Op Amp - Design question

Do you need to have the system remember the start-up voltage for ever (i.e. DC couple but reference to the start-up voltage), or do you need to be truly AC coupled?

If the latter then Jim's suggestion would be good, or an AC-coupled circuit with a switch to quick-charge the cap on startup.

If the former then you need something that will acquire the DC voltage and remember it for ever and a day, which points to digital circuitry; anything from an up/down counter, DAC and comparator to a processor, ADC, DAC, all that folderol.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

AC couple, but clamp after the capacitor at some small level, say ±30mV, to get quick recovery, otherwise the big DC step will hang up your amplifier for a long time.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Johnson:

This is exactly what I needed. Was going to get a micro and do myself what this chip does automatically.

Thank you so much. I have already ordered a sample.

Dave

The important facts you haven't given us are, how long does any particular "DC" level last and how fast do the small fluctuations happen? In other words what are the frequencys of the average bias value and that of the "fluctuations". Your manual null adjustments seem to be part of a low frequency high pass filter.

Could you use the TI auto-leveling op-amp, the TLC5402ADR.

It won't help you.

That device simply provides a calibrated input offset voltage.

No help in your application at all.

Graham

OK time for some more information. First I am an EE, unfortunately I have only done digital stuff. Let me back up and give some more information:

I have a sensor that is driven by a 150KHz sine wave 5V p-p. The output of the sensor is the 150KHz with anywhere from a 1V to a 3.5V p-p amplitude (this voltage is dependent on other things that leads to a change of value under different operating conditions) under steady state. With the sensor stimulated the amplitude can fluctuate +/- 20 mVolts. I had passed this signal to an envelope detector which gave me basically the dc offset signal I spoke of in my OP. Maybe this was a bad thing, and I should look at processing the AM signal directly. The rate of change is somewhat slow. The full 20mVolts shouldn't be traversed in less the 500ms. What I want to end up with is a signal that I can sample with A/D and small micro at about 2KHz. The resolution I need combined with the A/D part lead me to need a 0-3.6V source signal.

Thanks for the input so far.

Dave

Your needs and what you have seem a bit unclear to me, but I'll try to get it right:

• at the output the sensor will give you 1V to 3.5V peak-peak AC.

• The actual signal you are interested in is a 20mV variation in this peak-peak value.

• You need to somehow acquire and reject 1 to 3.5V of signal so you can see 20mV of variation.

That seems like a lot of noise and a tiny bit of signal. Do you have confidence that you'll be able to do it at all? How much resolution do you need on that 20mV signal? Do you just need to resolve the 1-3.5V signal down to 20mV?

In general with these AC-coupled systems you want to demodulate as late as possible, with the best circuit you can come up with, because once you've demodulated you're subject to DC offsets. But it looks like you already have a whopping big offset as it is, so I don't know that your envelope detection won't be as good (read "as bad") as anything else.

This sounds suspiciously like you're asking more from your sensor than it can deliver. Either I'm misreading what you're trying to say or you're going down a rough and rocky road, with darn few rosebushes along the way.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Well, with taking the output of the envelope detector and passing that to a difference amp as described in my OP, I am amping ~150 times bringing the output of my amp to ~3-4 volts. It isn't very noisy and I am able to sample and get the resolution I need. The sensor is actually functioning perfectly. My only issue is in pratice, I won't have a $4000 programmable PS available to null out the DC component at the difference amplifier.

Being a digital guy, what I was thinking was to take the steady state out of the envelope detector, passing it through a A/D, and storing it a small micro. Then I could use a SW programmable pot or similar to "null" the effect of the DC component into the difference amp. This approach will work, but I figured I would investigate an analog equivelent that would be cheaper as well as take up less real estate.

Dave

Also , is the output of your sensor a function of the level of the 150 kHz drive that you apply to it. If so , you will also be subject to errors due to changes in the drive level...

Is there a fixed phase relationship between the 150 kHz drive and the

150 kHz output? You may want to make some kind of bridge circuit that combines the 150 kHz drive signal and the 150 kHz return signal and creates a null at 150 kHz, then envelope detect the residual..., then you can sense any changes that occur after that....

Mark

First, if you're going to use Google Groups could you cut and paste the past history of the message into your follow-ons, so it looks like you have a real newsreader?

Second, I'm happy that you get what you need out of the envelope detector, but I'm still having trouble wrapping my brain around the notion that you're getting that much AC that you're having to null out, and that after doing so you're getting a suitable signal.

Second-and-half, you can almost always get better performance converting AC instrument readings to DC with a synchronous rectifier. You generally need an instrumentation amp and a few switches, and a reference signal from the sensor excitation.

Third, if you have a micro in there already perhaps the easiest thing to do would be to drive the difference amp from a good DAC. Drive the DAC around until you've found center using your existing ADC to close the loop. You can get 16-bit serial DACs in 8-pin packages, so you're not going to use much board space. Acquiring and holding that large of a voltage with that much precision for an indeterminate amount of time is going to be difficult if not impossible;

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

In order to solve this, you need to define "vary" and "fluctuate" in terms of frequencies or rates of change. You could periodically reset a sample and hold if you have some way of knowing when the sensor's output should be zero. That is presumably what you're doing manually with your programmable supply. If your durations are long (hours) then a digital sample & hold would probably be best. You'd probably want a minimum of 12 bit resolution. That's still 1.22 mV with 5 volt full scale. 16 bit resolution would be better.

If your desired signal (measured parameter) is intrinsically AC with a longer term average of 0, then you could use a longer-term average to subtract from your signal -- but you need to define what "long term" means in order to make intelligent tradeoffs about circuit approaches.

If the information part of your sense signal (the 20 millivolt part) does not itself have a DC component, another way of looking at this might be as a high pass filter, albeit with a rather low corner frequency. This is probably best done digitally. If there are known measurement events and other times when the sensed parameter is known to be zero (e.g., a scale that either has a load or it doesn't), then an auto reset (sample, hold and null to zero) would probably be the easiest and probably best approach.