Software based test for ADC

You could read a known voltage.

Normally, you arrange a known voltage(s) onto the ADC IPs. If you have any PWM channels, they can be good for this, as they give very precise Vcc ratio test levels thru a LPF. For a less precise test, but quicker test, a slowish ramp into the ADC, coupled with timed-samples, could be used.

-jg

Nutty posted:

"[..] Now as a part of this project, i m working on the Power ON Self diagnostic test for all the major components. In our project the ADC is not yet finalised but it is 16 bit SAR type. Is there any standard procedure to test the functionality of ADC, like memory testing. Could any one please refer some information in regards."

Hello,

In general this can not be accomplished, especially in the case of a self diagnostic test in a deployed product which you do not have access to. For example, how can the software control (and therefore know the true value of) whatever physical quantity which the ADC is trying to measure? If the true value is not known, the value reported by the ADC can not in general be validated.

After you have finalized which ADC you will use, your software may be able to perform some checks without human intervention, for example if the manufacturer has documented that the ADC will take at most a predetermined amount of time after being powered on to function correctly, and if the manufacturer has documented that some reply (unfortunately not actually the analog to digital conversion readings) pins will behave in a particular way in response to particular commands from the computer (for example command it to change some parameters and afterwards read its configuration to check whether those parameters are reported back), you could try something but testing whether analog to digital conversions actually work would not be possible.

Manually test your product's ADC (e.g. by providing it with input which you vary with a potentiometer or if your product's ADC will be connected to for example a speedometer, move it at known speeds) before releasing the product to the customer.

Regards, Colin Paul Gloster

In some data logging applications, I have been able to get a startup confidence check on the ADC by connecting one of the multi-channel ADC inputs to a resistor-divider connected to an IO pin on the MPU. You set the I/O pin high and measure the signal with the ADC. Then set the pin low and measure the signal again. If the two values are within your confidence limits, you can assume that the ADC is working.

If you have just a single-channel ADC, you would have to add a multiplexer on the input to use this technique, and that has other implications for the stability and accuracy of the ADC data.

Mark Borgerson

None that I know of, the analog world is much too diverse to standardise. One has to figure out whether and how to do it on a per case basis. For example, in a medical device I recently designed I had both control over a synthesized analog output and the ADC readback of it; that makes it easy. In other devices, where the input is unknown and switching input to some inbuilt known source would be too costly because of technical or financial reasons - or both - testing may be just impossible, or hard to do, like observing some sane behaviour pattern etc.

Dimiter

------------------------------------------------------ Dimiter Popoff Transgalactic Instruments

------------------------------------------------------

While the decisions about the product design are still in the melting pot it would be worth working through what sort of tests you would be satisfied with and ensuring that the hardware support for those tests were designed in.

I would suggest having perhaps a zero and span check where you switch in the appropriate levels while doing the data read check. You might have to fight for a know mid value check. This is not an exhaustive check and, with a multiplying DAC you could even self calibrate the input.