Opamp inputs

[Rochey]

Guys,

I'm not sure if this is written elsewhere - thought i'd ask here.

Does anybody have a link to a website that explains the difference between opamp inputs (i.e. BJT input, JFET input etc)?

I'm interested in knowning how this can affect the sound of a device.

Cheers

R

 

[Samuel Groner]

I'm not aware of a home-page right now, but I'm sure Jung writes some informative stuff in his two opamp books.

As a first starter let me say:
* bipolar tends to be lower noise than JFET for typical audio impedances
* bipolar tends to have lower distortion than JFET
* JFET has lower/"zero" input bias and offset current. This allows usually for simpler surrounding circuitry (i.e. no coupling caps) which may have an effect on performance
* JFET is often less prone to RFI demodulation
* JFET tends to have higher slew rate (though this is an effect which results in the typically higher distortion)

These are very rough and general statements and I would like to add the engineering wisdom: it all depends.

Samuel

 

[Rochey]

thanks Samuel -- if anybody else has some pearls of wisdom this this to share, I'd really appreciate it.

(gets the printer fired up...)

Ciao 4 now

R

 

[daArry]

Hey Rochey,

Here's some links, dunno if ur answers lies within em - but ya can never know!

http://www.analog.com/library/analogDialogue/Anniversary/6.html
http://cygnus.ipal.org/mirror/www.passlabs.com/diyopamp.htm
http://www.dself.dsl.pipex.com/ampins/webbop/opamp.htm

:thumb:

 

[tk@halmi]

* bipolar tends to be lower noise than JFET for typical audio impedances
At 600 ohms, yes the bipolars will have lower total noise. At 10K many FET inputs (OPA134, OPA604) will have lower input noise than BJTs .

* JFET tends to have higher slew rate (though this is an effect which results in the typically higher distortion)
Please explain this one to me. I have not learned how higher slew rate causes distortion.

There is one aspect of JFET inputs neglected by most hobby audio designers: if the FET input is p-channel type and it is intended to be used in a non-inverting configuration mismatching the input impedances of the inputs (inverting and non-inverting) will cause capacitance modulation (capacitance varying with input voltage). This means increased distiortion. This is one of the reasons why many JFET opamps will sound poorly when people swap IC opamps in/out of existing circuits. In most cases the circuit has to be designed to match the device used.

 

[bcarso]

tk: Amen (w.r.t. the capacitance modulation distortion).

I gassed about this in an earlier thread but I'm too lazy to look for it now.

 

[Gus]

There were good articles written by W. Jung IIRC about the differences the input types have on the sound and even some of the math behind it. I think it was in Audio about 1980.

And as posted by others sometimes the details count alot.

 

[daArry]

http://home.comcast.net/~walt-jung/wsb/html/view.cgi-showresources.html-TopRes-Walt-27s-20PDFs.html

 

[Samuel Groner]

Please explain this one to me. I have not learned how higher slew rate causes distortion.

I do not say that high slew rate causes distortion. I say that there is typically a design compromise between high slew rate and low distortion. I'm sure you have Selfs and Hoods books at hand, they explain this better than I could.

Samuel

 

[clintrubber]

Wasn't the trick to fix this simply having the same impedance at both inputs? In inverting configuration with 10k and 10k resistors you simply just shunt the positive input to ground with 10k?

That last 10kOhm should be a strawberry pancake then, sorry, 5kOhm I mean: R_in // R_feedback

- wait, with inverting you don't have the problem altogether.

Of the link gven above there's some about this in the #4-pdf:
http://home.comcast.net/~walt-jung/wsb/PDFs/WTnT_Op_Amp_Audio_4.pdf

 

[bcarso]

Thanks for that ref. I was not ready to slog through the entire list of pdf's to find it ;-)

I would also point out that even if the input capacitance/input voltage relationship were ~linear you would still have a distortion mechanism. * Also, assuming the two inputs behave the same is a simplifying assumption that will get you started, but what goes on at each FET drain internally will usually be different and with that difference the cap modulation differs as well.

One of the appeals of tubes is that the grid capacitance doesn't vary much with signal swing. Does anyone know of any measurements or for that matter theoretical analyses that say by how much it does change?



*this reminds me of one of my pet peeves--calling transistor beta "linear" when what is meant is "constant".

 

[CJ]

Note that a bjt opamp front end needs some bias current. How much depends on the opamp.
The fet front end needs only pico or nano amps bias current.



I think this is right. :?:

 

[Samuel Groner]

I think this is right.

I think so too. In fact, I think it is so important that it is well worth saying twice! :wink:

At 10K many FET inputs (OPA134, OPA604) will have lower input noise than BJTs.

I quickly did the math for a 5534 and an OPA604; at 10k the later is still a tad behind.

Samuel

 

[tk@halmi]

I quickly did the math for a 5534 and an OPA604; at 10k the later is still a tad behind.

Did you take current noise into account or just voltage noise? The BJT input will contribute current noise as well.

 

[clintrubber]

Note that a bjt opamp front end needs some bias current. How much depends on the opamp.
The fet front end needs only pico or nano amps bias current.

As I see it, the good news is that the circuit outside of the opamp can stay the same. Like for the situation at hand (with the 10k+10k feedback network), inserting 5k will give offset-compensation for a BJT-input and THD-compensation for a JFET-input. In both cases the required value is 5k :grin:

The latter is of course negelecting the 'complete' impedance, as could be seen in a certain AD-datasheet (which mimics the total network, including all kinds of caps, but i forgot which datasheet it was - we did have a thread about it though, but someone - again, forgot who- really went down the road to mimic a very elaborate network).

regards,

Peter

 

[bcarso]

As far as the voltage-dependent capacitance mechanism is concerned, there is I believe one old Burr-Brown dielectrically isolated part that bootstraps the drains to the sources so that this mechanism is thwarted. And if you are rolling your own discrete amp this is available as an option as well.

 

[PRR]

> One of the appeals of tubes is that the grid capacitance doesn't vary much with signal swing. Does anyone know of any measurements or for that matter theoretical analyses that say by how much it does change?

Some data in RDH 4th, chapter 2, around page 55 (PDF page 43).

One tidbit: for many purposes, the change of input capacitance with bias can be negated with a cathode resistor that reduces gain to 0.6-0.75 of the fully-bypassed gain. That can't be perfect, but the change is small to begin with so computing the last hair may just be fun-with-numbers.

 

[bcarso]

Excellent! Thanks PRR. This changes ( a bit anyway) some wacky circuit ideas I have been developing. Good old RDH.

 

[Samuel Groner]

Did you take current noise into account or just voltage noise?

Hey, my Excel sheet does it all! (I'll share it as soon as I documented it better.) It's based on the math presented by Jung in his audio opamp book.

Samuel

 

[gar381]

related

http://www.diyaudio.com/forums/showthread/t-23030.html

GARY