2

According to this question, it is possible to build instrumentation amplifier out of resistors and operational amplifiers. I am more interested in the version where the gain is set by changing Rgain resistance.

enter image description here

The goal here is to build INA125 or INA122 in-amp. Hence, a couple a questions:

  1. Which op-amps could I use in this circuit? Specifically, would LM124 or LM158 be appropriate? Would you recommend some others? Note that the circuit will be used in 5V low-power design at sub-zero temperatures.
  2. What drawbacks can I expect when using this substitute instead of INA125 or INA122?
  3. If this whole idea is silly, is there some other way to amplify the signal coming from a Wheatstone bridge?

Here's some extra information about my question. I am building a weight scale and I'm using four load cells connected in Wheatstone bridge. I get some 2.3 mV for 100 kg and need to amplify this to use with 8-bit Atmel AVR MCU. Unfortunately, where I live, suppliers do not have any kind of instrumentation amplifiers, so my only options are to try to build one or to wait 30+ days to be delivered from online sellers.

Community
  • 1
Nikola Malešević
  • 616
  • 2
  • 7
  • 20
  • 2
    Nikola - the problem with the circuit as shown above (lumped parts design rather than integrated) is the required equalness of the four resistors R. These resistors determine the resulting common mode properties of the amplifier. Hence, it depends on your requirements which approach is suitable for your application. – LvW Sep 07 '14 at 10:40
  • Exactly, take very low tolerance resitors + a compensation trimmer. E.g. make R=10kOhm+100Ohm on one side and 10kOhm+200Ohm Trimmer on the other side (1% tolerance for the 10KOhm). Then clamp the inputs, put them to the maximum voltage common mode and trim the CMMR away. And i cant believe that it takes 30 days to get an instrumentation amp delivered. – Sascha Jan 05 '17 at 13:36

1 Answers1

3

  1. You can use any amplifier you like. You will get better performance if you use a modern $5 dual op-amp than a 2-cent/amplifier LM324. Sub-zero means nothing to me. If you want to the amplifier to have guaranteed characteristics below 0°C, then you should buy one that is guaranteed for that temperature range. Below -55°C you may have to qualify them yourselves. For temperatures down to -20°C it's unlikely that anything untoward will occur other than possibly slightly worse performance for a chip guaranteed over the commercial temperature range. A single 5V supply will impose significant limitations on the input common mode voltage. You can add pullup resistors to the output of an LM324 to get a bit more range (you probably do not want to do this on the output amplifier). Note that the left two amplifiers can saturate at either rail, so even if the output should be within range that doesn't mean that the amplifier will work properly. I presume your common mode input voltage will be close to 2.5V so this may not be a problem. LM158 and LM124 ancient designs are particularly crappy amplifiers for low level DC- large offset voltage and large offset drift with temperature.

  2. See above. Offset voltage, offset voltage drift. Probably useful range of input. Closed loop gain accuracy will likely be lower too, because of op-amp gain.

  3. If you could use an analog switch with your micro to auto-zero the amplifier you could improve performance. A SPDT switch is sufficient (eg. 1/2 74HC4066). Just transfer one input to be shorted to the other at ~2.5V and measure the amplified offset voltage. Average over many measurements and subtract from the measured signal.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • 'sub-zero temperatures' means below zero degrees C, the freezing point of water. – Pete Kirkham Sep 06 '14 at 17:25
  • @PeteKirkham Yeah, I know, but *how much* sub zero? Just about every IC is rated to work down *to* 0°C, and for a hack a few degrees below zero isn't going to be a problem, so it's utterly meaningless to say sub-zero. Unless you work with cryogenics in which case it's just wrong. – Spehro Pefhany Sep 06 '14 at 17:29
  • 2
    Nice answer, I wonder if any/which opamps would work down at 77 K. (I should try some.) – George Herold Sep 06 '14 at 23:47
  • @GeorgeHerold It's possible the CMOS ones would work. Bipolar carriers freeze out, as you know. I'm going to be testing some discrete stuff at ~4K in the next few weeks (want to get a hold of some papers first- it's expensive to boil off He). – Spehro Pefhany Sep 06 '14 at 23:52
  • Here's at least one... http://ridl.cfd.rit.edu/products/cryogenic%20electronics/opa350.pdf – George Herold Sep 06 '14 at 23:55
  • Re 4K: Way back when I had some mosfets that worked at 4K. (at least I think they were mosfets... ) – George Herold Sep 07 '14 at 00:00
  • @GeorgeHerold Thanks for the link. Just the sort of weirdness I'd have feared, if not expected. – Spehro Pefhany Sep 07 '14 at 00:05
  • Subzero temps in my case would not go below 258K, so that escalated quickly into cryogenics domain. I've put that info just to justify my recommendation of LM124 instead of LM324. Thanks for the answer, it really is helpful. I see that drift for LM124 is 7µV/°C. That would mean that I would get 13 kg of error at -15°C, which is unacceptable. Do these independent drifts add up in op-amp configuration depicted in my question? Also, do you have any recommendations on op-amps I could use instead? To get an error of ±1kg, I'd need drift below 2µV/°C, if I'm right. – Nikola Malešević Sep 07 '14 at 12:20
  • 1
    @NikolaMalešević there are many types-- look at so-called zero drift (or implement yourself as I suggested). Drift can be <50nV/K. White noise of these tends to be higher than the best precision op-amps. – Spehro Pefhany Sep 07 '14 at 12:25