ADC Noise when using Sampler Class

Hardware
21

Hi All
I’m looking at the Imp for a new project. Fascinated by the idea of WiFi in an SD card and all the other concepts.
To validate if the Imp might work for me, I’m looking at checking the stability of the ADC - probably to 0.05% or approx 11bits. The circuit total is to 0.1% - so the ADC of 12bits with low circuit noise and Vref stability has worked for me on other parts.
I’m looking at taking samples every 15minutes with short term stability in hours and longer term stability for days - both as voltage and also current 4-20mA. With the 4-20mA I’m looking for 0.1% or at least 10bits which means in practise 11bits. At the same time for the short term stability its quite feasible the device temperature could shift 10C.

It looks like the processor is being powered down in the wait cycle.
Is that guaranteed that nothing else can interrupt ?
hardware.sampler.start();
imp.wakeup(2, StopSampler);

For testing, seems like I should ensure a board with out a switcher on it. (eg SparkFun April-v11 with direct power)
I am planning on using LiFePO4 batteries which are nominal 3.2V (actual 2.9-3.6V) as its solar powered.

The library references to the ADC signal are ratio-metric - proportional to the Vcc, but there is also an option for measuring the Vcc voltage.
How do I figure out the specified accuracy - are these all as per the STM32F20 manual + onboard resistors? If so can you say what accuracy are the resistors and temperature coefficient ppm.
http://devwiki.electricimp.com/doku.php?id=electricimpapi:pin:configure
http://devwiki.electricimp.com/doku.php?id=example:adc

Also the spin specification seems to suggest all pins can be allocated as ADC for card, (and 8pins for the module) but the STM32F20 only has 3 ADCs. Have I missed something?
http://devwiki.electricimp.com/doku.php?id=imppinmux

many thanks

22

The imp is an interesting concept, but unfortunately, in the effort to incorporate as much functionality as possible, certain aspects have suffered. One in particular is the ADC performance. The noise performance is fine for backyard hobby projects, but in a low-noise, real world situation, you’ll need an external device with proper layout techniques - stable reference, filtered supplies, stitched ground planes, temperature compensation, shielding etc. The module version (unlike the SD card) has the provision for an external reference, which should help in reducing the noise, provided it’s implemented accordingly. You’ll still only have a maximum of 12 bits though. Depending on your desired sample rate, it is possible to oversample at a much higher rate, then average an entire buffers worth of data to provide a single reading. This works nicely, and provides a significant reduction in noise, but with the added overhead of software complexity. All I can suggest is that you try it, and see if it’s acceptable for your needs.