Diff input bootstrapping variations

[atavacron]

Anybody familiar with the Michael Suesserman patent? 1994. Who bootstrapped what, when, has been covered here in years past, so i’m not trying to retread that. Whitlock clearly nailed it in terms of wide acceptance. Suesserman’s approach is focused on non-audio industries - his examples use very large values and require FETs or all-in-one instrumentation amps that aren’t designed for unity.

What I’m curious about are the advantages or disadvantages of this circuit vs. Bill’s circuit, in our world. Both require two well-matched resistor pairs for bootstrap biz; Suesserman’s could more easily be four equal value resistors, and high ones at that. C1-C4 and R1-R4 are described as equal value, but one might scale them so that the cutoff of C1 into R1 and C3 into R2 remains the same (like 1 or 2Hz), with their differential complements doing the same. The main advantage is one less op amp, and it seems more applicable to large gains….mic amps being the obvious AC-coupled application. I’ve tried to work out servos for bootstrapped inputs and haven’t ever found one that sticks (since any value in parallel with the bias resistors negates the bootstrap), but this feels slightly more servo-able, somehow.

I came across it while thinking about buffered -6dB line inputs and front end capacitor/ bias resistor values. It’s pretty easy to visualize either topology without R5-R7 and Rf/Rf/Rg, respectively.






And of course, if @MisterCMRR cares to drop some knowledge, i’m all ears.

 

[ccaudle]

That circuit simplifies to pairs of the circuit in figure 4 for unity gain input buffers (i.e. no R7).
I think the idea is that as R7 decreases it will decrease the differential mode signals fed back to the input, but will have no effect on common mode signals, kind of the converse of how it increases output gain for differential signals but the common mode gain is always unity.
That would mean the effective differential mode input impedance changes with gain, so you might want to put a lower valued resistor from input to input to reduce that if it is something you care about.
The matching concerns for the components on each side still come into play, so the InGenius circuit still probably has an advantage there. Might have just been patented as a way to get around the InGenius patent.

 

[atavacron]

Attached are 1k CM mismatch versions of the Whitlock (direct coupled), the Suesserman (direct coupled), and the Suesserman AC coupled, in that order. There's a capacitor error in the first two, but it's shorted so doesn't matter. I can't grok the last one, it's a notch. Notably, the first two graph out the same in terms of CM rejection. I'm using common values here; Suesserman was using 1u/10M and 100n/1M IIRC. It's buried on page 11 of his patent, linked in the OP.

 

[user 133392]

….mic amps being the obvious AC-coupled application.

I'm not sure bootstrapping can benefit a mic preamp input.

The common mode impedance will be limited, when phantom is on, by the 6K81 pull-up resistors. It will be 3K4Ω in parallel with any other CM impedance.
When phantom is off, and the 6K81 resistors not back-grounded, there has to be some common mode impedance to discharge the left-hand side of the input coupling capacitors. The discharge time constant bounds the CM impedance to something low to mid-ish.

A high-ish CM impedance on the right-hand side of the coupling caps is somewhat beneficial to improve LF common mode rejection arising from capacitor mis-match. THAT's 1510 data sheet shows the "T-bias" configuration where the lower-value bias resistors share a common larger-value resistor to ground.

I've found electrolytic capacitors from the same lot to typically match within ±2%. Films are usually closer to the published tolerance since the tight-tolerance ones get screened out and sold for a premium. That being the case the benefit of a super high CM impedance to improve LF CMRR isn't worth the overhead but a single CM T-bias resistor is worth considering.

As I have mentioned elsewhere T-bias can also be used to obtain a high CM impedance line receiver.

 

[atavacron]

@KA-Electonics.com , nice to see you back here. Are you saying that a bootstrap would lengthen the discharge time constant? I thought that the DCR of the discharge resistor(s) was all that mattered. The impulse to bootstrap those came from wanting to use film caps rather than 'lytics, thus less charge stored. I see your point about 'lytic matching as well.

@ccaudle indeed, the unity gain sim will always show peaking around ƒc with equal values for R1-R4 and C1-C4. It confused me until I thought back to your point and realized that peaking might be non-existent once that feedback source is attenuated. I was able to eliminate the peaking by adjusting the input and feedback high pass capacitor ratio, but that also messed with the CMR (not necessarily in a bad way, but it was hard to keep track). [EDIT: Nope, still peaking with gain - not shown here.]

Anyways:

Here's a pretty-much-real-world CMR comparison with 100R input mismatch (which is on the extreme end for normal wiring i believe). T-bias, the Suesserman, and the Whitlock, in that order. I used skewed 5% caps (470nF) and 1% resistors (50K). ƒc is the same across all three; the Suesserman required slight embiggening of the four resistors in question to match the cutoff (which I suspect would be unnecessary with gain but haven’t checked).

