Op Amp Voltage Offset Compensation, Looking for temp compensation cct.

I'm building a little circuit with a TL072 dual op amp (basic inverting amp configuration) and because of the very low signal levels that I'm dealing with, it will need offset voltage compensation. So far, I have a simple compensation circuit, using a voltage divider and series resistor, but I'm not sure how well this will behave as the temperature changes (I guess I can get out the heat gun & do some testing).

Somewhere, I recall seeing a compensation circuit that includes a diode arranged so that its Vf vs. temperature tracks that of the opamp inputs so that the compensation remains stable over a wide temperature range. I want to keep the parts count as low as possible, so I don't want a super complicated compensation scheme. Does anyone recall such a circuit?

Pls post your circuit.,....

The tempco is not specified for sign (I believe Vos and its tempco are magnitude, not signed), so you'll have to calibrate each and every one for tempco, and pray it doesn't drift over time.

Better to use a precision op-amp in the first place, which JFET input amps are not.

Also, the TL072 is unusually noisy, 18nV/rtHz, more at low frequencies. Fancy amps these days are below 8 or so. No clue if that makes a difference, just something I hadn't noticed.

Tim

Thanks Tim.

The old Nat. Semiconductor app notes that I've been reading were based on the older crappy op amps that likely drifted all over the place. So, maybe it won't be that big a problem.

There doesn't seem to be any noise problem at the signal levels that I've been using so far. I do have some LM833's that I started out with. However, I had more compensation problems with them than the TL072s due to higher input currents.

What would you suggest for a precision op amp that doesn't break the bank? I'd like to use a dual op amp because of space considerations.

Well, what's in your bank? If you have sixty three cents left, better grab a hat and head to the street corner...

'Course the good amps are more than $0.63, usually a few bucks. But even a $5 amp is chump change out of a whole project. Yeah I know, it ain't like the resistors you pulled off that board for free, but frankly, even my hobby time is worth more than that!

I haven't had to delve into precision amps yet myself, but I've ran across a few. Let's see...
http://cds.linear.com/docs/Datasheet/12134fa.pdf
http://www.ti.com/lit/ds/symlink/opa2227.pdf (typ. open-loop gain 160dB -- that's one followed by eight zeros!)
http://www.analog.com/static/imported-file...0_8512_8513.pdf this one's somewhat cheaper (wow, never thought I'd say that about an ADI part )
http://cds.linear.com/docs/Datasheet/16789fs.pdf (very slightly cheaper than the other LT part)

There's a couple of other manufacturers in the show, too. Lots of stuff to find on Digikey and etc., just don't let yourself get overwhelmed by the choices, concentrate on what you want (Vos(typ) in the uV, bandwidth sufficient for purposes, required supply voltage, output current, etc.) and play around a bit with parameters. See what pops out. And sort by "In Stock" to eliminate unavailable items, packaging (bulk, cut tape, tube) to remove multiple choices (Digi-Reel/full reels, boxes), and case if you have some expectation of how it's shaped (e.g., DIP8, SO-8, etc.).

Nice thing about high gain, low offset amps is, you can design a circuit that *works*, the first time, every time, within guaranteed limits, without requiring any trimming or calibration whatsoever.

Mind the resistors, too -- you'll get a lot of tempco, even (now this is down in the nV) thermocouple junction effects from gradients and inhomogeneity, from cheap resistors (speaking of: note that 1% chip resistors are the same price as 5% in most quantities), let alone resistive tempco and plain old matching. 0.1% or tighter resistors also get expensive.

Tim

Yes, this is always the problem with DigiKey.

Yes. That's my philosophy too. Well, up to a point. If an el-cheapo part does the job, that's fine, but it's not a bad idea to start out with good stuff, and then simplify & cheapen later.

I'll check out the parts you listed.

It's possible that I'm just overthinking this thing, because it started out as a little mental exercise that, initially, I didn't think would be practical at a reasonable cost. Then after more thought, it did seem possible.

What I'm building is a wattmeter (DC to 10kHz or thereabouts) that will measure down to 1 picoWatt. So, I was naturally worried about noise problems, but upon further consideration, realized that if voltage and current are considered separately (each has its own amp) then it's not that big a problem. Noise measured in the actual load will translate into an actual positive power reading as it should, but noise generated separately in the voltage amp and current amp will be uncorrelated, and should cancel out. So, a bit of noise generated in the op-amps, as long as it's not excessive, should be okay. Of course, voltage offsets are not okay. It's working fine on the breadboard, but I'm concerned about long term stability. I don't want to have to go through the re-zeroing procedure (4 adjustments) every time I use it. Time for the heat gun test, I guess.

Just had a look at the premium op amp specs. Interesting that after all these years, they can't get the offset voltage down into the nV range.

I have to put in an order to DigiKey in the next few daze, so I'll order a few OPA2227's. I guess I can afford 5 bucks a pop (for one or two). Still, with a 75µV offset, I'll likely need to add some compensation. It's all academic anyway, because the multiplier chip requires offset adjustment, and since it all accumulates, it can just be done once.

Ah, if it's extremely low offsets you want, and fairly low bandwidth, try shopping for chopper amps!

I expect you'll be doing it in ranges, in order to maximize SNR? Doing 12 decades of *anything* all at once is just about impossible, electronically.

When you need ludicrous gain, consider using several amps cascaded. The first amp is the most critical, but you'll want to make sure you don't pick up a harmful amount of offset and noise in subsequent stages either. You probably won't get enough GBW (gain-bandwidth product) from just one -- to resolve 10kHz at 1uV on a 10mV scale, you need 10^4 (80dB) gain. An amp with GBW = 20MHz will get you only 20M / 10^4 = 2kHz bandwidth (voltage op-amps are all dominant-pole compensated, so it's always a straight 20dB/decade relationship, give or take a couple dB around roll-off). However, two such amps, cascaded, will get you 200kHz each, for probably 50kHz bandwidth combined. The exact nature of roll-off might have to be scoped out, in which case you can apply "peaking" capacitors to tweak the response for best response (probably a Bessel or Butterworth filter response; make sure it's matched on both voltage and current channels!).

Note that multipliers themselves have annoying errors. You might consider a simple microcontroller and ADC combination; dual 16 bit ADCs with ample bandwidth are highly available. Then you can do filtering, multiplication and integration (for energy) or averaging (for power) in software without drift. Plus the uC + ADC is cheaper.

Tim

good ideas tim .

man , i remember when THE name in high performance opamps was "burr brown" .
and they were very proud of their stuff ... much more expensive than the other stuff available at the time .

Since their purchase by TI, they've been up and down in terms of production and design, but they're still a big name as far as I know. Want something unique with jaw-dropping specs? Well pull your jaw up sonny, 'cuz wait 'til you see the sticker price...

Tim

Was going to do more 'sperimenting tonight, but I got a better offer which involved the sampling the output of several local boutique breweries. So, no contest.

Meanwhile, I've relaxed the specs somewhat. I've removed the DC requirement, and reduced the upper frequency limit to maybe 5kHz or thereabouts.

And also just to clarify, the 1 picoWatt number is the minimum resolution, not full scale.

From the breadboard results, I'm pretty sure this can be done with standard low $ parts. May involve a bit of manual zeroing, but I figure I can use a bit of ingenuity to make it reasonably painless.

Will report back once blood alcohol level recedes to levels safe enough to operate electrical equipment.