op amp parameter tradeoffs for charge amp with servos?

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
murrayatuptown
Posts: 3
Joined: Mon Mar 10, 2014 5:56 pm
Location: Holland, MI

op amp parameter tradeoffs for charge amp with servos?

Post by murrayatuptown »

Looks like I typed too long and got logged out. Apologies if a dual post DOES appear, as well as my wordiness. Detail seems to be necessary.

Hello:

I will do my best to make this a relatively short post...very difficult for me...

I am experimenting with some large polymer piezo and electret sensors. The piezos have (documented) very strong pyroelectric output (temperature changes produce large transient outputs, essentially DC shift due to the low frequency most temperature changes exhibit in most scenarios). I need to eliminate this artifact.

The 'large' sensor parameter for my purposes is defined by the capacitance range (depends upon specific application): 250-11000 pF...

This could easily be dealt with something as simple as a high pass filter, which is inherent in any piezo RC-input voltage-amp.

However, I want to test some different circuits for specific reasons...partly proof of concept, partly to optimize some very specific results. This leads to a bunch of other problems.

The inspirational app note that is my starting point is here, for a visual aid. http://www.maximintegrated.com/app-note ... vp/id/1127

There are many changes: size, and impedance and variety of sensors, probably dual rail voltages with rail splitter, totally different purpose from app note with regard to what signal extraction is desired.

There are many advantages to using a charge amp, along with many more challenges. I have been researching app notes, patents etc., to learn methods, and try to be aware of any blatant infringement. This may or may not become 'commercial'. If it does it will be ultra-small numbers. It's really a proof-of-concept for a couple reasons.

Basically, the large capacitance allows large feedback (primary charge amp gain element) capacitance, and relatively low feedback resistance (parallel RC feedback across charge amp op amp i/o). I can get by with 1-20 Megohm depending on the application. The worst case scenario needs relatively flat response down to at least 31 Hz. I'm aiming for 15. This will use the larger sensors and larger resistance values. This is still much lower than hydrophone and photodiode circuits which can be 100 Megohm - 1 Gigohm, which puts high input bias and offset performance requirements upon the input stage op amp. 1-20 Megohm should allow some relaxation of this.

I also think that adding a servo for offset control may also allow flexibility in selection of the charge amp.

The inspirational app note circuit has a fully differential input - two separate op amps, for noise performance. That circuit wants to capture the pyro thermal signal, and acknowledges the acoustic pickup property of polymer piezos, as a 'contaminant' artifact that be dealt with in various ways. Upon achieving the differential common mode advantage for the Q-V conversion, the two half-circuit charge amps are combined in a differential-singe-ended op amp voltage stage. I discussed adding servo to this with Maxim, and they recommended servoing the 2nd stage to avoid phenomena that can occur with two separate servos if they are not identical between the two half circuits, in consideration of the pyroelectric signal being common-mode.

I need to eliminate the pyro signal which can be volts per degree C of temperature shift. If I want a 15 Hz audio low corner frequency, at first take, 0.15 or even lower for a DC offset management servo would seem appropriate...but the temperature change would be from environmental changes like relocation from outdoors to indoors, opening/closing a window, heater or air conditioner cycling. I think such changes are more likely to be roughly in the 0.1 to 1 Hz range. Making the servos as low as possible may not allow adequate control of this 'nuisance' frequency range.

For my purposes, once the Q-V conversion is performed with the charge amp and DC-coupling, I have no problem with moderate DC offsets so the servo need not be as high performance as is typically desired i say, a transformer-coupled mic preamp. I can even AC-couple a subsequent stage. Bear with the stubbornness, please.

What I am interested in discussing, learning, getting input on, is understanding what kinds of effects result from compromises in op amp choices. There is no end of choice for low offset op amps suitable for charge amps, with a wide range of string and weak traits. Ideally, aiming for lower power consumption is preferable. I'd rather have a maximum 500 uA per op amp section than 5 mA...a never-ending frustrating balance of input offset/bias, noise, CMRR, power consumption and cost.