The Suesserman is marginally better than T-bias at 1KHz, better than T-bias at 10KHz, and worse than T-bias at 100Hz. I was not expecting the Whitlock to so decisively outperform the other two (with a CM sample error, even!).

I'm not sure how phase shift in the common mode influences what we hear differentially, but the Whitlock is the only one that hovers around 90 degrees across the spectrum. The Suesserman reaches -270 degrees after leveling out above ƒc, and T-bias...well...I don't know what to make of it.

 

[ccaudle]

I have never seen the Whitlock circuit direct coupled like in your third simulation. That is positive feedback at DC, I don't see how it can be stable.

 

[atavacron]

I have never seen the Whitlock circuit direct coupled like in your third simulation. That is positive feedback at DC, I don't see how it can be stable.

I forgot the capacitor (well, got lazy) and these are ideal op amps.

I have been curious though, about how low of an offset one must have for the cap to be dispensable. Precision amps are pretty good these days. With no gain, and an AC coupled input, does the offset cause more havoc than just doubling?

 

[JohnRoberts]

I would defer to Bill on this subject.
===
My only first hand experience with a similar topology was inside a console back in the 80s. I needed to locate the mic preamp's gain pot some distance from the preamp active components using a long 2 conductor shielded cable. I was apprehensive about hanging that much cable capacitance on sensitive mic preamp gain control nodes. My solution to address that was to generate a CM voltage and buffer that with an op amp. I then used that low impedance buffered CM op amp output to drive the cable shield.

I don't recall spending much bench time on this, it worked so I stuck a fork in it, and moved on.

JR

 

[atavacron]

My solution to address that was to generate a CM voltage and buffer that with an op amp. I then used that low impedance buffered CM op amp output to drive the cable shield.

JR

Shield drivers! I’ve seen a few designs, they seem good for sensitive measurement but all seem to be calibrated for real specific applications. It would be nice to see one tailored for pro audio, just as a reference point. File with ground canceling line drivers…

 

[JohnRoberts]

Shield drivers! I’ve seen a few designs, they seem good for sensitive measurement but all seem to be calibrated for real specific applications. It would be nice to see one tailored for pro audio, just as a reference point. File with ground canceling line drivers…

This technology for audio is pretty mature, again I defer to Bill regarding SOTA for audio interfaces.

Using positive feedback to bootstrap inputs for higher impedance is not without tradeoffs wrt audio performance. There is nothing wrong with researching this technology but I wouldn't expect any easy low hanging fruit available to harvest.

JR

 

[Newmarket]

Shield drivers! I’ve seen a few designs, they seem good for sensitive measurement but all seem to be calibrated for real specific applications. It would be nice to see one tailored for pro audio, just as a reference point. File with ground canceling line drivers…

"Driven" cable screens are not uncommon in instrumentation applications to minimise capacitance "seen" by a transducer. Not understanding what meant by "calibrated" here ?
But be aware that if implementing by means of a triaxial connection with an outer screen connected to 0V then then those connectors (and cable) can get very very "spendy' compared to eg standard coax types.

 

[MisterCMRR]

Anybody familiar with the Michael Suesserman patent? 1994. Who bootstrapped what, when, has been covered here in years past, so i’m not trying to retread that. Whitlock clearly nailed it in terms of wide acceptance. Suesserman’s approach is focused on non-audio industries - his examples use very large values and require FETs or all-in-one instrumentation amps that aren’t designed for unity.

What I’m curious about are the advantages or disadvantages of this circuit vs. Bill’s circuit, in our world. Both require two well-matched resistor pairs for bootstrap biz; Suesserman’s could more easily be four equal value resistors, and high ones at that. C1-C4 and R1-R4 are described as equal value, but one might scale them so that the cutoff of C1 into R1 and C3 into R2 remains the same (like 1 or 2Hz), with their differential complements doing the same. The main advantage is one less op amp, and it seems more applicable to large gains….mic amps being the obvious AC-coupled application. I’ve tried to work out servos for bootstrapped inputs and haven’t ever found one that sticks (since any value in parallel with the bias resistors negates the bootstrap), but this feels slightly more servo-able, somehow.

I came across it while thinking about buffered -6dB line inputs and front end capacitor/ bias resistor values. It’s pretty easy to visualize either topology without R5-R7 and Rf/Rf/Rg, respectively.

View attachment 104627

View attachment 104628


And of course, if @MisterCMRR cares to drop some knowledge, i’m all ears.

