Something puzzles me about this buffer ckt...

[magicchord]

This schematic shows an IC-with-buffer circuit as used in a number of places as an output in my old console.

The circuit looks simple and straightforward and does a good job driving a 150:600 stepup transformer (represented in the schemo by the 150ohm load R).

What I wanna know is - what purpose does the 27pF capacitor C1 serve in the circuit? It seems to me it would hurt rather than help by making the IC drive a more capacitive load. But it must do something good or they wouldn't've put it in. What's goin' on here?

Have a good day

 

[bcarso]

It's a bit odd. But emitter followers usually exhibit a bit of a negative input impedance at high frequencies, so it may have been determined empirically that it quelled ringing or oscillation. It is small enough that by itself it wouldn't get the 356 in trouble, at least.

 

[PRR]

> it must do something good or they wouldn't've put it in.

Don't bet on it.

Some designers are minimalists. Muntz would study the prototype TV sets his people built, then reach in with wire-cutters (live!) and cut a part out. If the TV kept working (it often did), the designer was scolded for wastefulness. But you find other "designs" with lots of excess parts. Commercial designs tend to have less of this, but sometimes you find 100pFd caps sprinkled like salt.

> making the IC drive a more capacitive load.

Yeah, but 30pFd on a LF356 is small. I'd figure 30pFd stray capacitance at any node (true, I'm more an Octal-guy than a DIP-head). The BiFET output is around 300Ω, 300Ω and 30pFd is almost 2Mhz. As bcarso says, it probably slumps the 356 phase just enough to not ring from the added phase shift in the transistors. I'd grab the wire-cutters and try without it. Unless you "listen" on a 10MHz oscilloscope, it may not be needed.

For that matter: what the heck does R5 330Ω add to the plot? OK, it "fixes" the fact they use dissimilar devices at Q6 Q5 and at Q3 Q4. Note that if R5 is cut-out, Q6=Q4 and Q5=Q3, and R1 R2 are made zero, both stages run at the same current, fixed by resistors R3 R4. That's not very stable, but if R1 R2 drop 20mV to 60mV, the output stage runs at 1/2 to 1/10th of driver current and is very stable. R3 R4 pass 4mA, so you can set output stage current to 2mA to 0.4mA with R1 R2 of 10Ω to 150Ω. 5Ω will work well too. Now you not only save a resistor, you save the trouble and cost of stocking twice as many transistor types and getting them stuffed in the right holes.

I must note that +/-20V is probably illegal for LF356, though I would expect it to work for days or decades. (Yeah, I've put 500V on 60V-rated parts in simulation; it mostly works, though my 1N1002 model breaks-down at 100.1V exactly....)

 

[magicchord]

I believe the 356s in this 1981-vintage board (ADM) were a certain suffix that could take up to 22v.
I have experimentally tried such chips as OPA134 and OPA627 in this circuit and haven't blown any up yet :grin:
The 627 sounds really good and measures well. Expensive though. I have 'em in a couple of channel outputs.
I've also tried the NE5534 (the mfr must have tried them too since there's holes in the PC board for the compensation cap betwixt pins 5 & 8), the OPA604 also, and the LM318 works very well too.
They all surpass the 356 which I think is a poor choice.

I've tried taking out the 330ohm resistor and, though the idle current through the output Qs increases, the distortion into a transformer load increases too. So back in it went.

 

[bcarso]

Sims show the 27p about right to reduce what is otherwise about 2 dB of peaking in the response of the buffer alone when driven from 1k. One might think that the op amp's output Z would be much lower than this, but then recall that the output Z will be rising at high frequencies. In this case the peaking is at about 6MHz without the cap.

The impact of the 330 ohm is indeed puzzling. Again sims (taken with a grain of salt) show the distortion into the given 150 ohm resistive load to be much lower without the resistor there. The 330 ohm also steals away a lot of the current for the driver Q's (337uA with, 4.1mA without) and makes the whole buffer about an octave slower.

The distortion in the sim at 1kHz, 5V peak in is 0.55% with the 330, 0.12% without, and it is mostly third.

Typically, one sees resistors like that in a different topology, like Locanthi's "T-Circuit", with cascades of emitter followers of the same polarity on upper and lower halves, and a big bias network at the input between to make up all the Vbe drops.

 

[magicchord]

[quote author="bcarso"]...The impact of the 330 ohm is indeed puzzling. Again sims (taken with a grain of salt) show the distortion into the given 150 ohm resistive load to be much lower without the resistor there. The 330 ohm also steals away a lot of the current for the driver Q's (337uA with, 4.1mA without) and makes the whole buffer about an octave slower...[/quote]

Yes, what you say is true. I'm using as an analyzer a PC application called AudioTester which uses my soundcard as a signal source and as a spectrum analyzer or distortion plotter. Perhaps the transformer has something to do with the results I'm getting. Or my measurement method is flawed.

 

[bcarso]

It would be interesting if third order distortion in the trafo was being partially compensated by the buffer distortion. Can you measure the output independently with a resistive load?

 

[magicchord]

Hm. Whenever I get time I may do more detailed testing.
Keep in mind too that my tests are with the buffer in the middle of an opamp feedback loop - whatever difference that may make.
In any case the distortion is pretty low.