If I go with higher rail voltage op amps, I can relax the offset expectations of the charge amp front end. I'e seen charge amps without servos, but the pyro signal I fear is going to be a killer. Another application only went down to about 200 Hz. I used a voltage amp, there is capacitor coupling and I never gave it a moments thought.

I'm thinking if I can chose op amps (I'll start with a range I have on hand) that give me at worst 0.1-1.0 V DC offset from worst case input drift, and AC-couple the 2nd stage, the servo will possibly only have to control 'ordinary' DC drift as a secondary requirement, and protect against the pyro (and and sub-sonic transients).

I know, no pictures may make it a little hard to visualize...I'll work on that.

Thoughts on the challenge presented by the 1 Hz +/- signal would be appreciated.

I believe a quick summary of my op amp requirements would be as follows:

Charge amp:
Minimal input offset voltage, less concern about bias current as it can be orders of magnitude smaller in modern op amps.
Best CMRR & PSRR I'm willing to pay for to achieve other goals.
Low current noise density followed by low voltage noise density (is there any point in viewing rn = en/in as is done with mic transformers and op amps?; AD797 target, for example). Typical rn for the types of op amps used for charge amps is 330k and up into the megohms. I'm not sure how this applies to large feedback resistors and the sensor is a capacitor...implying a resistance in its equivalent is a large parallel value. en is less important and t
Charge conversion gain is minimal, if any. This will be more important as a stability concern than anything else.
Unsure what advantage there is if any with higher GBW and slew rate devices, other than potentially complicating overall stability.


Servo:
Lowish GBW
Low-mid slew rate?
As much open loop gain, CMRR and PSRR as balance of other factors will afford.
Given lack of an ultra-low DC offset requirement typical for, say, a mic transformer-coupled pre, MC pre, MM RIAA stage, etc., it seems a wider range of 'less-respected-in-common-application' op amps are suitable.

Temperature range is environment a human tolerates (high end perhaps 50C so not forgetting device temperature rise (one reason to limit rail voltages), temperature effects of ib are less worrisome).
Frequency response for the low frequency application will range from 15-30 Hz to several kHz...easy enough to target 20 kHz to avoid mis-perception.

What is it?

Some physically unconventional sensors for string instruments with integral electronics, for custom installation primarily by luthiers during instrument construction. Battery and phantom power versions rather than one-size-fits-all are being explored.

My goals/requirements struck a couple engineers I consulted with as very strange and complex vs. available solutions. Moving past that, the design approach was discussed fruitfully.

Unconventional approaches should not shock anyone in audio. There are several reasons to head in a different direction than existing methods. I've taken several complaints, dislikes and objections to current designs from some luthiers and am trying some approaches I think are not common in such applications.I think I need to prove to myself the benefits equal the trouble of attaining them, and there are many changes I can make if the results do not meet my expectation (I don't see how they cannot...I am using an approach that is proven in other applications tailored to my needs).

Before even attempting a charge amp and servos, I already find the differentially constructed sensors are easier to attain low noise with, even feeding a single-ended input voltage amp.

I've looked at the Maxim family of op amps in the app note, nice performance and reasonable cost, but I think lower power may be desirable and the addition of a servo may allow lower power Q-amp op amps as a result of other combined features. Choosing op amps has been one of the more difficult aspects of this! There are too many choices...who would think that was a complaint!

I'll try to sketch something if necessary, but I suspect it's not hard to envision the proposed changes. (Add servo on each charge amp, AC-couple either the 2nd stage differential input, or it's SE output, possibly add a voltage divider on one input of the 2nd stage to allow it to be unity gain for CM optimization.

Thank you.

Murray Leshner
Holland MI
Murray
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: op amp parameter tradeoffs for charge amp with servos?

Post by JR. »

I read most of your post, I was not familiar with the specific nomenclature "charge" amp , so I googled it. Apparently a charge amp is an amplifier configured as an integrator, or current integrator.

For you opamp selection you should probably look for opamps with low input bias current (like modern bifets) and low input error current.

