bipolar or fet buffer?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

featherpillow

Well-known member
Joined
May 14, 2005
Messages
214
Location
USA
For a simple transistor based output buffer, which do you prefer, and why? A bipolar transistor based buffer, or a FET transistor based buffer?
 
Thanks for the response!

The rule of thumb I've heard is FET for input, bipolar for output because of its ability to drive cables better. Do you think this is true?
 
This is somewhat true, if a very high input impedance is required. If a low input impedance is required, bipolars perform better in most instances. For output circuits, both are used, but I tend to believe bipolars are somewhat better at driving low impedance output loads.
 
Ummm---sort of (replying to featherpillow). The DMOS technology has gotten so good for some things that it's not as clear cut as it used to be, but at these relatively low power levels with care bipolars work well.

You have to make sure you don't saturate, and mind the thermal effects closely. DMOS doesn't have storage time per se, but the input capacitance to transconductance ratio is not as good (low) as bipolars, usually. And with linear operation of DMOS you also have substantial temperature coefficients for modern geometries unless you operate at impractically high quiescent currents. As well DMOS has a very wide distribution of threshold voltages, much worse than bipolars, so they are tricky to use in production as linear amps.

Power JFETs essentially don't exist, and if they did would have many of the problems of DMOS.
 
Also DMOS still has a lot of 1/f noise so generally is not a good idea for inputs, although the improvement over the years has been amazing.

Good big JFET's like the 2SK170 have gotten competitive with bipolars for low noise front ends, but the input capacitances are larger which can be a problem.
 
That's a great circuit indeed - I didn't dare to post it yet another time so nice you brought it up ! :wink:
It's alike (/about identical/related) to the circuit Dan K. was so kind to provide here before.

Fabio's '3.1.2 & '1.2.7.2-boards have Hi-Z buffers as well (FET+NPN), but different topology - I'm curious to eventual noticable difference (in sound).


Nice question this thread BTW - ever since I saw Boss FX-pedal schematics I've been wondering why they used a FET and then again a NPN for another pedal.

Regards,

Peter
 
[quote author="featherpillow"]For a simple transistor based output buffer, which do you prefer, and why? A bipolar transistor based buffer, or a FET transistor based buffer?[/quote]

IMHO, the question is too broad to answer meaningfully. Are there requirements for this buffer?
- What impedance will it see on the output?
- What impedance will it see on the input?
- What sort of power and voltage rails are available?
- How much is it going to have to swing peak-to-peak into the following stage? ...and so on...

It is best if you state the exact nature of the application. Otherwise you will get a lot of "gut-feel" opinions, and they may not be the best solution to your problem.

Cheers,
Tamas
 
Thanks for the circuit link...I hadn't thought of a FET/PNP combo as a buffer. I assume the theory is to take advantage of the strengths of both components in order to compensate for their individual weaknesses, correct?

Is input impedance set by the combination of R1 and R2 in parallel (about 1M)? How is output impedance set--by R4 and R5? The spec sheet says 200 ohms O/P impedance, I can't see how this value is set by R4 and R5, which appear to me to be build-out resistors.

For the sake of argument, Tamas, let's say we want a buffer that has about 1M input, and about 60-100 ohms out. Supply rails are a typical +/-15vdc.

From what I gather, Walt Jung's take on buffers is that they can make or break a design. Any opinions?
 
Tamas is of course correct, and now we know your requirements.

The output Z of the FET/PNP combo per se is quite low due to lots of feedback. That's why the 100 ohm R sets the Z out along with the other 100 ohm R.

It's not the highest-performance stage that can be imagined but with moderately small signals it will work fine. There are higher performance FETs but the 4393 is a good old standard medium-capacitance part.
 
Thanks for the reply!

So the feedback network is of crucial importance here?

What are the functions of D1, D2 and C5?
 
"So the feedback network is of crucial importance here?

What are the functions of D1, D2 and C5?"

The feedback network in this case is a short circuit---the collector of Q2 tied directly to the source of Q1, making the Q1-gate-to-"buffer out" gain close to unity, and drastically reducing input C, output Z, and distortion. When two bipolars are used in this config. it is known as a Sziklai follower (for trivia buffs ;-). Thanks to Peter Carroll Dunn's great book Gateways into Electronics for that piece of info.

D1 regulates the phantom power to about 12V for the buffer, and also decouples (with C5) the source of the power from the output drive. D2 clamps the buffer output to the D1/C5 potential, so that line never gets more than 0.7V above the buffer V+ on turnon or under other transient conditions.
 
Thanks for the answer bcarso!

So, if I were to use this circuit as an output buffer for something other than its intended purpose, say a mic preamp or some other application, I would

a) Omit D1, D2, and C5.
b) Drop the input impedance of the stage to something smaller, like 10k, reduce increase input coupling cap size. If I reduced the value of the input impedance, would I then want to reduce the value of all other resistors by roughly the same ratio except for the build out resistors? Although perhaps not...would I be reducing the circuit's 10v swing capability by doing this?

Is this at all correct, or am I off here?

What about the two 6.8k's here--what function are these serving?

I just noticed, too that this circuit seems to function as a phase splitter as well as a buffer...
 
You presumably want voltage gain for your preamp, so this wouldn't give you that. The input impedance set by the bias R's is not an issue---you want the mic impedance to be lower that the preamp input impedance when dealing with audio frequencies. (note added later: there may be subtle loading effects with certain mics that will dictate the input/loading R, but they are rarely all that low compared to the mic Z).

The circuit does not act as a phase splitter, although it could if you placed a resistive load of 1.2k in series with what is now the 12V supply and fed it from an appropriately higher voltage; the new inverted phase output will now be taken from that junction, and the noninverted from what is now the only active output. And you do pick up 6dB of gain this way feeding the balanced input.

And the FET is not the best for a low Z mic from a noise standpoint.

You could get a bit more gain out the the two transistor combo but not a whole bunch before distortion etc. will suffer.
 
You presumably want voltage gain for your preamp, so this wouldn't give you that.

Actually, I wouldn't be looking for voltage gain out of this circuit. I would only want something to act as an output buffer, but with different input and output impedances. I prototyped a variation of it last night, and I have the schematic on my computer at home, which I will be happy to post once I get home from work today.

Basically, I dropped R1 and C1, R6 and R7 (the two 6.8k resistors), and D1, D2, and C5. I'm feeding it from a +15v supply instead of +12. I I changed R2 to 2M, dropped R3 from 4.7M to 2.2M, and dropped the value of the two build-out resistors (R4 and R5) to 60 ohms. I tested it by driving into a 600:600 Edcor transformer. Seems to work OK thus far, tested it with a 1k square wave and sine wave both, about 10v pk-pk as the original schem suggested.

Any additional observations and suggestions?
 

Latest posts

Back
Top