求助 梯度误差对SFDR的影响

fhchen2002

The idea is like the following:

SFDR should be examined by FFT (i.e., signal power over the next largest power of in the FFT bins).
For ideal N-bit DAC (un-biased, no mistmach between unit cells, without process gradient effects), the FSDR is ~ 8.2 N + 2.2 dBc.
Now simply model your unit-cells with process gradient in X and in Y directions.
You get bit-by-bit DAC total output based on the sum of the (biased) unit cells.

Stimulate the non-ideal DAC (modeled by MATLAB script, with process gradient) with sinasoidal "digital" input (an ideal analog input going through an ideal N-bit ADC reaching just the full-scale of the ADC), then you get non-ideal DAC outputs with the effect of gradients.  Send your DAC output to FFT routine to get the SFDR.

You should find that when the gradient is small, the SFDR examined by the FFT routine is close to 8.2 N + 2.2.  As the gradient gets larger and larger, the SFDR starts to degrade.