Page 2 of 5

Re: Op amp matching

Posted: Wed Mar 27, 2019 10:36 am
by JR.
terkio wrote: Wed Mar 27, 2019 8:48 am
KMN wrote: Wed Mar 20, 2019 12:52 pm I just discovered the LMP2021. It has very low offset and it is pretty quiet.
http://www.ti.com/lit/ds/symlink/lmp2021.pdf
Fantastic Vos. Trouble is, for me, I need +15V -15V op amps.
I just couldn't get the choppers quiet in my circuit, though they are also working for me.
How bad is it ? How to cope with it ?
There may be a way to use a chopper for it's superior DC characteristic in a servo topology that uses just the DC and severely LPF the chopper noise.

No I have never tried this.

JR

Re: Op amp matching

Posted: Wed Mar 27, 2019 6:11 pm
by KMN
When I tried this with the OPA2189 and when it was still making my circuit noisy I gave up. Notice it way over at the right hand side. How much LPF is it going to take? lol

LMP2021 was looking perfect until I noticed the limited rails. The cost is really nice too. It's too bad.

Re: Op amp matching

Posted: Wed Mar 27, 2019 6:38 pm
by KMN
actually, just to be clear. I believe it was the opa335 I didn't like and not the OPA2189. I don't recall if I ever did the noise simulation on that one. 2189.

Re: Op amp matching

Posted: Wed Mar 27, 2019 6:54 pm
by KMN
OK Im at work should be working on non audio simulations but since I'm an ijit and wanted the whole internet to know...went back and had another quick look. For me, in an integrator such as the one I posted above, the OPA335 was horribly noisy even with a ton of LPF.

On the contrary. The OPA2189 is not noisy at all. I don't know why I missed that before. It has excellent Vos specs per the data sheet and also seems fairly reasonably priced on Digikey, or at least reasonable compared to the linear tech parts I normally simulate with.

Summary. Please do have a look at the OPA2189 I made false claims about it earlier.

Re: Op amp matching

Posted: Thu Mar 28, 2019 9:37 pm
by montemcguire
The OPA2189 is a relatively new amplifier - it had 'preview' status in its datasheet up until 12-2018 with the rev-D datasheet, when it finally went into production status. So, you're not sleeping - it's new!

I've been using the Analog ADA4522-2 for dual DC servo uses, and it's roughly equivalent to the OPA2189. I do want to try the OPA2189, and for some circuits, the 2189 might be quieter or have less offset, but for my uses, the 4522-2 works well enough. It gives me an 'end-of-the-day' simulated offset of around 8-10µV, and a measured offset of about the same value, which is frankly at the limits of my Keysight 34461A meter, so that number should probably not be trusted. Still, back to reality, having a direct coupled amplifier with a net DC output offset of 10µV is fine with me.

One hurdle with a dual servo using a chopper amp is the possibility of beat frequency problems that could happen when using two unsynchronized single chopper amps (such as the OPA189 or the ADA4522-1). The ADA4522-2 seems to use one oscillator for both amplifiers, and this prevents beat frequency problems. It also uses a relatively high switching frequency (800kHz?), and a typical integrator circuit will indeed lowpass filter this noise into oblivion. I extensively tested this theory with an APx-555 (which can do 1M sample, 24 bit, 1.2MHz BW FFTs) with the DUT inside of a steel and aluminum Faraday cage, and aside from AM band leakage into the star-quad test leads and the APx, there is no HF gunk from a dual servo built with an ADA4522-2. I also tested the servo integrator with an RC post filter, and that produced the same results as without the post filter - all the hash I measure is due to metropolitan Boston and the imperfections of my test leads and the APx-555.

So, while the OPA2189 is a very attractive amplifier, I spent enough time qualifying the ADA4522-2 (and have a lot of other work to do) so I haven't tested the OPA2189 yet. However, that doesn't mean that you should not consider it - it is truly a fine amplifier, and while I'm just guessing, I would assume that there's only one oscillator in the dual 2189, making it suitable for a dual servo circuit.

Re: Op amp matching

Posted: Fri Mar 29, 2019 10:38 am
by terkio
Thanks for these reviews on ADA4522-2 and OPA2189 for dual DC servo uses.
Giving more than datasheets.
Good to know about beats from asynchronous clocks, an effect one can hardly guess beforehand.
BTW: Which dual servo do you use ? A differentiel Deboo integrator ( that has one cap ) or a more usual differential integrator with two caps ?

