybe not too much.

Hmm, the DC-offset is indeed not relevant here, but that nulling of 2nd we rather hadn't either perhaps.... :?

Regards,

Peter

 

[Guest]

[quote author="PRR"]IIRC(?): Phase Linear used that general plan, with the wacky Q103 deal, in their early amps.

Probably comes from instrumentation amp tinkerers. Unlike most other simple schemes, it keeps the Q101 Q102 currents very nearly equal without hand-trimming or many extra parts. This reduces DC offset (who cares here?) and near-nulls 2nd harmonic nonlinearity. Of course it adds its own nonlinearity, but maybe not too much.[/quote]

Strictly speaking, opamps come to audio from instrumentations (better would stay there), including cascoded diffstages like drawn on the schemo.

Just a fashion, nothing more...

 

[PRR]

> that nulling of 2nd we rather hadn't either perhaps....

If it bothers you, change the 68K and 68K to say 50K and 75K.

However, the BJT's nonlinearity is so steep, you probably want to just minimize all distortion. And for the diff-pair, the null in all distortions at equal-current is pretty deep.