I do not need to know about your application but if you have such large environmental (temperature) related inputs, is it possible to set up a bridge or dual opposite polarity sensor inputs where the ambient temperature changes cancel out common mode, while valid signal does not?

Have you tried melting solder and seeing how your circuits work?

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
murrayatuptown
Posts: 3
Joined: Mon Mar 10, 2014 5:56 pm
Location: Holland, MI

Re: op amp parameter tradeoffs for charge amp with servos?

Post by murrayatuptown »

Apparently the temperature signal is common mode. Most charge amp schematics I see are not differential, but I know there are some commercial products for vibration analysis that are SE or differential...I guess one chooses what one needs...

I haven't yet because I want a good handle on what specs I do and do not meet with the op amps I already have. I'd like to avoid buying more and more until I need to. Two approaches...spec and purchase something that's close to ideal, hopefully with a 2nd source/alternate...I think I would err toward overspecifying and two options come out of that: Wanting/needing to find '2nd generation' design choices...

If I look back at early documentation for what people used years ago, op amps that are considered 'inadequate' nowadays, it's sometimes based on some idealization. If an older device was sufficient, it suggests I should get a good handle on what the limitations were and how close they come to meeting my needs.

I have a simulation of a modified version of the posted circuit with a single servo on the output stage and some alterations to the sensor and gain distribution. It included a thermal signal and looks like it works...but I do want to move the servos to the input.

I think the only question about 'function' will be verification my prototype is error-free, and assessment of performance results will surely lead to changes...I'd rather be in a position of understanding what compromises I made to start with so I know what I would want to change.

I am getting closer to building a prototype but keep sorting through op amp parameters to balance all the wish list items. I think it will be as frustrating to feel the need to make changes and not be comfortable with which parameters need modification. I think I should include the ability to disable the servos. I considered variable servo ratio, but that will interact with feedback in the overall composite structure. I think yes servo vs no servo to start, then analyze a bit.

On the other hand, I could underspecify, anticipating what I consider less demanding, then upgrade. This might kill two birds with one silicon chip. It's amazing what performance was attainable with available technology decades ago...like putting a JFET LTP in front of a 741 back when there weren't so many options ;O).

Thanks
Murray
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: op amp parameter tradeoffs for charge amp with servos?

Post by mediatechnology »

I discussed adding servo to this with Maxim, and they recommended servoing the 2nd stage to avoid phenomena that can occur with two separate servos if they are not identical between the two half circuits, in consideration of the pyroelectric signal being common-mode.
This jumped off the page. I see the issues with two SE servos having different time constants and it not necessarily being monotonic. Why not use a differential servo at the output of the INA?

Take a look at the Differential Deboo. Last time I looked there were only two references to it.

One is here in this forum: viewtopic.php?f=6&t=559

The other I found here that looks oddly familiar (down to the component values): http://designideas.cocolog-nifty.com/bl ... erent.html
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: op amp parameter tradeoffs for charge amp with servos?

Post by ricardo »

murrayatuptown wrote:Apparently the temperature signal is common mode.
Where did you hear this? Got a link?

Piezo films are 2 terminal devices. Bit difficult to have a common mode signal generated inside the film.

AFAIK (and please correct me if you have more reliable info) the Pyro effect just appears in series with the source capacitance like any other piezo signal.

So in http://www.maximintegrated.com/app-note ... vp/id/1127, C1/2 & R3/4 provide LF rolloff as usual. You need to choose your rolloff to maximise your signal and minimize the LF noise from the pyro effect.

You may want a higher order filter but then you'd do that on the differential stage or after.

A servo is simply a complex way to add a simple filter. It may be justified when a simple filter requires Unobtainium caps & other Unobtainium parts.

BTW, the Maxim article certainly doesn't assume the pyro signal is common mode. It also has several naive & misleading statements but the circuit should work.
User avatar
murrayatuptown
Posts: 3
Joined: Mon Mar 10, 2014 5:56 pm
Location: Holland, MI

Re: op amp parameter tradeoffs for charge amp with servos?

Post by murrayatuptown »

Thank you for the replies!

