I use ad620 for my load cell conditioning circuit. It amplifies load cell signal to +/- 10V range. I need to improve drift cause by temperature. When I apply just a little hot air to my pcb and output changes.
I have ad620 with a gain of 660. I have +10V and -10V regulated from +/-15V. I use reference pin of ad620 in order to remove dead weight. because after ad620 I further amplify my signal with a gain of 12. then a low pass filter. My filter is around 3Hz.
I use 10ppm thin film resistor on gain adjustment and 25ppm in my filter and voltage regulator circuits.
does any one has an idea on getting better temperature performance.
I use 16bit A/D module with a range of +/-10V. 20/65535= 305mV for a decimal change.
I power my circuit in the morning where there is not much temperature change in the environment, maybe 4-5 degree change from morning to afternoon. From morning to afternoon there can be 150 decimal change at the A/D module. that means 45mV change. The change is not constant for all circuits. Some of them change 10mV some 70mV. But if you take average it is around 45mV.
Also as you have mentioned I need a controlled environment for to test temperature. But even if you have constant temperature there is change so it is easy to guess where the output will go if you change temperature 40 degrees.
For 40 degrees change I need to have maximum of 10mV change at the output.
I have limitations;
The output must swing +/-10V, upto now I boost output of AD620 with the filter so for a 10kg loadcell I can measure upto 800gr. Therefore I need a reference adjustment for AD620 or for the new instrumentation amplifier in order to remove dead weight.
Yes 305uV. It is measured at the A/D module.
150 decimal change 45mV.
It is complete analog module. Panasonic AD8VI. Used as a module in a PLC rack. I check values from HMI device.
At pages 6-3 and 12-3 there is info about module.
The typ. zero drift of AD620A is 0.3µV/°C. Gain 660 gives 200µV/°C. Then comes the output drift typ. 5µV/°C.
This 205µV/°C is amplified with 12 giving 2.5mV/°C. Over 40°C it is 100mV. Added to this the load cell will contribute with more drift. Do you have a preload on the load cell? E.g. using it between 8 to 8.8kg? Then the poor gain TC of the AD620A, 50ppm/°C will add a lot more drift. 16mV*50ppm/C is 0.8 µV/C, added to the 0.3µV/°C on the first row.
You can measure the drift of each unit and program a D/A-converter to drive the reference pin to balance the drift as a function of temperature or use another amp. A quick search of analog's Inamps led to AD8228. It can be set to 100V/V gain and then you have to increase the following gain from 12 to 79.
To get low gain drift, select among inamps using only internal resistors for gain selection.
I use my load cell between 3kg to 3.8kg basically. I did not get the following
"Then the poor gain TC of the AD620A, 50ppm/°C will add a lot more drift. 16mV*50ppm/C is 0.8 µV/C, added to the 0.3µV/°C on the first row."
But there is some questions about your statement. How can you explain some circuits have nearly zero drift and some has much more.
In my application I have nearly 15 circuits working at the same time. you power on the system wait about 10-15 minutes in order to build temperature of IC's. And zero all circuits. Then start to wait and see the drift. You will see some circuits output becomes +1gr some is 0.5gr some is -0.5gr some is -2gr. so their drift is not similar.
How about effect of power supply voltage. I have 10V reference IC suppose its output changes 10ppm/C or my negative regulator has 50ppm/C.
Will drift of a circuit always be the same. Suppose;
temperature is 20 degrees. you power up system,make zero the output and wait 3 hours. output becomes 1gr. one day later again you make same measurement at same temperature. will output be same?
Let me summerize it
FOR AD620 or any in-amp there are 2 sources for the drift.
1-) Voltage Offset Drift
a) input offset drift
b) output offset drift
2-) Gain Drift
Input offset is 0.3uV/C with 660 gain 198uV/C
output offset is 5uV/C
Totally 203uV/C. My output stage gain is 12. Therefore total voltage drift is 2.436mV/C
Since there is 3kg dead weight on my load cell. 3kgx10Vx2mV/V = 6mV. Gain drift is -50ppm so; -0.3uV/C. with a gain of 660 => -198uV/C And 12 filter gain. -2.376mV/C
For power supply;
I only use 10V for loadcell So I think TC of positive supply is more important. negative supply is just for ad620 and op amps. can we make average of it like you did?
about +/- signs of errors. Gain is always ( - ) voltage offset can be +/- 2.436mV/C and power supply can be +/-. With a good combination you can have very low drift. This will explain that some of my cşrcuşts has nearly zero drift.
Do you say that
1-) an in-amp with a gain of 1000 and plus a filter with gain 5 is same as
2-) an in-amp with a gain of 50 and plus a filter with a gain 100