0

I want to know about the possible solutions to achieve precise automatic gain correction of the difference amplifier ?

Description: I am using a differential amplifier and all resistors have 1% tolerance according to specs. After some time the gain changes due to a change in resistance (aging effect/tempco etc). Therefore, I need an automatic gain compensation. There are digital potentiometers available, but they are in the range of 5 to 100 kOhms which do not fit for my case because my feedback (R3) resistor is 1 kOhm. If I use multiple digital-pots, I will not be able to achieve the desired resolution of less than 0.5 Ohm. Here is a description of the difference amplifier.

Which solution would you suggest?

Khalil
  • 1
  • 2
  • 1
    What is your question? – Kartman Aug 13 '21 at 06:36
  • 5k-100k pot is perfect for 'trimming' 1k resistors to a few percent. You certainly don't want to realise the entire resistance as a digipot! – Neil_UK Aug 13 '21 at 07:09
  • Use much better quality resistors. – Andy aka Aug 13 '21 at 08:09
  • Khalil, How do you expect to get "automatic gain compensation" without having a reference known to be "true?" How would you measure gain-drift, exactly? Think about this. Do you expect to insert a "magic crystal ball" into your circuit? Or do you imagine that perhaps you'd need something *even better* in order to observe any drift? Not only would you need better resistors but you will also need a better set of ICs, I suspect. To achieve better, the *entire system* needs to be examined, carefully, and an eye to absolute drift over time, temp, etc. This is not a "stick this here" question. – jonk Aug 13 '21 at 08:52
  • what jonk says is very relevant, Khalil. there are autocompensation schemes that do not need references (e.g. zero-drift inputs) but these use switches and your post doesn't appear to go into that direction. – tobalt Aug 13 '21 at 09:31
  • thanks for your response everyone, @jonk I want use ADC to measure the voltage level and compare it to the calibrated ADC value (reference which was set previously) to measure the gain-drift. then I want to compensate the gain, since 1% tolerance results in 200mV change, I want to compensate this change automatically. for this one possible solution would be to use digi-pots but their resolution is not feasible to achieve this accuracy. – Khalil Aug 13 '21 at 09:56
  • Kartman, I have written the question. – Khalil Aug 13 '21 at 10:08
  • Andy aka, Using precise resistor won't solve the problem for long time. – Khalil Aug 13 '21 at 10:13
  • "my feedback (R3) resistor is 1 kOhm" The most obvious solution is to change he resistance values in your diff amp. Is there any particular reason that your current values are sacred? – WhatRoughBeast Aug 13 '21 at 13:34
  • You calibrate the system with a known input and measure the output. – Voltage Spike Aug 24 '21 at 22:48
  • A suggestion for a valuable solution to this specific problem involves doing a lot of design work for you. That is something most people here don't want to do. You should break down your problem in smaller problems. This will help you as well to better understand what to do and you will get better answers. – Ariser Aug 26 '21 at 15:12

3 Answers3

2

In my own DIY-level reading and playing with pencil on paper, I've met a related problem a while ago. Precise analog gain control - the goal being to provide high-linearity AGC at a medium-wave frequency range. I understand that yours is more like DC or audio, which should make it easier for you to find a solution.

If you try googling about the topic, you will find references to JFET's and their possible use for gain control. JFET's do have a so called linear region in their total range of operation, so the catch is to fit within that relatively small region in the JFET, which isn't very easy or practical AFAICT, for the general case (about +/-50 mV of Vds if memory serves). But it might as well suit your needs, if your desired adjustment range isn't very broad, and you could wire the JFET in your feedback network such that it would never see the full amplitude at the op-amps (-) pin for instance, if that's a problem at all. Modulo the JFET's peculiar transfer curve, with a gate threshold voltage that's underground for the N-polarity device etc.

The other part that you're inevitably going to learn about, are analog multipliers, almost synonymous with the Gilbert cell. There are chips for that job, similar in their properties to operational amplifiers - as to the supply voltages, differential inputs and outputs including their typical impedances, some (not all) have broad input ranges etc. I suggest that you take a look at the AD534 / AD734 / AD834. The 534/734 have a single output of the "voltage" type and apparently full-range op-amp-like inputs. The AD834 features radically broader bandwidth, its input is limited to about +/- 1V differential (common mode range is wider) and has a balanced, current-coupled output. I'd like to recommend two appnotes by AD. You may also like this report of a DIY multiplier built out of a quad op-amp, four matched schottky diodes and a handful of precision resistors.

frr
  • 2,573
  • 12
  • 14
2

When you have a fixed gain and require better than 1% accuracy, you do a tolerance stackup of all parts from all environmental stresses and define a specification. You may consider an auto-calibration method if you have a reference that is better than your spec. Then you can consider how to compensate for this error by software gain or hardware changes.

0.05% tolerance, 5 ppm TCR

But generally the normal solution is to use resistor arrays, which are far more accurate (due to laser ratiometric trim-processing) than I expect your (unstated ) requirements.

for example for example

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
1

Please provide a schematic of the circuit you have actually built. Go into Edit of your question, then hit ctrl-m to bring up the schematic editor. Make sure to specify the resistor values you are using, along with the op amp models. Also add the input and output voltage ranges you are using.

Then, edit your question to provide some measurements you have taken, and discuss WHAT problem you are seeing (numbers, please!) and WHY you think that resistor drift is the culprit. Trust me, if you're seeing 1% changes in 1% resistors, you are abusing your components. Among other things, temperature changes that big will probably kill your op amp. Not only that, but if your resistors are near your op amp, they will probably be at the same temperature (more or less), so drifts will tend to cancel out.

I suspect that you are asking what is called an "xy question". That is, you have a problem x, and are trying to fix y, when y is not the problem.

Based on what information you provide, and more is better in this case, I'll edit this question to provide what help I can.

WhatRoughBeast
  • 59,978
  • 2
  • 37
  • 97