I had two laptops die & have been traveling and am finally able to get back here!

I am not sure where I got the idea that the pyroelectric signal was common mode.

Ultimately it's based on my misunderstanding..I cannot see anything I would argue on the side of it being common mode right now, but that may have been influenced by some discussions I've had elsewhere. I asked a couple people who are EE's about the circuit and my envisioned servo additions, but since it's not something everyone sees in common use, it took a few rounds to agree on what it was and how it worked. I agree there are something sin the app note that seem to go down a wrong path, but don't recall exactly what I objected to.

I do not have a severe thermal environment, but MSI states the polymer piezo sensors can produce volts per degree C due to pyroelectric response. The Maxim app note example with a hpf corner frequency of 360 Hz still has a rather large output from a soldering iron several inches away.I assume that single temperature transition is essentially a very low frequency drift (up then down). I realize there is gain in both the charge amp and differential to SE stage, so I may be unduly disturbed by the magnitude of the temperature shift shown on the scope trace.

In principle, the thermal transients (the desired signal in this app note) are interference for me, and to not rigorously address them where audio is the desired signal appears to me to defeat any effort to use an op amp with offset qualities appropriate to a charge amp.

Most applications I have been able to find literature on are for single-ended piezo sensors that are quartz or ceramic. I don't have comprehensive data to generalize, but the phenomenon of pyroelectric transients so seem to be discussed for all sensor materials. Most literature I have read discuss ceramic and quartz materials, but the polymeric piezoelectric and 'quasi-piezoelectric' ferroelectret materials (like Emfit film), seem to have the highest pyro response. I say this based more on cumulative inferences in papers and not significant published data. One problem comparing these widely different materials and the range of application-specific circuits is that they are very different. There are some common traits, but vibration systems appear to never use polymer piezos and gas flow acoustic applications seem to favor polymer sensors.

I agree that more aggressive high pass filtering can fundamentally solve the pyro signal problem. I never paid attention to it in my last design, as the lowest fundamental frequency was 190 Hz or so, and I used a capacitively coupled voltage amp...I knew about the pyroelectric response but never noticed it.

The unique quality of the inverting op amp charge amp is the virtual ground and the counter-intuitive low input impedance.

There are several other things to deal with, but the overly simplified main characteristic of the inverting op amp is that the virtual short affords charge transfer without the RC time constant 'charging' time that a voltage amp fed by a piezo has. The feedback RC network time/frequency coefficients absolutely do matter and apply as far as frequency response and noise gain are concerned, but the rise time is much relieved in this configuration.

That is what I wish to preserve, so DC-coupling at least to the differential-single ended stage is desired.

As far as the servo on the output stage is concerned, I did discuss this further with a Maxim app engr. who added a servo at the output in a simulation. A LF sawtooth was superimposed as a thermal transient signal and the output was pretty well controlled. It still seems poorly implemented to allow a random large amplitude 'error signal' in the charge amp stage superimposed over a much smaller desired signal. The charge amp output must be prevented from saturating (due to offset characteristics OR a large pyroelectric output because a servo on the 2nd can only control it meaningfully if the first stage remains linear).

I think the distorted thinking about the pyroelectric response being common mode came from interpreting the audio signal as being obviously differential relative to ground but the thermal signal being present on both on the sensor leads. There is something bothering me about using a 2-wire sensor (signal and 'common') in a 'floating' configuration. Some people apply copper tape to shield them, but usually connect the shield to the 'return'. Some people fold the ribbon in half, with the 'return' on the outside, to provide some measure of 'self-shielding. To copper-tape-shield the entire sensor and isolate both from the ground is probably about all that can be done to call the sensor 'balanced' but there the capacitance from return to ground vs signal to ground seem inherently non-symmetrical. I guess that's a fact of life. I don't know how symmetrical XLR connectors allow a microphone signal to be, but they work. Then there are those dual twisted-pair Mogami XLR mic cables...

Back to charge amp published circuits...Linear Technology has some hydrophone app note circuits with servos applied directly to the charge amp stage. Composite amps may apply some kind of servo at a single stage or in feedback around two stages, or in some kind of 'split-band' arrangement where much higher BW is desired and DC precision is probably more about offset accuracy than LF signals.

