Page 3 of 5

Re: Op amp matching

Posted: Mon Apr 01, 2019 9:56 am
by mediatechnology
Interesting, good to know. Otherwise, I see no advantage in the Birt.

There is a second when used as a line receiver.
You can vary the gain of the inverter stage and vary the gain of the input without affecting CM rejection.

Self thinks he may have invented the SuperBal in 1982 but I see cites going back to 1970 in National's AN-31 authored by Bob Dobkin: https://proaudiodesignforum.com/forum/p ... 810&p=9537 (page 9 bottom figure.)

The AN-31 cite uses a different polarity of feedback than Self otherwise they are quite similar.

Servo'ing the way montemcguire does pins both differential and common mode DC to 0V which is good for the application.

Re: Op amp matching

Posted: Mon Apr 01, 2019 11:30 am
by JR.
I nominate Wayne to be the circuit design historian... the ancients are constantly stealing our ideas. :lol:

JR

Re: Op amp matching

Posted: Tue Apr 02, 2019 12:14 pm
by JR.
I generally do not pay attention to this sector of uber op amps but Microchip just pushed an ad at me for a 15uV op amp. MCP6V51

https://www.microchip.com/wwwproducts/e ... l-features

http://ww1.microchip.com/downloads/en/D ... 06136A.pdf

10 nV rt Hz is not uber low but respectable.

JR

Re: Op amp matching

Posted: Tue Apr 02, 2019 3:17 pm
by terkio
Interesting: MCPV6V51.
CMOS Vdd 45V
Vos typical 2.4µV
Chopping clock rate of 100kHz
BW 2MHz
Low cost 1 buck

Re: Op amp matching

Posted: Wed Apr 03, 2019 3:09 pm
by montemcguire
mediatechnology wrote: Mon Apr 01, 2019 4:34 am Are the inverting/noninverting inputs reversed on U3/U4?
OOPS! Yes indeed. Sorry for the mistake and good eyes catching it. I "cleaned up" one of my simulations, replacing the auto-generated yellow blocks with triangle amps and got the polarity wrong. It's always something... :)

Re: Op amp matching

Posted: Thu Apr 04, 2019 6:50 am
by mediatechnology
@JR. Thanks I chalk up the "circuit historian" to having (sometimes) a photographic memory.
I have a keen memory of circuits, outdoor locations/geography, and something I can't mention here* that's often heart-shaped.
(* I may run for office some day or be nominated to serve on SCOTUS. ;) )

@Monte. Take a look at the National AN-31 page 9 the lower figure for the variable gain diff amp I cited previously. There you'll find inverting/non-inverting inputs reversed that are actually supposed to be. When I see circuits I always check those feedback signs. viewtopic.php?f=12&t=810

This one threw JR: https://proaudiodesignforum.com/forum/p ... f=12&t=765 another cool one here: viewtopic.php?f=6&t=224
Some of Jim Williams buffers introduce polarity inversion so the error op amp has its inputs reversed. Those are always fun: viewtopic.php?f=12&t=13 Figure 4 here is a double-take: https://www.ka-electronics.com/images/p ... es_pt1.pdf

The input reversal on your servos looked deliberate enough that I had to double-take since positive feedback is either (1) useful, (2) really isn't positive but just looks that way or (3) is unintentional.

@terkio We had a servo discussion that began with the OPA2188. I did a test check with slow op amps as simple integrators to see when/where they stopped being integrators due to bandwidth and load-driving. This exercise reinforced the point JR has made regarding the importance of passive input poles when slow op amps are used. https://proaudiodesignforum.com/forum/p ... ?f=6&t=419

