jockeymonto wrote on Jan 29th, 2014, 2:05am:Thanks RobG. Well i agree that the output of any ADC is (Vin/Vref)2^N - 1 but there are situations inwhich we can use the differential input voltages to make Vref (like BGR circuit in BJT based temperature sensor) instead of allocating dedicated hardware for Vref. Any thoughts on how ratio-metric ADCs differ from conventional ADCs?
OK, you seem to be saying doing it without an explicit voltage reference. If you mean injecting a reference
charge derived from BJTs (or a single BJT with different currents at different times), then yes I've seen that done in a pipeline (or was it a delta-sigma?). Might be impractical for a SAR since IIRC a practical design required two caps with a ratio on the order of 8:1, the 8x device switching in a PTAT charge and the 1x device switching in the Vbe. If you aren't talking about that then I don't know
.
By ratio-metric I assume you mean the circuit loose-electron talked about - I've never seen it but would be interested in learning more. Seems like a way to divide two signals.