Hi there,
we are measuring from a quadrature encoder the raw sine and cosine and need to extract the angular position [1]. The position is then fedback to a PID which drives a motor.
We've been told that dither may be efficient in removing spurious harmonics due to quantization errors which may affect the control loop only few days before pcb production, therefore the only possibility was to add few resistors from the FPGA to Vref used in the opamp which feeds the ADC. Noise generated from the 4 toggling pins of the FPGA will be injected right at the input of our ADC.
We met the deadline for the pcb production (great!), but unfortunately we just postpone the issue and moved it into the FPGA (imagine how happy was the fpga team!).
Now I have the following questions:
- how do I prove I need a dither generator? Indeed we didn't have even the time to figure out if we really need it. Is there a dedicated test I can perform which may undoubtly confirm the need?
- how the dither generator noise spectrum should look like? I've read about both in-band and out of band dither but I'm still confused.
- any reference for such generators on FPGA? A long LFSR (say 64bit) and only 4 pin used might seem silly, but is simple. OTOH I read LFSR do not provide random enough patterns therefore introducing other issues.
Any help is appreciated.
Al
[1] by tracking demodulation.