ADXRS623 Null

Based on a careful review of the ADXRS623 datasheet, what is your expectation for bias accuracy and noise?  Also, I am afraid that I do not understand why you believe that a 5mV change in bias/offset causes >200% change in scale factor.  That seem like quite a leap to me.  From my point of view, the 5mV change in bias should have ZERO influence on the scale factors, as they are independent behaviors.

Thanks for replay.

yes you right it does not effect on SF. But 5mv*2.5=Scale Factor=1deg. so that means a serious problem for my application. 

If you paid attention to my Design, i just used RateOut and SUMJ differential to adjust the Nul Voltage to 0 V. That's why the 5 mv cause an error.

And if you please answer the question that; Is it normal or not(Null Voltage error) ? what can be solution(to fix Null voltage)?

Looking forward to hearing from you.

I fear that we are digressing in this discussion.  Let me take another shot at this. 

Q: Is 5mV of offset normal? 

A: Figure 4 in the ADXRS623 would suggest that this is well within the normal distribution.  

(more below plot)

Q: Did I look at your schematic?

A: Yes, but only briefly, because it didn't seem to relate to your question and I needed clarification on your scale vs offset statements.  To be honest, I am not sure that I understand why you are using the differential between SUMJ and RATEOUT to derive your bias adjustment.  Most applications run this into a Ratiometric ADC interface and calibrate the behaviors in their digital processing routines.  That doesn't mean that your method will not work, but it does mean that it is outside of what I given consideration to, so I am not in a position to offer judgement on how effective this method can be.

Q: How do we solve this? 

A: Solve what?  Zero error is not possible, so I am not sure what to offer at this point.  What level of error is acceptable?  Can you application observe and zero out bias on a regular basis, perhaps when it is on the ground?

I hope that this helps get us started.

Hi and thanks for replay. That helps a lot.

I have some limitation on size so it is not possible to use digital processors.

Q:What level of error is acceptable?

A: because of application (Platform stabilizing for Quad) the error should be least.

Q: Can you observer a zero out bias on ground?

A:No! it has -8mv to 8mv float out put.

"Least" is a great value, but is not very useful as a goal. How much inaccuracy is acceptable?

+/-8mV seems to be an appropriate level of noise, for your current circuit. Here is my hand-drawn calculations..

max 3mV Peak to peak !!!

Is 3mv peak to peak is possible?

Yes, but you would need to lower the bandwidth. If you did that, would you still be able to provide adequate responsiveness for your controllers? If that is a valid consideration, the hand-drawn notes that I posted would provide a guideline for that trade off.

Thank you so much. that helps so much.