What is an instrumentation amplifier and why do we need them? On the face of it they just look like a differential amplifier using more parts than necessary. For example, this is the “classic” three opamp instrumentation amplifier:
So why not use a differential amplifier and cut out the first stage?
An instrumentation amplifier is normally used where you have a small differential signal which is probably buried in a large common mode signal, often from a fairly high impedance source. An ECG (electrocardiogram) for example, will have very small signals from the heart between electrodes with a large signal picked up from 60Hz or 50Hz mains electricity and other interference sources nearby. So, high input impedance, high gain and high common mode rejection ratio (CMRR) are what is required. A conventional differential amplifier fails on the first hurdle – high input impedance. While additional opamps could be added to fix the input impedance problem, that doesn’t solve some of the other issues.
This is one possible solution to the low input impedance of the single opamp differential amplifier – simply buffering the two inputs. Opamp part numbers shown are simply for drawing purposes rather than a recommendation of “best” choice for a particular application.
One problem with just buffering the inputs to a differential amplifier like that is the CMRR is dependent on precise resistor matching. With a 1% resistor mismatch in one resistor your 100dB CMRR will reduce to 40dB. With 0.1% it will be 60dB.
With the “classic” three opamp instrumentation amplifier shown at the top, the two opamps forming the gain stage (R5/R6/R7) do not require precise resistor matching for high CMRR. The first stage provides differential gain with unity gain for the common mode signal so the CMRR only needs to be good on the output stage. The subsequent resistors (R1/R2/R3/R4) do require good matching for high CMRR though. This seems to disagree with some texts which suggest that the resistors in the output stage of a classic three opamp instrumentation amplifier to not need to be precisely matched but a quick simulation will show that a good CMRR is reduced to just 46dB when one resistor has a 1% error.
Instrumentation opamps are usually bought as a single device rather than made from individual opamps and resistors. In that way the CMRR is defined and the problem of resistor matching is taken care of for you (usually by laser trimming the critical resistors in production). Also, the two resistors R6/R7 for gain setting are usually internal with only a single external resistor R5 needed to set the gain. Other versions have all resistors internal with several values for R5 selectable from input pins to give a selection of available gains with no external resistors at all.
As an example, an instrumentation amplifier such as the INA121 from Texas Instruments will give a gain of 1 to 10,000 using a single external resistor and a CMRR of up to 106dB. However, the CMRR will be lower at low gains. Also bear in mind that the CMRR starts to drop at quite low frequencies – well under 1kHz for the INA121. So at 10kHz and a gain of unity the INA121 will only have a CMRR of around 50dB.
A configuration that is useful for applications where CMRR is not critical but cost is important is the two opamp version below.
While you can get good CMRR from it, you will need precise resistor matching. A 1% error in one resistor will reduce the CMRR to 40dB. It is a useful circuit though where you want a simple differential amplifier with a high input impedance without the expense (and performance) of a true instrumentation amplifier. One inconvenience of this configuration is that changing the gain requires two resistors to be changed whereas the classic instrumentation amplifier only needs a single resistor change for a gain change.
chami says
You mentioned in passing the ECG signals: does either configuration serve to amplify ECG signals well? How about EEG? I think they are in microvolts range but I do not know the frequency.