Hey all,

I hope I'm not repeating something someone else has posted already, but if I am, please let me know.

I'm trying to amplify a DC signal on the order of microvolts (tens of microvolts at times, thus it needs programmable gain). I've tried using auto-zeroing amplifiers (the AD857x series to be exact), but they seem to only take GND and +5 as power supplies and thus do not allow negative signals (they also can't be made into inverters for some reason...).

Second, I tried using reed relays to take the DC signal and make it into a square wave. But, that didn't work either since (as you probably know) there's noise whenever the relays switch, if you go above about 1 Hz, even if you use another relay to short the inputs whilst other relays commute the leads.

Anyways, I then tried using ADG453 chips (semiconductor swithces, basically), but I can't seem to get that circuit to act reliably. Sometimes it switches according to my logic signal, sometimes not. But it definitely will not switch the signal's polarity: it just grounds the signal. The problem with semiconductors (beyond them not working) is that they have thermal emf's that one must worry about, which are on the order of tens of microvolts, so that's a no-go probably.

In short, I'm out of ideas. Any and all help would be immensely appreciated.

 

You may want to consider using an instrumentation amplifier. Here is a link to the tutorial on this website.

http://www.allaboutcircuits.com/vol_3/chpt_8/10.html

The value of Rgain shown in the schematic is a good place to adjust the gain when it needs to be changes.

hgmjr

 

Hi,

Amplifing signals in the range of microvolts is quite feasable. How much gain do you need, and why must the gain change? If you were trying to use the relays to chop the signal, it is normally done after some significant gain has been applied, so the noise atrifacts can be more easily dealt with.

Unless there is some information about the signal we do not have, modern op amps should be able to apply gain in the 10's of thousands without having to resort to signal chopping.

What's the signal source - high or low impedance? And what is the magnitude of the amplified signal?

 

We can't use regular op amps right away due to input offset voltages. Our signal will be on the order of 10's of microvolts (it's a differential thermocouple measurement, essentially) and thus any input offsets must be (at most) on the order of a few microvolts. That's also why we can't use instrumentation amps right away, because even the precision ones like the AD524 have, at best, 50 microvolts of input offset. This is why we tried chopping the signal: that way, any dc offsets would shift the entire waveform and could be subtracted out in the software.

We need our gain to be able to take the signal into the volts range. However, we can't guarantee how big the signal will be, thus we need a programmable gain that goes between (probably) 10^4 and 10^5. We were just going to use a Burr-Brown PGA204 for this. This chip would be used in a later stage, after amplifying the signal out of the range of DC offset issues.

In the hopes that this answers some questions, I will tell you that we are a physics lab and we are trying to measure the Seebeck Effect.

 

Hi,

Life in a lab setting makes things possible that might otherwise be too much trouble. Arranging for temperature control of your circuit, for instance.

All amplifiers have an input offset voltage. No exceptions. Many also have an arrangement to concel this Ios voltage by means of an external trim pot. If you can hold the temperature stable, then the Ios trim works very well.

Many packaged instrumentation amps do not have this trim available (I'm not searching the literature, so perhaps several do). The work-around is to construct the instrumentation amp from three op amps, all of which can be trimmed to negate the Ios error. You can generally get better input resistances out of individual op amps, too. Look at the OPA134, for instance.

If you want to take the time, you can also get several of each resistor and hand select for exact matches. Starting with .1% resistors helps, as does a good Kiethly meter for doing the measurement.

If you're going to take the measurements into a computer, look into an A to D board that has selectable gain - that will eliminate some external circuitry.

Good luck with eliminating all the thermocouple junctions in the connections out to your experimental ones.

 

Have a look at the Maxim ICL7652. It is a chopper stabilised instrument amp with an input offset of 5uV max (0.7uV typical). You can download the datasheet here.

http://www.datasheetcatalog.com/datasheets_pdf/I/C/L/7/ICL7652.shtml