Mohamed,
send to me datasheet or link for loadcell(i didnt find with Google).
Scales , which our company produces, are typically
"legal-for-trade". We use strain gauge loadcells, which
have about 3000...6000 reliable divisions on full scale.
In Your case this is equal to 600g/6000=0.1g .
For technical purposes we done some small scales 1500g/0.1g.
They are checked/passed with milligrams on some points:
100g,200g,500g,1000g and results are ok.
But i can't guarantee linearity on whole range. On other hand loadcell has hysteresis and nonlinearity, which maybe moves real precision back to 6000 divisions(o.25g...0.5g). So we have quaranteed 6000 divisions (12.5..13 bits).

Let me say again - here we have 2 problems - mechanical precision and electronic precision. I believe You can get
60000 divisions from electronic part and 6000..15000 divisions
from mechanical part(for example Vishay Tedea lowcost loadcell 1022 has combined error 0.2% by datasheet).
How to get 60000 divisions from mechanical part i dont know-
some producers include ADC and interpolation firmware into
loadcell and give this way about 10000...20000 reliable divisions.

About boards: we use 2-layer boards, AD7730 , 5V AC excitation,
1/3 divider on reference voltage.ADC runs on 57Hz samplerate,
microcontroller uses digital low-pass filter ( 3..4 order), implemented with floating point math. Microcontroller is ultra low power MSP430 .
regards,
Stefan
stefan63@abv.bg