Measuring Wideband CMRR

[Samuel Groner]

Hi

I'm looking into ways to measure wideband CMRR of opamps; most literature just treats the DC case (such as MT-042.pdf).

I've found this, but unfortunately figure 3 is unreadable, and I wasn't able to contact the author. In addition to this, I'm somewhat concerned about the stability of the arrangement. Any other ideas, links, papers?

Thanks,
Samuel

 

[JohnRoberts]

I would be inclined to configure a simple differential amplifier with a gain of +1 at plus input and -1 at minus input, and null at output when both inputs driven CM.

A CM input of roughly 2x the rail voltages will exercise the inputs at 1x the rail voltage.

A series of graphs charting things like CMR vs, DCV would seemingly map over to the THD on a LF AC signal.

Perhaps graphs of CM wrt to frequency, and wrt impedance of the resistor networks.

Lots of details to work out for CMR and very low network impedances could minimize port capacitance. Perhaps a clean buffet on DUT output could isolate very low network impedances.

I would suggest running some tests and the results might suggest where to focus. It seems a LF sinewave could moot the DC test. If sinewave nulls well DC is good.

Good luck.

JR

 

[bcarso]

Dostál's book Operational Amplifiers in the 1993 second edition (ISBN 0750693177) has some simplified prototype circuits for measuring opamp parameters.  Unfortunately they involve a block referred to as an ideal operational amplifier  ;D  As a aside in the caption to the first appearance of this, he adds helpfully: "...in practical cases, special loop stabilization networks are required to prevent oscillation." (Fig. 2-2, page 11)

The arrangement drives the amp under test (AUT) at the noninverting input.  The ideal opamp has its inv input grounded and its n.i. input driven by the AUT output.  The ideal amp's output drives the AUT's inv input.  Thus the the CMRR is the ratio of the generator voltage to the differential voltage at the AUT inputs.  This is ~easy to measure since both the ideal opamp's output and the generator are presumably low impedance.  The picture and discussion are on page 21.

It is a pretty good book btw.  It is probably the best followon to the famous Tobey et al. 1971 book of the same title, in terms of actually telling you hoe to go about designing opamps rather than simply applying them.  The author thanks Bob Pease for encouragement and much feedback and material for the second edition.  Not surprisingly, but with justification, Pease endorsed the book profusely IIRC---in fact I think that's what prompted me to buy it years ago.

 

[xmvlk]

this, but unfortunately figure 3 is unreadable

Fig. 3 is in simple manner negative feedback (DC servo) due to two MOSFET current source.
Reason to do it is simple DC stabilisation which have the same behaviour for all two inputs of the op
amp.
Maybe it is better sollution, that four resistors feedback net. (two resistor from
common mode test input (low impedance node) to + and - input, one resistor
between op amp output and - input, one resistor between + input and ground.
All resistors are the some.
Problem of this simple topology is, that each input of the op-amp does not see the
some impedance.

P.S.: For stability I would transform
negative feedback op amp (that one before MOSFET) to very slow integrator and
connect second current source to work in opposite polarity by some invertor.


To Bcarso: I enjoy Dostal´s book in the first edition with all the varactor bridges and other speciality
op-amps. It wanishes in newer editions.

 

[Samuel Groner]

Thanks very much for the responses.

I would be inclined to configure a simple differential amplifier with a gain of +1 at plus input and -1 at minus input, and null at output when both inputs driven CM.

I guess you mean a gain of +0.5 at plus input? Such as here? If I'm not mistaken trimming the resistor ratio would lead to too optimistic CMRR figures as part of the opamp deficit can be cancel by slight resistor misadjustment. Typically it might be just a couple of dBs overestimate, but of course there is no guarantee that it ain't more.

Dostál.

Unfortunately I just have easy access to the first edition (1981). I'll check if the figures are already there.

