Page 4 of 5

Re: Op amp matching

Posted: Sat Apr 06, 2019 5:58 pm
by mediatechnology
I guess I'm left wondering why the need for extreme DC accuracy in the uV level.

I know that you have an interest in minimizing clicks during gain switching but unless you do it at zero cross the odds are pretty good its going to generate a click anyway.

Re: Op amp matching

Posted: Sat Apr 06, 2019 7:54 pm
by terkio
With 10µV accuracy at the integrator input, I get 10mV spikes when switching gain 1000 to/from 500.
That is with no input signal.
The spike is a steep front followed by a RC like decay.

I do not switch at signal zero crossing.
With signal, the spike is blended with the signal. I have simulated this and can see the signal when switching gain.
With 0mV signal, I see the said spike.
With 10mV signal, 1KHz the spike is blended in the signal, the result doesn't look ugly.
With 100mV and more signal the spike is swamped by the signal, it is hardly visible but.......I do see the trouble when the gain change happens at the wrong time. When at the signal peak there is a fast edge.

Thanks, I forgot about this trouble. How bad does this sound ?
Normaly, sound engineers do not change the gain of their preamp mikes once recording is started.

I will switch in 3dB steps, and I assume the user will not span all steps from 0dB to 60dB full speed ahead.

Re: Op amp matching

Posted: Fri Apr 12, 2019 11:34 pm
by KMN
I've been amazed lately at how much bang for buck an OP07 seems like in integrator circuits. Maybe they aren't THE best but they seem pretty good.

Re: Op amp matching

Posted: Sat Apr 13, 2019 5:11 am
by terkio
From the data sheet, one can expect 5µV Vos with trimming using OP07E.
It comes in two flavors OP07E ( better ) and OP07C.
There is another flavor OP07D ( not so good, lower cost ).

Re: Op amp matching

Posted: Sat Apr 13, 2019 10:22 am
by KMN
Wow. The little engine that could.

Re: Op amp matching

Posted: Mon Apr 15, 2019 5:12 am
by mediatechnology
I agree the OP07 is a bargain.

We should be saying the same thing about the NE5532 or, in some apps, the NJM2068.
Those two ICs were about the only 4" product JRC didn't EOL.
Unfortunately JRC did EOL the OP07.

The dual OP07 isn't such a bargain.

Re: Op amp matching

Posted: Tue Apr 16, 2019 5:17 pm
by KMN
Indeed. Though, while appreciation of the NE5532 seems fairly ubiquitous, the NJM2068 seems to fly under the radar for many. I have recently been gaining quite a healthy respect for it in some low noise circuits. When it is the winner it often beats out parts I didn't think could be beat. I am still trying to understand it better.

Re: Op amp matching

Posted: Wed Apr 17, 2019 2:57 am
by terkio
About low Vos balanced integrators I am afraid we are stuck with OPA2277 or go the route of choppers with adaptation board.
I do not see how to use two OP07 with two trimmers. I see no way to trim correctly those two that will interact.
OTOH. One trimmer for the two op-amp makes the adjustment possible, but there are notes telling that an op-amp offset trimming should not be used for overall circuit compensation because it doesn't guarantee optimal drift.

Re: Op amp matching

Posted: Wed Apr 17, 2019 5:36 am
by mediatechnology
It may be ineffective to reduce Vos much further below 100µV given that the 1/f noise of the preamp will dominate.

Re: Op amp matching

Posted: Wed Apr 17, 2019 1:20 pm
by terkio
I am not familiar about the effect of noise on an integrator.
May be naive, I see integration, filtering out noise. I see the noise at the input averaged, buried in the integrating capacitor.
1/f noise looks scary because going to infinity for f going to 0 while gain of an integrator goes to infinity as well for f going to 0.
However,
a) 1/f is noise density. Noise is density x rt ( bandwidth) where bandwidth go to 0 when f is going to 0. So, noise does not go to infinity.
b) A theoretical integrator has gain going to infinity for f going to 0. A real integrator does not. It's gain at DC is limited by the open loop gain of the op-amps AND is limited by the insulation resistance of the integrating capacitor.
These two scary infinities at DC, do not exist in real circuits.
My approach is certainly naive, I have not done the maths, do not know yet, how to tackle this issue and I lack background.
How could I improve and learn to relate 1/f noise with integrator quality ?