Electrical – ny difference between AD620 vs INA128

You don't need a dual supply, but you'll need more design work than you seem aware of. You can use the AD620 with a single 5 volt supply since the load cell outputs ought to be about 2.5 volts and are within input limits. Set the reference to 1.1 volts. The ouput will then be restricted (according to the data sheet - and you DO have a copy of the data sheet and you HAVE tried to understand it, right?) to the range 1.1 to 3.7 volts. If you're willing to deal with 200 gram resolution you can stop right there.

If not, you need a second amplifier / offset generator. But first you would set the AD620 reference to something useful like 1.5 volts using a pair of resistors, and the gain for a 2 volt swing. Then, using a rail-to-rail op amp you would make a gain of 2.5 difference amplifier, and add in a 1.5 volt offset. The result would be (nominally) a 0 to 5 volt output. And before you ask - no, I won't give you a design. Do some research first. Google "differential amplifier" and actually try to understand what you read. Make a preliminary design and mess around with it. If you can't get it to work, describe your problem in detail on another question. Folks, including me, are glad to help - but I'm not going to do your work for you.

There is another consideration. First, you want to set your zero point at very slightly above zero volts. You should be able to check your data at zero load and see a little bit of noise. If the output of the amplifier is dead zero you know that your zero load level is less than 0 volts, but you don't know how much less. You can always subtract out a known zero offset, but there is no way to recover a completely unknown quantity. Losing a little bit of dynamic range or resolution is much less of a problem than not knowing what your zero point is.

I sample 4 times per second and I repeat it for about 20 cycles for smoothing of values using moving average.

OK this is the basic information needed. To prevent the effects of noise causing aliasing you need to filter the signal before it enters the ADC at below 2 Hz. This might be best done on the final op-amp (gain stage) by using a sallen key filter: -

This is a typical circuit although the values will be different for your application. Here's the link for the circuit above - it may be useful for you. Here is a very useful link - it's a calculator and this is what the calculator looks like when you scroll down the link to the relevant part: -

You might also consider putting a capacitor across the lower of the two 10k resistors that feed the 1st AD822 - a value of 33uF creates another 1Hz low pass filter that, along with the sallen key filter gives you a third order filter. Even so, depending on the noise this may not be good enough and a final rC filter on the output of the op-amp stage converted to a sallen key filter may be needed. 10kohms and 10uF would be about right.

And, if all this seems a little over-the-top, consider sampling at a much higher rate (maybe 100 times per second) because the filters require smaller capacitors and you can average more in your MCU.

One final thought - the IA may be subject to EM Interference and this may require 1k ohms in each input leg with a 100nF across the two input pins.