Re: Op amp matching

Posted: Sun Mar 31, 2019 9:16 pm
by montemcguire
I use a non-inverting integrator. Here's a picture of a Superbal amplifier with a dual servo for reference:
superbal-with-servo.jpg
This has the advantage of a 'pre-filter' RC lowpass filter, which keeps fast stuff out of the integrator amplifier. The disadvantage is that you need twice as many capacitors to get the job done as a standard inverting integrator. However, I can use small 100nF C0G ceramics for this filter, and still get a low corner frequency, so having to use 4 capacitors for the pair is not a big deal. The circuit shown provides a -3dB frequency of around 0.66Hz, which is a little low, but it's very well behaved both in simulation and in reality. The reason that the cutoff frequency is low is because the 'force' signals of the DC servos are attenuated. This limits the correction range to less than a volt, but also attenuates integrator noise. You can adjust that tradeoff by changing the RC constant (66K5Ω and 100nF) and altering the 33K2Ω and 66K5Ω servo output resistors.

In the Superbal, the two amplifiers can be thought of as being in series, and that's why the tuning is the way it is. By having one servo amplifier attenuate 2x the amount of the other (and thus have a 3dB point 2x lower than the other) you get a maximally flat, optimally damped response with no ringing. I wish I could tell you the grand theory behind this, but I stumbled onto it and it seems to work really well for this Superbal circuit.

Re: Op amp matching

Posted: Mon Apr 01, 2019 4:02 am
by terkio
the two amplifiers can be thought of as being in series
I see a Birt topology.
I do not use the Birt to make differential outputs from differential inputs.
I use twice the same op amp circuit, one to derive Outa = K x ( Ina - Inb) the other to derive Outb = K x ( Inb - Ina).
This is perfectly balanced, more balanced than the Birt where the two op amps don't work the same way.
I use a non inverting integrator too.
Two op amps, the differential input signals go directly at the op amps non inverting inputs, a resistor between the op amps inverting inputs and two caps that feed back to inverting inputs.
This simple integrator is damn handy. It has hight input impedance, there is no need for any matching, the time constant is R x ( C1 + C2 ). The only thing that matters is the offset voltage and the capacitors isolation resistor, where it is likely the stray resistance of the PCB will come in.

Re: Op amp matching

Posted: Mon Apr 01, 2019 4:34 am
by mediatechnology
montemcguire wrote: Sun Mar 31, 2019 9:16 pm I use a non-inverting integrator. Here's a picture of a Superbal amplifier with a dual servo for reference:

superbal-with-servo.jpg

This has the advantage of a 'pre-filter' RC lowpass filter, which keeps fast stuff out of the integrator amplifier. The disadvantage is that you need twice as many capacitors to get the job done as a standard inverting integrator. However, I can use small 100nF C0G ceramics for this filter, and still get a low corner frequency, so having to use 4 capacitors for the pair is not a big deal. The circuit shown provides a -3dB frequency of around 0.66Hz, which is a little low, but it's very well behaved both in simulation and in reality. The reason that the cutoff frequency is low is because the 'force' signals of the DC servos are attenuated. This limits the correction range to less than a volt, but also attenuates integrator noise. You can adjust that tradeoff by changing the RC constant (66K5Ω and 100nF) and altering the 33K2Ω and 66K5Ω servo output resistors.

In the Superbal, the two amplifiers can be thought of as being in series, and that's why the tuning is the way it is. By having one servo amplifier attenuate 2x the amount of the other (and thus have a 3dB point 2x lower than the other) you get a maximally flat, optimally damped response with no ringing. I wish I could tell you the grand theory behind this, but I stumbled onto it and it seems to work really well for this Superbal circuit.
It is a Birt topology. If the Birt is tuned properly to manage overshoots it can be a great line in, line out or A/D modulator driver.
The advantage of the Birt is that the second op amp's noise appears in common mode, not differential. (Assuming good R match.)

We've talked a lot here about the advantage of passive input poles on servos which the non-inverting "simple" integrator has.

Are the inverting/noninverting inputs reversed on U3/U4?

Re: Op amp matching

Posted: Mon Apr 01, 2019 9:20 am
by terkio
mediatechnology wrote: Mon Apr 01, 2019 4:34 am The advantage of the Birt is that the second op amp's noise appears in common mode, not differential. (Assuming good R match.)
Interesting, good to know. Otherwise, I see no advantage in the Birt.

We've talked a lot here about the advantage of passive input poles on servos which the non-inverting "simple" integrator has.
I missed it, please give me links.
I am very interested in an overview of the solutions for a differencial integrator, to know the advantages and drawbacks.
So far I have experimented with the diff Deboo, then switched to the diff non-inverting "simple" ( 2 Op-amps, 2 Caps, 1 Resistor as described above in this thread ).