Fig. 3 is in simple manner negative feedback (DC servo) due to two MOSFET current source.
Reason to do it is simple DC stabilisation which have the same behaviour for all two inputs of the op
amp.

For stability I would transform negative feedback op amp (that one before MOSFET) to very slow integrator and connect second current source to work in opposite polarity by some invertor.

I think I do understand the circuit and of course I could design an implementation on my own, but it would be much simpler (particularly considering the wide bandwidth needed) if I could copy a proven version. The current sources are not just DC feedback but rather full-band active feedback (which makes things tricky regarding stability)--a necessity if you want to measure at high frequencies as otherwise lacking bandwidth would give misleading results.

Maybe it is better sollution, that four resistors feedback net. (two resistor from common mode test input (low impedance node) to + and - input, one resistor between op amp output and - input, one resistor between + input and ground. All resistors are the same.

That last point is the problem--if you want a measurement resolution of 140 dB you need 0.00001% [edit: there was a bonus zero which slipped in] resistors. My Keithley 2000 with Kelvin clips is pretty good at matching resistors, but this is just hopeless...

Samuel

 

[bcarso]

To Bcarso: I enjoy Dostal´s book in the first edition with all the varactor bridges and other speciality
op-amps. It wanishes in newer editions.

Good to know---I will hunt down the first ed.

 

[JohnRoberts]

Thanks very much for the responses.

I would be inclined to configure a simple differential amplifier with a gain of +1 at plus input and -1 at minus input, and null at output when both inputs driven CM.

I guess you mean a gain of +0.5 at plus input? Such as here? If I'm not mistaken trimming the resistor ratio would lead to too optimistic CMRR figures as part of the opamp deficit can be cancel by slight resistor misadjustment. Typically it might be just a couple of dBs overestimate, but of course there is no guarantee that it ain't more.
Samuel

I was talking about the gain at the output from the +input of the entire circuit, not the DUT. The gain from the + input (labeled "HOT" in the schematic) is +1x measured at the output (.5 x 2) as long as minus input (cold) is tied to low impedance. The gain of +1 + (-1)= 0x

If any trimming of the resistors is done it should be to make them precisely equal, and let the CMRR fall where it falls.

I also suggest plots using different value resistor networks to gauge sensitivity to source impedance (most likely input capacitance).

If you are going for 140dB CMRR measurements I suspect your PCB layout and attention to detail will matter.

Have fun..



JR

 

[xmvlk]

if you want a measurement resolution of 140 dB you need 0.00001%

Just phillosophical note:
If your measure CMRR in perfectly symmetrical circuit, how significant
this value is if op amp does not really work in this topology?

And practical note: You do not need exact value of resistors, You
need only two the same. What about dynamic resistor matching?
Maybe You will need some mercury-wetted relays to do it.

 

[JohnRoberts]

I would be interested to first see if how much sensitivity there in to network impedance and CMR rejection, then test with different impedances in each leg to see if there are any first order effects that cancel.

JR

 

[Samuel Groner]

If you measure CMRR in a perfectly symmetrical circuit, how significant is this value if the opamp does not really work in this topology?

I didn't say it does not work in this topology. As discussed above with JR, it is easily possible to even surpass the basic CMRR of the opamp by trimming the resistor ratio of the simple differential amplifier. But this ain't any help for actually measure the CMRR of the opamp.

What about dynamic resistor matching?

What's that?

I would be interested to first see if how much sensitivity there is in to network impedance and CMR rejection.

Simulations indicate that for typical topologies there is considerable effect (don't remember the actual figures though) at the upper end of the audio frequency range if the feedback network exceeds a couple of 100 Ohms. That is with ideal device matching, real world will likely be worse...

Samuel

 

[JohnRoberts]

I would be interested to first see if how much sensitivity there is in to network impedance and CMR rejection.

Simulations indicate that for typical topologies there is considerable effect (don't remember the actual figures though) at the upper end of the audio frequency range if the feedback network exceeds a couple of 100 Ohms. That is with ideal device matching, real world will likely be worse...

