Attempt at a 100 V/V gain low signal amplifier - NE5532 offset correction

[Kleinstein]

The larger resistors are OK. The typical values suggested  for the pot for offset trim are a good guess to start with. There is normaly a voltage of some 1 V over these resistors, while the OP07 can swing more like 10 V instead. So the resistor there could be something like 10 x larger to have a similar effect. Chances are one could get away without R4 altogether (one would need a rager value anyway), so just no connection there. The resistor to the positive side and to the OP07 together can than get zero effect when they form a divider to get the initial 1 V or so from the positive supply.
One could go higher with R5 as well - this only reduces the range for the correct.  A little like the divider after the integrator it also effects the cross over frequency / suitable capacitor size.


For the OPs it is not about presence of large resistors, but absence of a low resistance path that can be an issue. The trim inputs have low values resistors inside and work with open pins as well.

[Verdefluox]

So basically to sum up for what I've come up so far.
Referring to ATTACHMENT 1: for the inverting, I adjusted some values and we should have come somewhere for now.

The larger resistors are OK. The typical values suggested  for the pot for offset trim are a good guess to start with. There is normaly a voltage of some 1 V over these resistors, while the OP07 can swing more like 10 V instead. So the resistor there could be something like 10 x larger to have a similar effect. Chances are one could get away without R4 altogether (one would need a rager value anyway), so just no connection there. The resistor to the positive side and to the OP07 together can than get zero effect when they form a divider to get the initial 1 V or so from the positive supply.
One could go higher with R5 as well - this only reduces the range for the correct.  A little like the divider after the integrator it also effects the cross over frequency / suitable capacitor size.

I think I was getting confused then, because I thought that I had to go up A LOT or else I couldn't correct the offset properly.

For the OPs it is not about presence of large resistors, but absence of a low resistance path that can be an issue. The trim inputs have low values resistors inside and work with open pins as well.

I think that this the main point of my doubt. I still don't know if it's a good idea (out of curiosity) placing something > MOhms there, but if it also reduces my correction range there's no point for me going there.
Btw I didn't think about the obvious: they can work as open pins; and regarding the OP07, it shouldn't have problems even if there is a 1 MOhms at its output, right?

Btw in ATTACHMENT 2 I tried to arrange something for the non-inverting too, trying to mix all the various answers. I'm looking carefully the non inverting since I can easily cascade 2 stages of 10x (instead of 100x). I'll check better regarding the noises, even though I don't know how to simulate such a circuit with exposed offsets pins (I have to see if LTSpice has something to offer similar to the NE5532 **with** offset pins).

[Kleinstein]

One wants the adjustment range for the offset correction to be small, as this reduces the additional noise coming from the OP07 and also the gain error: there is a tiny bit (e.g. 1/1000) of signal going via the offsest correction path and one would like it to be small. A a side effect one can than get away with a smaller capacitor. One may still need more than indicated in the plan.
The sensitivity of the trim inputs is often not directly specified and it may need a bit try and error / trim to find the right values.

Equal values for R4 and R6 would do essentially nothing. As the corecting OP has more votlage below it's trim range is more one sided and it may want a high R4, possibly open is R6 is large enough.

There is not real issue with using large resistors in places where they a correct - it is often worse to work around it, e.g. with a T type divider and similar to first divide a signal down in stages. There is a small point that resistors > 1 M may be a bit harder to find with low TC. Many of the series of moderate precision resistors (good thick film or low end thing film) end at 1 M.

[Verdefluox]

So I leave it there in case someone may be interested in the future. I increased the values for R4 (left it as in AN21) and R6 to make the parallel connection more similar to the original 12k value. For R5, I increased to 600k to take some distance from R6, and has to be tweaked either up or down to adjust the offset sensitivity. In any case, one could tweak a bit also the integrator cap C1 to slow it down if necessary, but I think that overall this is a good starting point, and the rest are adjustments.

EDIT: Also attached a tweaked inverting configuration after trying to perform myself something some noise calculations with LTSpice. C1 helped me at lower frequency, and one can also tweak C2 in case wants a different BW (obviously, this affects the overall output noise level -because of the the different BW-, but the density itself doesn't change). Many thanks to user magic for helping me with this.

EDIT 2: As a final afterthought, one can also think of splitting the system into cascading 10x (which also seems nicer for a simple amplifier project), which overall lowers the noise a little bit further. There's also a rough noise analysis done by myself. For 10x, I made a test trying to increase the resistors to 10k/100k for an higher input impedance, but obviously one gets lower noise with 1k/10k divider, with integrator resistor still at 10k. After further trials, around 3k/30k seems a good upper limit.
Finally, mind that this simulation still ignores the NE5532. Results are below

[bson]

What is it going to drive?  You might want to add the load to the output; the NE5532 in particular is load sensitive and won't like going far below a few hundred ohm.  It's universally measured with a 600Ω load.  (This is the test circuit you will find in any datasheet.)  However, the spice model may not effectively simulate its propensity for gain compression as load currents go up, so this is best verified on a breadboard.