The only literature I have seen other than this app note discussing differential charge amps has been for dual sensors that are actually differential (I haven't seen the sensor themselves physically or represented as more than a block diagram. No schematic as the companies are offering systems and not their proprietary circuitry.

So I am thinking very hard about why the entity of a differential charge amp seems to be uncommon. This is the only schematic I have found that is this detailed. One or two others were patents that were more focused on the concepts and overall function than the circuit details.

Someone told me I am thinking TOO hard about it, but it has been evident in most discussions that this circuit has led to misunderstanding at the first impression. One person was certain the servo-equipped schematic I showed him had a gyrator that the feedback RC pair controlled the LPF cutoff. I have indeed seen a paper that said this as well. There is something not entirely accurate about the discussion within this app note, so 'thinking' every last detail through is my task.

I deferred to that first individual who had two more degrees than me and I brought him another paper that explained the inverting charge amp characteristics and the next day he agreed with me - he said he vaguely remembered that from school probably 20 years earlier and had never seen it since ;O).

That's part of the fun of fundamentals and unusual applications. I have made several mistakes with piezo electronics in the past based on not fully analyzing the fundamentals and using assumptions. The worst mistakes were in areas I actually tried to address things! My first 40 megohm-input impedance voltage amp went into the trash. The thermal noise made it unusable and I could not achieve sufficient (electrostatic) shielding on the first sensor I made. It basically had a 'hum vector' axially. I did not attempt magnetic shielding.

If we diverge to the general concept of the range of advantages of balanced circuitry for audio, I see significant appeal in the dual differential circuit, despite it's rare appearance elsewhere. Photodiode preamps offer alot of good principles but I have not yet seen a differential one. I've seen 'dummy' photodiodes and other 'placebo' accelerometers used to keep circuitry balanced.

The only truly differential charge sensors I've seen documented in detail were actually accelerometers. One was in a 1937-ish patent for a 'carrier microphone' that exploited common mode vs differential properties to achieve LF response in one mode that was rolled off in the other mode. (I think it was applied as a phonocardiogram (acoustic vs electro.) but became controversial as it was not an accepted diagnostic method.

At the beginning I wondered why not just use a single op amp as a diff amp since the MSI piezo ribbon is not 'balanced'. I have Emfit material that is 3 wire, seemingly non-symmetrical, but at least isolated, suggesting 'balanced' use is invited. One day I convince myself the double circuitry for a fully balanced preamp is excessively complex, then the next day I think it's brilliant and want to retain it.

The instrumentation amp 'form' in this app note is very interesting. The admission that the CMRR balance the circuit offers is not as good due to difficulty matching capacitors compared to the classical In-Amp resistor balance, but better than a single-ended stage. Since I will be using very different values of sensor and feedback impedances than in this app note, I want to examine the 'mutation' of the three-amp in-amp and how badly the CMRR will be affected for capacitive feedback. The common (+) inputs are analogous to a zero ohm gain-setting resistor in a resistive feedback in amp. I suspect the value of the resistor across the feedback capacitor isn't relevant to the CMRR above DC. I can match the resistors and make some effort to match capacitors, but I suspect DC CMRR is all I'll gain. (No pun intended).

Well, I haven't really asked anything specific. I sure hope someone finds this interesting.

If I were to work on formulating questions...what I'm thinking about is

1) Why do I care about slight variation between two servos...that so-called infra-sonic oscillation possibility? I will probably change the diff-SE stage to unity gain in principle, and the charge stage will likely be run at the minimum gain the selected device is stable at. I can treat the differential-SE output downstream with filters all I want. I wish to preserve the speed of the charge transfer at the DC-coupled input but do not need that in subsequent voltage stages.

2) Does a Deboo integrator function symmetrically with respect to both it's + and - inputs? Maybe it doesn't matter with open loop DC gain...(-) Aol vs. (+) (Aol+1)...or does that make for two different time constants depending on the +/- input? I'm asking before thinking...
Murray
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: op amp parameter tradeoffs for charge amp with servos?