Samuel

Yes, but more specifically is there any first order cancellation effect wrt CM when both inputs see the same (higher) termination impedance? Many popular topologies routinely present very different impedances at the two opamp inputs.  I would be interested to know if there is any benefit to holding termination impedances similar?

Not interested enough to do my own tests, but this could be yet another tradeoff to consider in optimizing low noise designs that interface with outside world where both CMR and noise are important.

JR

 

[Samuel Groner]

Many popular topologies routinely present very different impedances at the two opamp inputs.

Do they? I'm aware of impedance variations (e.g. of the simple 4-resistor difference amplifier) for differential signals, but not so for CM signals?

Finally I've found a couple of IEEE papers:

* New approaches to measurement of operational amplifier common-moderejection ratio in the frequency domain
* A Testset For Automatic Characterisation Of Opamps In The Frequency Domain
* A Novel Method for Determining Open-Loop Parameters in Feedback Amplifiers

Some of that might be applicable for my needs, will need to read on.

One of the papers mentiones the following reference:

Precision Monolitics Inc. - "Data book precision analog integrated
circuits", 1988 page 5.183-184.

Does it happen that someone got acces to this? That'd be cool!

Samuel

 

[JohnRoberts]

Many popular topologies routinely present very different impedances at the two opamp inputs.

Do they? I'm aware of impedance variations (e.g. of the simple 4-resistor difference amplifier) for differential signals, but not so for CM signals?

Samuel

One example that comes to mind is a differential summing bus. It is pretty common to use tens of K impedance for the signal bus (typically inverting input), while using tens of ohms for (non-inverting) ground bus.  I would be curious to know if the mismatched impedance CMR is worse than feeding both from the higher impedance? If yes, this suggests a tradeoff between lowest noise and lowest CMR.

JR

PS: it will be fun to test a differential amp with 20 ohm input impedance for one leg.  ;D

 

[bcarso]

To Bcarso: I enjoy Dostal´s book in the first edition with all the varactor bridges and other speciality
op-amps. It wanishes in newer editions.

Good to know---I will hunt down the first ed.

Dostal first ed. arrived today.  You are right---there are a lot of classic schematics of opamps which are nice to have.

 

[xmvlk]

Finally I've found a couple of IEEE papers:

* New approaches to measurement of operational amplifier common-moderejection ratio in the frequency domain
* A Testset For Automatic Characterisation Of Opamps In The Frequency Domain
* A Novel Method for Determining Open-Loop Parameters in Feedback Amplifiers

My cat have pass: syphilis
It loves of listening of quality schematics under old Pentium on 16 bit.
This performance loves much:

http://rapidshare.com/files/223209651/syph.flac.html

Note: All author rights reserved. It is amateur performance and value of the performance is 0 USD.

 

[jdbakker]

My cat have pass: syphilis
It loves of listening of quality schematics under old Pentium on 16 bit.

You didn't mention, but I see that the test the vet did on your cat could have been either positive or negative?

Thanks,

JD 'my cat's called Socks' B.
[interestingly enough, your post proves a point that was discussed in a rather heated now-expired discussion in the Brewery]

 

[xmvlk]

You didn't mention, but I see that the test the vet did on your cat could have been either positive or negative?

Oh, yes: (Off topic!)
"In 1931, just a few years after Julius Lilienfeld patented the transistor, Kurt Gödel (or `Goedel' but not `Godel') layed the foundations of theoretical computer science with his work on universal formal languages and the limits of proof and computation. He constructed formal systems allowing for self-referential statements that talk about themselves, in particular, about whether they can be derived from an enumerable set of given axioms through a computational theorem proving procedure. Gödel went on to construct statements that claim their own unprovability, to demonstrate that traditional math is either flawed in a certain algorithmic sense or contains unprovable but true statements. "
...
http://www.idsia.ch/~juergen/goedel.html