True 'continuous' ADC on a single source

[bobert]

Posted on November 04, 2014 at 06:36

Hi,

I want to sample a single source 'continuously'.  By that I mean I don't want a moment to pass without sampling being done.I read 

http://www.st.com/web/en/resource/technical/document/application_note/CD00258017.pdf

 and the dual fast interleaved moved looks closest to what I am trying to accomplish but it shows an overlap in the sampling areas.I would like to figure out how to set it up where the exact moment conversion starts on one ADC, the second ADC is sampling, and the exact moment conversion starts on that ADC, the first one is sampling, with no overlapped sampling of the source, and no gaps between sampling being done on the source.Is it possible to do this?Thanks,Robert #adc_dual #nyquist? #adc #dma

[ivani]

Posted on November 04, 2014 at 21:21 If you are using STM32F1xx then you could combine slow interleaved mode with sampling time of 13.5 cycles. Then you will have only 0.5 cycle gap every 14 cycles. With STM32F4xx the gap between sampling intervals could be variable but it can't be less than 2 cycles. So, the minimal gap will be 2 cycles. But, the question is do you really need such a ''continuous'' sampling? You must take care to provide a proper sampling rate according to signal spectrum and sufficient sampling intervals depending on signal source impedance but the gap between sampling intervals is not important at all.

[bobert]

Posted on November 04, 2014 at 22:36

I am trying to get the exact average of the source over time.  It doesn't matter how long the sample timing is, nor my sample frequency.  It could be uS or mS or just seconds... it does not really matter to me.  I just want it to be exact.

Example...  air flow in a tube.Lets say the sample time is 1ms.  If for 0.5ms the flow is 20 units, and for the next 0.5ms the flow is 0, I expect to get back a flow of 10 units for that 1ms sample time.  By continuously sampling I won't miss a thing.  I don't care about the frequency components of the source signal, only that I have an exact average of the data, for every moment it is happening.  Higher frequency would be nice, but not important.Does this make sense as to why I don't want any overlap or gaps?

[raptorhal2]

Posted on November 04, 2014 at 23:00

The sampling is not an averaging process. During the sampling interval, a capacitor is charged to some acceptably close voltage to the input voltage, then the capacitor is disconnected from the signal and the ADC converts it. So you are getting the value at the end of the sample period, not an average.

The signal source example provided has a max frequency of interest much lower than the ADC conversion frequency. Numerically averaging the converted values is probably what is needed. Cheers, Hal

[Tesla DeLorean]

Posted on November 04, 2014 at 23:04

Does this make sense as to why I don't want any overlap or gaps?

No not really, you aren't really taking instantaneous measurements, you're charging a capacitor. If the input voltage is moving throughout the sample period, you'll see it trying to follow as fast as the RC constant will permit it. Really it just sounds like you need to sample faster than the input is changing, and have a low pass filter to enforce that.

[bobert]

Posted on November 05, 2014 at 00:51

Thanks guys.  I misunderstood how the ADC worked.  I figured it was some sort of capacitance charge but I also figured it would deplete in the absence of the signal, leaving you with an average for that time.