Re: Op amp matching

Posted: Fri Apr 05, 2019 11:48 am
by terkio
mediatechnology wrote: Thu Apr 04, 2019 6:50 am @terkio We had a servo discussion that began with the OPA2188. I did a test check with slow op amps as simple integrators to see when/where they stopped being integrators due to bandwidth and load-driving. This exercise reinforced the point JR has made regarding the importance of passive input poles when slow op amps are used. https://proaudiodesignforum.com/forum/p ... ?f=6&t=419
They stop being integrators at DC, where they become high gain amplifiers instead of infinite gain as would do a perfect integrator.
This DC high gain comes from the cap(s) resistor leakage ( stray PCB resistance included ) or the open loop gain of the op-amp(s) witchever comes first.
This high gain, one whishes to be infinite, will decide of the residual error of the servo.
From this POW, I do not see how a RC cell at the input or the output of the intergrator can give an improvement ( apart from filering noise ). May be I am missing something, I must look harder in the informations from the servo discussion with the OPA2188.
Notes:
_The Deboo's resistance tolerances turns it into an equivallent resistor leakage at the cap. This is the reason, I have abandonned the Deboo.
_The input voltage offset of the op-amp(s) participates in the servo residual error, in a direct way.

Re: Op amp matching

Posted: Fri Apr 05, 2019 12:23 pm
by JR.
terkio wrote: Fri Apr 05, 2019 11:48 am
mediatechnology wrote: Thu Apr 04, 2019 6:50 am @terkio We had a servo discussion that began with the OPA2188. I did a test check with slow op amps as simple integrators to see when/where they stopped being integrators due to bandwidth and load-driving. This exercise reinforced the point JR has made regarding the importance of passive input poles when slow op amps are used. https://proaudiodesignforum.com/forum/p ... ?f=6&t=419
They stop being integrators at DC, where they become high gain amplifiers instead of infinite gain as would do a perfect integrator.
This DC high gain comes from the cap(s) resistor leakage ( stray PCB resistance included ) or the open loop gain of the op-amp(s) witchever comes first.
This high gain, one whishes to be infinite, will decide of the residual error of the servo.
I'll save you the trouble of reading my old comments

Typical op amps have DC voltage gain on the order of 100dB (100,000 : 1) , some even more. Not infinity but more than adequate for typical applications. All this gain has to be attenuated down below unity to assure stability with 100% capacitor feedback, before reaching the high frequency where internal lag flips negative feedback to positive. :o

From this POW, I do not see how a RC cell at the input or the output of the intergrator can give an improvement ( apart from filering noise ). May be I am missing something, I must look harder in the informations from the servo discussion with the OPA2188.
Notes:
The front end passive pole scrubs high edge rate signals from the servo loop, making life easier for modest speed op amps. Even with 100% capacitor feedback the output may not have to swing much voltage in response to HF transients, it still has to deliver the HF current, so internally is slewing like crazy.

Contrary to popular opinion servos remain in the audio path, so transient perturbations could be audible and are therefore undesirable. Of course the value/benefit of this additional real pole depends on the servo topology, if the signal feeding it is already adequately LPF it could be redundant.

JR

_The Deboo's resistance tolerances turns it into an equivallent resistor leakage at the cap. This is the reason, I have abandonned the Deboo.
_The input voltage offset of the op-amp(s) participates in the residual error, in a direct way.

Re: Op amp matching

Posted: Fri Apr 05, 2019 1:15 pm
by mediatechnology
My point was at HF where they stop being capacitors and become "inductive."

The Deboo works fine for what I want. I like its passive input pole.

Re: Op amp matching

Posted: Fri Apr 05, 2019 2:25 pm
by terkio
Thanks, you save me a lot of time pointing to me in the right direction.
The Deboo can be fine, depending of the application and one cap only is nice for a differential.
I abandoned the Deboo in an application where I want the servo loop to give a very best servo error correction.
I bumped on the resistor tolerance issue, where from simulations I saw I needed unrealistic resistor matching.
Then I saw that resitors tolerance boils to an equivallent parasitic resistor across the cap. ( When I get a round tuit, I'll do the maths, but I know I will find: "R_parasitic = R / tolerance" , where R comes from the used resistor values, for R_parasitic infinity you need tolerance zero ).
So, in this application of mine, I am better using an integrator where the limitation doesn't come from resistor tolerances.
So, I switched to a simple non inverting integrator ( differential ). It has two caps that do not need matching ( good luck, because, otherwise it was good for the trash bin ). To optimize, I need caps with Insulation_Resistor times C as high as possible, Op-amp, open loop gain as high as possible ( OPA 2277 is fine with 5,000,000 gain ( from memory ) ).
I wonder what are realistic capacitor IR and PCB realistic insulation resistance.