Dear Forum Members,
Hope all of you are enjoying the weekend.
This post is a bit lengthy but informative and highlights the practical analog signal conditioning problems and some remedies to deal with. I have mentioned some problems [solution to which I dont know] for our kind Analog Gurus to analyse and share their techniques of dealing with them. I apologise for such a lengthy post kindly read with patience.

With mixed signal system designs we often make use of signal conditioning of analog signals. For this purpose we use Op-Amps and In-Amps etc. There are several errors introduced in the signal conditioning process we can broadly classify these errors into following categories.

1. Absolute Errors.
    a. Input offset Voltage.
    b. Output offset Voltage.
    c. Input offset current.
    d. CMR dB.
2. Drift Errors.
    a. Input offset voltage drift.
    b. Output offset voltage drift.
    c. Gain drift.
3. Gain Nonlinearity.
4. Error due to passive component tolerances.

1. For nullifying the drift errors the Op-Amps and In-Amps have offset adjustment pins and using a pot there the offset can be made zero or near zero.

2. Drift is usually very small few micro volts per degree C and can be left uncompensented in smaller resolution applications but if its really matters than the tempreature compensation can be made by measuring the actual temperature using some temp. measurement chip on the system. [ correct me if I am wrong not very sure about this.

3. Another way of correcting for Drift + Offset is to connect the input of conditioning element to ground or 0V and measure the output this output is exact measurement of Drift + Offset. Then measure the Signal which will give output including drift and offset. Then in software the drift + offset measured earlier can be subtracted to nullify drift and offset. [ I request members to contribute their methods of connecting the input to ground and signal alternately do you use MUX or mechanical swithcing element such as relay etc.?? ]

4. The gain linearity errors of OpAmp and InAmps seems to be the factors against which hardly anything can be done but they are very minute. [ I request the members to please contribute the methods for this correction if they have done this ].

5. The last factor is the accuracy of the passive components such as Resistors Capacitors etc. used with these conditioners. I agree in some applicaion designers dont use passive components and directly make use of In-Amp [ suh as INA1xx AD62x etc ] having built in laser trimmed resistors and get rid of this problems. but some times the application is such that it is absolutely necessary to use a conditioner involving passive componenets. The Real problem here is the maximum accuracy components that could be fetched easily is 0.1% resistors [ MFR ] and 1% tolerance Polypropelene Capacitors. If the component tolerances are so high [ 0.1% ] its difficult to achieve even 9 bit resolution lets take the simplest hypothetical example of Inverting amplifier with gain = -2. The resistors used are Rf= 20k and Rg= 10k both having 0.1% tolerance. Lets say the 20k resistor has +0.1% maximum error thus the true value is 20.02k and the 10k resistor has -0.1% maximum error thus the true value is 9.99k thus the practical gain will be 2.004004. Lets say a signal of 2V is applied at input.

      Output      Output    ADC output   ADC Output    Error
    Theoretical   Actual    Theoretical    Actual   

8bit    4.00V    4.008008      204      204.4 := 204     0 LSB

10bit   4.00V    4.008008      820      821.6 := 822     2 LSB
12bit   4.00V    4.008008      3276        3283          7 LSB  

Vref=5V in all cases.
From above table it is clear that even with using so high prcecision components the signal conditioning accruacy attained [ Ignoring Drift + offset + Non Linearity ] its not even sufficient for a 9 bit system. How to achieve better accuracy?? In cases where the monolithic In-Amps and Diff-Amps is not a solution.

Thanks & Regards,
Prahlad Purohit