Why have a pull-up resistor to bias the input of an op-amp gain stage?

[gaussmarkov]

This question may sound like it has the answer in it.

But I realized that sometimes there is no pull-up resistor to bias the non-inverting input of a divided feedback op amp gain stage.  Take the clipping stage of the tubescreamer, as described by R.G. in his Technology of the Tube Screamer. After the coupling capacitor, there is a 10K resistor to 4.5V.  If you take this resistor out, this clipping stage works almost the same way.

In all of the stompbox circuits that I have looked at, there is a resistor after the coupling cap.  But what purpose does this resistor serve?  It seems like an op-amp biases the noninverting input automatically.  I don't understand why, but that is what I see.

Any help is appreciated, gm

[PerroGrande]

Hi GM,

I've been a little curious myself about the tube-screamer's use of the resistor to VRef instead of to ground.  I've run the circuit in my simulator with the 4K7 resistor in the clipping stage to ground instead of to Vref and the behavior does not change.

In reality, the stage is de-coupled from any DC via the capacitor, so all we have to think about is the path for signal frequencies.  For signal frequencies, there is really little difference between Vref and ground.  All the filtering in the supply is going to make signals see a path to ground -- so I guess in the case of the TS, it doesn't matter.

[gaussmarkov]

hi! thanks for replying.

i think we are talking about different things.  i am referring to the 10K resistor connected to the non-inverting input and Vref and you are talking about the 4k7 resistor hanging off the feedback loop, right? 

peace, gm

[PerroGrande]

Sorry -- you're right -- we're talking about different things.   


I think this article will provide the answers you seek.

http://www.analog.com/library/analogDialogue/archives/35-02/avoiding/

[gaussmarkov]

I think this article will provide the answers you seek.

http://www.analog.com/library/analogDialogue/archives/35-02/avoiding/

fantasitic!  that is a great article.  thanks very much.

[PerroGrande]

Yeah -- I enjoyed that article, too.  I remembered reading it a few months ago when a similar discussion came up.    I knew anecdotally that some performance aspects of operational amplifiers are lost when run on a single supply.  However, I had been out of the field for too long to remember all the nuances.  That article really helped bring it back into focus.

[gaussmarkov]

Thanks again.  By the way, this recent thread TS-808 Circuit Question for the Tube Screamer experts has some discussion of the topic that you have been thinking about.

[gaussmarkov]

I found the following App Note from Texas Instruments that doesn't always bias the input of the op-amp.  Look, for example, at Figure 3 for the noninverting AC-coupled gain stage.

Edit:  oops.  forgot the link:  A Single-Supply Op-Amp Circuit Collection

[gez]

Bipolar input amps require bias current, just as a discrete transistor requires a bias current.  A resistor is required.

As far as FET amps go, then perhaps the leakage of/charge on the coupling cap is enough to set a reference for the + input (assuming there's a previous stage), but I should imagine there'd be quite a DC offset at the output (wouldn't bias at half supply).  With diode clippers, that probably isn't critical.  Having said that, some CMOS amps have such low 'input bias' current, that they probably bias up fine.

Call me old-fashioned, but I just think it bad design to leave off said resistor (if AC coupled).  Input bias is so low on amps these days that many would probably work without the normal bias, but I'd be amazed if the older stuff did - and you never know what people are going to stick into their circuits (especially the salvagers out there)!

[12afael]

coated paper work very well for me

[gaussmarkov]

Bipolar input amps require bias current, just as a discrete transistor requires a bias current.  A resistor is required.

As far as FET amps go, then perhaps the leakage of/charge on the coupling cap is enough to set a reference for the + input (assuming there's a previous stage), but I should imagine there'd be quite a DC offset at the output (wouldn't bias at half supply).  With diode clippers, that probably isn't critical.  Having said that, some CMOS amps have such low 'input bias' current, that they probably bias up fine.

Call me old-fashioned, but I just think it bad design to leave off said resistor (if AC coupled).  Input bias is so low on amps these days that many would probably work without the normal bias, but I'd be amazed if the older stuff did - and you never know what people are going to stick into their circuits (especially the salvagers out there)!

thanks, gez.  when i see odd stuff like this, and then simulate it and it works, it's pretty hard to know what to make of it.

and then i wonder whether it might be a way to make an interesting difference in sound.   i haven't actually put it on a breadboard to see.

cheers, gm

[gez]

and then i wonder whether it might be a way to make an interesting difference in sound.   i haven't actually put it on a breadboard to see.

cheers, gm

Andrew (Tone God) posts schematics now and again without said resistor.  I thought it was an omission at first, but he says it works so it obviously does.

[PerroGrande]

They'll certainly work without the resistor, but I agree with Gez that it is bad practice to leave it out.  At the very least, it allows for a "stiffer" voltage divider without destroying the input impedance of the stage.

[gaussmarkov]

cool.  what does "stiffer" mean?  close to an ideal DC source?  so low resistor values in the voltage divider? and then use a high value for the bias resistor for high gain stage input impedance?  if that's correct then i think i am following you.