Post by thread:resolution /was: low ext. Vref for A/D?

At 03:59 PM 7/18/97 -0400, you wrote: >I'm not familiar with the 16C72 but I am familiar with analog to digital >conversion. The fact that your reference is different from VCC does not >reduce the resolution of the ADC. In other words you'll still get 8 bits. > > if VREF = 5 volts your resolution will be 5/256 = 19.5 mv per count > if VREF = 3.3 volts your resolution will be 3.3/256 = 12.89 mv per count > > just scale the input so it ranges from 0V (or something slightly greater >than 0V) to whatever your reference is. > > Right, and thanks for the fast response, Microchip specifies Vref from 3.0V to Vdd+0.3V. But will I still get acceptable results at 3.3V. Or can I forget about the LSB (and maybe Bit2!)? I think turning Vref won't affect the number of bits I get, but how many of them will be useless to me? Maybe somebody has already experience with this, and can tell me that even under good (unnoisy) conditions I will loose the lower X bits when my Vref gets under X.X volts or something like that? Thanks again, Franky X. ----------------------------------------------- Man is always a victim of his own truth - Camus

Perhaps it would help if you told me a little about the signal you are converting. Its hard to tell how many bits can be lost without knowing the application. Its true that your signal to noise ratio (SNR) is less with a 3.3Volt reference than with a 5Volt reference. You have to look at the resolution of the parameter your are measuring to determine if the loss of an LSB is tolerable. For example if you are measuring pressure and the pressure range is 0 to 100 cmH2O you would amplify the transducer output so that it was = to VREF at 100 cm of pressure. Your resolution would be 100cmH2O VREF --------------- * --------------------- = .39cmH2O per count VREF 256 counts You have to know the system requirements to determine if loosing .39cmH2O (1 LSB) is significant. If you are worried about white noise being larger than 0.39cmH2O you can take the average of multiple samples (if you have enough processing time). This will do to things, 1, it will remove noise and 2 it will increase your resolution. I don't know if I'm helping or not so let me know. Good luck...