I've made the following circuit, but I just can't get it to work. My LTspice simulation suggests that it should work, but it doesn't. I can get it to work in a normal inverted/non-inverted setup, but I need a differential reading.

I understand that some precision opamps have built-in precision resistors at a fixed gain?

I'm using the INA214 which has a gain of 100, but even that is giving me +-5% noise

Do you have any suggestions regarding a good 'through-hole parts only' opamp (e.g. DIP8) which is great for this purpose?

I've found a lot of SMD-type opamps, but they're a pain to work with. I'm looking for about 50-100 gain, with input range 0-80mV.


simulate this circuit – Schematic created using CircuitLab

enter image description here

  • \$\begingroup\$ what is the current that you want to measure? (Order of magnitude) \$\endgroup\$ – Marcus Müller Mar 19 '17 at 19:06
  • \$\begingroup\$ So, are you using INA214s, or are you using the circuit shown with LM358s, or the circuit with LT1014s? What is your power supply like? Are you using decoupling capacitors on whichever amp you're actually using? Also, 80 mV out of a 750 uohm shunt implies currents of 100 amps. Are you sure you want to do this? \$\endgroup\$ – WhatRoughBeast Mar 19 '17 at 21:18
  • \$\begingroup\$ @WhatRoughBeast, Yes, everything is dimensioned for high current. I'm using the LM358 as well as the INA214, for comparison and different gain setups, to give wider current range. I'm currently powering the opamps with 5V, to protect the ADC which is limited to 5V. Although I'm now considering supplying the INA214 with a higher voltage to have a larger range, and use a voltage divider to provide 5V into the ADC. \$\endgroup\$ – user95482301 Mar 19 '17 at 21:34
  • \$\begingroup\$ @user95482301 if you load is .5 ohms then you're looking at a power of roughly 300W. Is this true i.e. is it a heater you're monitoring? \$\endgroup\$ – user1890202 Mar 19 '17 at 21:54
  • \$\begingroup\$ Check out auto-zero or choppered OpAmps. Many new ones work 0/5v rails. \$\endgroup\$ – analogsystemsrf Mar 19 '17 at 23:03

You have left out a number of important requirements, so I'll work from the assumptions which are easiest to deal with.

You do not specify the precision you need. Since you specify a 750 uohm shunt and an 80 mV signal range, you are clearly interested in 100 amp signals. Let's say you want 0.1 amp accuracy.

You do not specify your power supply. Since you want a DIP package, you're clearly pointing toward older op amps, so I'll assume you are happy with the older requirement for +/- 15 or +/- 12 volts.

You do not specify the frequency response you need. I'll go with 100 Hz or less.

With these points in mind, you can do pretty good work with a breadboard provided you take some precautions. The most important is establishing a ground point. What you should do is take the horizontal row just below your op amp and connect the two sides with a 1/2 inch jumper. Nothing longer, and if you can go with a 0.3 inch jumper that's better. Tie all ground points to this.

A very reasonable op amp for this application is the OP177, which is still available. Vos is 10 (typ) uV (equivalent to about 0.013 amps) and TC is about 0.3 uV/C max. Of course, it's horribly slow, with a GBWP of less than 1 MHz, but you can't have everything. It should do for the frequency range you're (apparently) interested in.

A circuit would look like


simulate this circuit – Schematic created using CircuitLab

The input RCs will reject common mode frequencies above about 100 Hz, so the op amp will be able to deal with the remainder, and pulse response is about .25 msec.

Note that, for accuracies even approaching 0.1 A, you'll need to use 0.1% resistors, and you'll be well advised to match your input caps. The 0.1 uF should be ceramics.

All connections should be as short as possible, and the use of jumpers to connect the components is forbidden.

Since you are obviously working in a high-current environment, ground noise coupling is likely a problem. You may well be advised to put your breadboard in a grounded metal box, and possibly use filtered feedthroughs for your signals. If so, putting the power supply in the box as well would be good idea.


breadboard-compatible opamp which is great for this purpose?

Low-noise measurements and breadboard – might be a bit hard. Breadboard is generally know for the following properties:

  • bad connection reliability
  • high line reactance, making it hard to observe fast-moving signals at all
  • high cross-talk between neighboring lines
  • relatively high resistivity at contact points.

In other words, when dealing with mΩ resistors, you simply can't use breadboard, the unwanted effects can very easily overshadow that.

Then, you're really driving your Opamp rail-to rail – that might make accurate measurement very hard.

I'd use a boost or inverting SMPS circuit to simply increase the supply voltage of my opamp. That way, I could use a single shunt for both directions, since I could guarantee that the top and bottom voltages of said shunt would be well within the input range of the Opamp.

  • \$\begingroup\$ Indeed I mean to solder the component to a prototype board. I meant to say "'through-hole parts only' opamp (e.g. DIP8) ". Is rail-to-rail opamp, or providing a negative supply, the main problem here? I believe some opamp ICs come with an integrated -0.3V supply? Or are those typically what is called "rail to rail"? \$\endgroup\$ – user95482301 Mar 19 '17 at 19:22

This might better be done with negative supply, say -5V. Common mode range and output range will come into play. Your desired limitation of 'through-hole parts only' will tend to paint yourself into a corner. In general micro-ohms and breadboards don't mix well. You should, however, be able to make this work if a Kelvin-connected shunt is carefully wired outside the breadboard.

Reading the datasheet, the INA214 might be expected to have about 30mV p-p noise at the output, any excess is probably not due to the in-amp. You can reduce the noise by further reducing the already fairly miserable bandwidth, but the typical 25nV/\$\sqrt{\text {Hz}}\$ noise of the INA214 is not atypical for this class of product.

An LM358 as you show is totally unsuited due to the poor DC performance- you want something with microvolts of offset and very low drift with temperature. Only 75uV of offset represents a 0.1A error in your measured current. With a single supply the output must swing close to the negative rail to get close to 0A reading and the offset may add or subtract from the required voltage.

  • \$\begingroup\$ What about OP37 then? I'm reading: 80nV/p-p noise, 0.2uV/C drift, 10uV offset. I can't seem to find any through-hole rail-to-rail opamps with microvolt offset \$\endgroup\$ – user95482301 Mar 19 '17 at 19:29
  • \$\begingroup\$ @user95482301 I don't know the OP37 datasheet by heart, but try something that doesn't have a two- or three-digit type name – those were mainly designed in the seventies and have bad performance by modern standards. TI's website's opamp catalog might help. \$\endgroup\$ – Marcus Müller Mar 19 '17 at 19:30
  • \$\begingroup\$ @MarcusMüller, I see. As far as I can tell, all of them are SMD type. How do you prototype them? Do you solder them yourself to an SMD breakout board? Or have a PCB manufactured with a pick-and-place machine for testing purposes? \$\endgroup\$ – user95482301 Mar 19 '17 at 19:35
  • 1
    \$\begingroup\$ But as said by Sphero, only using through-hole components will be a very limiting factor here – for example, the OPA388 (Offset Voltage: ±0.25 µV,temp drift: ±0.005 µV/°C) would probably be a fine choice, but only in the SMD package. Prototyping: yes, solderig them to a breakout board works. \$\endgroup\$ – Marcus Müller Mar 19 '17 at 19:35
  • 1
    \$\begingroup\$ You can use a breakout board, make a quickie PCB, or even use dead bug prototyping (easy with SOIC not so easy with smaller packages), as @Marcus says. Or buy eval boards, sometimes they are reasonably priced. \$\endgroup\$ – Spehro Pefhany Mar 19 '17 at 19:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.