I'm going to make some very quick comments here. I'm up to my ears in design work, so detailed answers will have to wait - sorry!
Suesserman bootstraps each input separately (not with derived CM) and also has the problem of the bootstraps being driven from (a non-zero source impedance) the minus inputs of A1 and A2, which will affect differential and CM gains. A feature of my circuit is that, regardless of Rg and Rf (setting diff gain), the common-mode gain of A1 and A2 remains unity - always.

 

[MisterCMRR]

Attached are 1k CM mismatch versions of the Whitlock (direct coupled), the Suesserman (direct coupled), and the Suesserman AC coupled, in that order. There's a capacitor error in the first two, but it's shorted so doesn't matter. I can't grok the last one, it's a notch. Notably, the first two graph out the same in terms of CM rejection. I'm using common values here; Suesserman was using 1u/10M and 100n/1M IIRC. It's buried on page 11 of his patent, linked in the OP.

View attachment 104681
View attachment 104682
View attachment 104683

Not clear to me what circuit corresponds to which plot ("in order" meaning top to bottom plots?). It all looks a bit suspect ... don't see how the plots could be from circuit with ideal op-amps. I'll puzzle over this later ....

 

[MisterCMRR]

I have never seen the Whitlock circuit direct coupled like in your third simulation. That is positive feedback at DC, I don't see how it can be stable.

Nor do I understand how a simulator could avoid overflow and underflow errors using positive feedback with infinite-gain (ideal) op-amps!

 

[MisterCMRR]

Shield drivers! I’ve seen a few designs, they seem good for sensitive measurement but all seem to be calibrated for real specific applications. It would be nice to see one tailored for pro audio, just as a reference point. File with ground canceling line drivers…

At Jensen, back in the late 90's, we designed and built playback and record electronics as plug-in replacements for Magna-Tech 35mm magnetic film recording/editing systems. Because head impedances were quite high and cables from head to preamp were long, I used a driven shield to reduce effective load capacitance on the head to about 5% of the cable's actual capacitance. But, as John mentioned, it takes some careful handling of high-frequencies - op-amps can easily oscillate at any frequency below their GBW figure unless you deal with HF phase shifts!

 

[atavacron]

First thing, I've edited all the op amp characteristics, which were just the program defaults. This is not a snazzy simulator, it's just the one I'm using until I have a chance to dig in to KiCad. I entered the 85° values for the OPA2210; It's much nicer than this on average. I'll offset the input capacitor values by 5% when I run common- and differential mode sweeps on this.

DC analysis demonstrates why the cap is necessary (bottom schemo). Facepalm emoji.

 

[abbey road d enfer]

Shield drivers! I’ve seen a few designs, they seem good for sensitive measurement but all seem to be calibrated for real specific applications. It would be nice to see one tailored for pro audio, just as a reference point. File with ground canceling line drivers…

In the late 90's, I won a contract for an intercom system using the RTS party line system, which is based on a common unbalanced line where the individual stations and beltpacks inject audio current. The signal is between one conductor and the shield. The nominal impedance is 200 ohms. The lines circulated all over the building, for a total of about 3km (2 miles), resulting in about 400nF capacitive load, which limited the audio BW to about 2kHz.
Since all lines were actually shielded pairs, I drove the unused conductor with a voltage follower. It instantly doubled the BW.
I don't remember the details of the circuit I used, probably a TDA2050, which I used extensively at the time.

 

[atavacron]

It all looks a bit suspect [ ... ] I'll puzzle over this later ....

Puzzle no more! These should be easier to digest, and I wonder what people observe.

1 - Whitlock, common mode
2 - Suesserman, common mode
3 - T-bias, common mode
4 - Whitlock, differential
5 - Suesserman, differential
6 - T-bias, differential

100Ω source resistance mismatch, with worst-case 1% resistors and 5% capacitors all around. Op amps are again OPA2210 with worst-case 85°C values. That's a 5µV Vos / 300pA Ib amplifier at 25°C on average.

It took a while to figure out the right common-and-realistic values to make the CMR notch of Bill's circuit and Suesserman's circuit line up. They're pretty much apples-to-apples here, and it's interesting that the Suesserman is almost as effective at CM rejection as Bill's circuit, even with the various component mismatches. The resultant differential mode plots are as expected for Bill's circuit and T-bias (ƒc around 3Hz, 0.1dB down around 20Hz), but the Suesserman differential mode plot is pretty wacky as a result. It can be made to look like the others, but not with similar CM performance.

I was not expecting to see the peaking around ƒc at the buffered common mode point in Bill's circuit. It smoothes out with larger capacitance values, but to get it anything close to Butterworth, you're looking at mF range, or increasing the resistance to ground like with T-bias. I'm guessing it doesn't matter (?), but it would be interesting to dig in to the load seen by the previous stage at various frequencies with various bootstrap cap values.