Biasing non inverted into inverted opamp

[Max9999]

Dear diystompboxers,

I am trying to do this inverting buffer with 1M input, see picture. Because of hiss considerations I stuck a non inverting buffer in front of it.
Problem is that I don't know how to bias it now.

Is there a coupling cap needed between the opamps?
Do I need to apply Vbias to both the non inverting input of U1 and the non inverting input of U2?

[ElectricDruid]

You don't need to worry about a coupling cap for the second stage, because the output from the first stage is already biased. You do need to tie the +ve op-amp input to Vbias, because that is the reference level for the signal.

HTH

[GibsonGM]

+1 what the Druid said. Welcome to the forum, Max. 

[Vivek]

Electricdruid has a secret schematic of a "Buffer with gain"


That would need only one Opamp instead of 2

[ElectricDruid]

Electricdruid has a secret schematic of a "Buffer with gain"

That would need only one Opamp instead of 2

It's not a secret, but it's not inverting either. So I think it still needs two op-amps.

I like the way Max9999 did it, to be honest - solid non-inv buffer at the front, and then you don't have to worry about the impedance for the inverting stage, so you can choose low resistor values and keep the noise down. There's a temptation to try and do it in one inverting op-amp with a pair of 1M resistors or something, but I expect this way is quieter.

[Vivek]

Would this idea help here ?

[iainpunk]

> buffer
> gain

choose one, only one

welcome to the forum by the way.

cheers

[Max9999]

Cool!

Just to double check Electrid Druid and GM, you mean I re apply vbias to U2 right? Like in this schematic below.

Electric Druid, my spice simulations show a lot less hiss like this compared to using 1M resistors in single inverting configuration.

Vivek, thanks but I am looking for a gain of -1.

[Vivek]

Could you run an LTSPICE noise analysis without specifying the values of the capacitors and without specifying the model of the Opamps ?

Maybe then it is a very theoretical noise figure calculated based only of the resistors,

and maybe the real life Opamps make more noise than the resistors but LTSPICE used ideal noiseless Opamps since you did not enter Opamp model number.

[GibsonGM]

Yes, Max, but place a resistor before U2's "+" input.  10k is pretty standard.  If you want lower noise, also decrease the 1Meg bias R in the first stage.

You could just breadboard the thing and try 1Meg's, and 'something lower' to see what the R is doing re. noise....

[anotherjim]

The buffer input bias reference can be 1M or more without contributing resistor noise since the least noisy resistor in parallel is infinite resistance. An audio designer might use 1k for the inverter since the least noisy series resistance is zero and 1k is an ok load for the buffer opamp and it doesn't need a coupling cap which would have to be around 10uFNP into 1k.

Then there is the opamp type since input bias current and noise depend on those resistors. A Bi-FET likes the 1M input for the buffer but a BJT type can easily be found that's less noisy for the inverter. A dual opamp containing one of each TL071 and a NE5534 would be useful.

[GibsonGM]

Or if your rig is as noisy as mine is, you won't hear any real difference! (LOL)  But the Facts ARE important.

[ElectricDruid]

Just to double check Electrid Druid and GM, you mean I re apply vbias to U2 right? Like in this schematic below.

Yep, exactly!

Electric Druid, my spice simulations show a lot less hiss like this compared to using 1M resistors in single inverting configuration.

Yes, that's what I'd expect to see practically too. Years ago when I realised you could get the same gain as long as the *ratio* of the resistors was the same, I dod some experiments to see if it made any actual difference. What I learned was that (roughly) fo the decade between 10K and 100K it makes no odds. If you go lower by another decade, current starts to be a consideration. If you go up by another decade, noise becomes a consideration.
So I wouldn't build a one-stage inverting op-amp with 1M resistors, I prefer it the way you've done it, with the buffer in front, and then the inversion done with much lower value Rs.

[ElectricDruid]

Yes, Max, but place a resistor before U2's "+" input.  10k is pretty standard.

Why, please? It might help with DC offset, but since we're dealing with audio and the first thing after the op-amp is a cap, I don't see the point. Am I missing something? (It wouldn't be the first time... )

[GibsonGM]

My bad - you'd need it if you flipped U2, tho, LOL     Good eye Tom.

[amptramp]

Would this idea help here ?

In a booster with gain as shown here, if the source impedance is low, the pair of 2.2 meg resistors are in parallel above the R-C turnover set by C4 and R6 which is 0.7 Hz, making the input resistance for noise calculation values to be 1.1 megohms.  If you also parallel the series combination of R4 and the guitar output resistance which is typically a 10K pickup in parallel with a 500K level pot on the guitar, the largest resistance seen looking into the guitar is 255K so 256K (guitar plus R4) in parallel with 1.1 meg (=207669 ohms) is the maximum value of the resistor for noise calculations.  If you leave the guitar control dimed, you have an equivalent noise resistance of slightly less than 10K.

It doesn't matter whether the first stage is a booster with gain or a buffer going into an inverting stage, this is the input resistor contribution to the noise.

[ElectricDruid]

For what it's worth, I think I've stopped using 2M2 resistors in that position. At one point I was worried about input impedance and thought 2M2 might be better, but I've seen no real benefit over using 1Ms in the same place (R5/R6) so I've gone back to 1M.

Ultimately, I don't think it matters much. There may be measurable differences, but my old ears can't hear 'em!

[Vivek]

my spice simulations show a lot less hiss like this compared to using 1M resistors in single inverting configuration.

Could you share your noise simulations on SPICE.


I see that 2 caps don't have values. SPICE normally would refuse to do any analysis at all.

And the Opamps you used are the ideal noiseless opamps. It might be possible that real opamps contribute a large percentage of the noise.

[Max9999]

Busy weekend, I have time to respond now.

Vivek the schematic shown is just a quick mock up to show it here. I did the noise simulations before that but you are correct that I used the generic opamp model.  I dont know if those are noiseless but the noise came above the opamp that I want to use, opa2134.

Electric Druid and GM, I actually saw examples of a resistor on u2's + to bias. The recommended value was the feedback in parallel with the input resistor. But as I understand it has only something to do with the dc offset, which is no factor here right ( ?) because I use coupling caps as said by Electric Druid.

Thanks everybody, time to build it!

[GibsonGM]

Yes - Tom is WAY more up on those things than I am!  If he says it's not a factor, I'm happy to accept that