Post by JR. »

murrayatuptown wrote: If I were to work on formulating questions...what I'm thinking about is

1) Why do I care about slight variation between two servos...that so-called infra-sonic oscillation possibility? I will probably change the diff-SE stage to unity gain in principle, and the charge stage will likely be run at the minimum gain the selected device is stable at. I can treat the differential-SE output downstream with filters all I want. I wish to preserve the speed of the charge transfer at the DC-coupled input but do not need that in subsequent voltage stages.
Is that a question?

2) Does a Deboo integrator function symmetrically with respect to both it's + and - inputs? Maybe it doesn't matter with open loop DC gain...(-) Aol vs. (+) (Aol+1)...or does that make for two different time constants depending on the +/- input? I'm asking before thinking...
Wayne is our resident expert on things Deboo...I've never used one. My understanding is yes, the Deboo responds symmetrically to + and - inputs.

JR
Cancel the "cancel culture", do not support mob hatred.
Craig Buckingham
Posts: 17
Joined: Mon Jan 05, 2015 7:49 am

Re: op amp parameter tradeoffs for charge amp with servos?

Post by Craig Buckingham »

Ricardo, you took the words out of my mouth. For common mode output there needs to be a third terminal with some interaction with signal common or ground.

With the Maximum AN1127 circuit I see no need for a DC servo to correct IP offset errors. I make it approximately 66mV maximum common mode error due to first stage worst case IP bias currents (1000pA). Ignored non-inverting IPs as their contribution is so minimal. First stage OP error due to IP bias current offset error I make it 4.4mV per side, 8.8mV in total.

There is no need to servo any common mode from the IP of the first stage AFAICS. The second stage could be servoed to remove the approximate worst case 88mV offset. R5 and R6 provide negligible offset if 1% tolerance type are used. However the overall offset has very little impact on the dynamic range even for a single 5V supply IMO.

Sorry, couldn't follow the original posts. Can someone summarise the issue(s) I may have overlooked.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: op amp parameter tradeoffs for charge amp with servos?

Post by mediatechnology »

I never read the app note too carefully until now but this statement bothers me:
Thermal noise generated by these resistors is not amplified by the differential amplifier. Instead, it appears as a common-mode signal at the differential outputs and is attenuated by common-mode rejection in the following stage.
That statement seems to be an over-simplification.
I'm not sure how the uncorrelated noise of R3 and R4 can be in common mode without at least mentioning the differential to common mode conversion by the capacitance of the sensor bridging the two inputs.
At least that's the only mechanism I see for the two to appear in common mode.
Am I missing something?

And yes I agree the sensor cannot generate a common mode signal without a third terminal.
The temperature effect observed in the app note was differential.
If it was a common mode signal how did it manage to appear after common mode rejection by the third op amp?

And then there's this:
The electrical analog of a piezofilm sensor is a capacitor in series with a voltage source. The sensor has high output impedance and requires a high-impedance buffer amplifier. The circuit shown includes a differential charge amplifier followed by a differential-to-single-ended amplifier. The differential topology reduces line-noise pickup, which is a problem in high-gain circuits.
Can't we (almost) say the same thing for a condenser mic?
Is that where we're headed?
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: op amp parameter tradeoffs for charge amp with servos?

Post by ricardo »

mediatechnology wrote:
The electrical analog of a piezofilm sensor is a capacitor in series with a voltage source. The sensor has high output impedance and requires a high-impedance buffer amplifier. The circuit shown includes a differential charge amplifier followed by a differential-to-single-ended amplifier. The differential topology reduces line-noise pickup, which is a problem in high-gain circuits.
Can't we (almost) say the same thing for a condenser mic?
Yes. The 'only' difference is that a piezofilm sensor has nFs of capacitance while most condensor capsules are in the 10s of pFs.

Scott Wurcer, in his Linear Audio articles on condensor mike circuits, mentions his first circuits of these kind were for piezo transducers.
Post Reply