1
\$\begingroup\$

I need to build a system to monitor several (around ~20) voltages over several days. The ~20 voltages need to be measured differentially. Something which could in theory be accomplishes with 3 of these picologgers.

I was wondering, what is the state of the art way to do something like that?. Do the fine people of Picotech use: a) 8 differential voltages → 8 to 1 Mux → 1 Differential ADC → µController, or b) 8 differential voltages → 8 Differential ADC in parallel → 8 GPIOs in µController ?

Option a) is cheaper but the 8 to 1 Mux needs to be very fast to cycle thourgh 8 channels and not degrade performance. Do they use CMOS Mux? Also it should be a differential mux.What do you guys think?

\$\endgroup\$
8
  • 2
    \$\begingroup\$ how often do you need to sample/what's the signal bandwidth? What's your signal's dynamic range? What is the resolution / number of bits you need? \$\endgroup\$ Apr 17, 2020 at 20:01
  • 2
    \$\begingroup\$ Because, honestly, 24 bits resolution is a lot, and it's likely the effort to make that work is not only in the ADC architecture, but very much in the signal conditioning leading up to that. When you look at the speciifcations page, you'll see they only manage a limited SFDR/ an ENOB of less than 20 bits, even for their 24bit ADC version. And Pico have been making digital measurement devices quite some time – high resolution measurement is hard, and you'd want to buy if possible. If, on the other hand, you can do with significantly less bits, solutions much easier might be appropriate. \$\endgroup\$ Apr 17, 2020 at 20:09
  • 1
    \$\begingroup\$ Hi Ken, then that does sound like you might want to do something like simply using multiple microcontrollers, each with their own built-in ADC muxing. If your bandwidth is limited to 1 kHz, then anything sampling each channel above 2 kSamples/s allows you to reconstruct the signal at any arbitrary point in time, so, the "sampling instants need to be aligned" gets easy, because at these rates, you can do that easily in software postprocessing. \$\endgroup\$ Apr 18, 2020 at 8:00
  • 1
    \$\begingroup\$ now, the differential signal does pose an interesting aspect, but you could just use an opamp configuration used to differentially amplify (instrumentation amp!) to convert to single ended. \$\endgroup\$ Apr 18, 2020 at 8:03
  • 1
    \$\begingroup\$ use one oscillator source for all micrcontrollers. configure one of their pins as interrupt input to kick of the main program sufficiently synchronized for you application. Go through the channel conversion sequence on each microcontroller. Later, on a PC, you'd either use fractional delay FIR filters to re-align, or just resample each channel to a higher (typically number-of-channels higher) sampling rate, and shift by an integer number of samples to align. \$\endgroup\$ Apr 18, 2020 at 9:15

1 Answer 1

2
\$\begingroup\$

In the 6.5 digit multimeter level of measurement (around 27 bits of resolution) It falls under the name "Data Acquisition Unit"

Things like the Keithley DAQ6510 which for the 80 channel (I think most of which can be measured differentially) is about $1800 US, note that in most of these systems the switching cards are actual relays, as such they do have a number of cycles before there performance degrades (something around 10 million cycles usually)

There are cheaper alternatives, but the switching will be your issue, silicon switching at that level of precision is extremely hard, non-linear with voltage and non-linear with temperature,

This is making an assumption about the resolution your after based on the devices sold in you link, if your resolution requirements are something like 12 bit at a reasonable voltage, then mux chips can come back in to play,

\$\endgroup\$
5
  • \$\begingroup\$ Thanks @Reroute. A 12 bit resolution would be OK, even 8 bit would be fine. Range is 1 to 2 V. So you think 1 mux plus 1 ADC would be the best choice? \$\endgroup\$
    – Ken Grimes
    Apr 18, 2020 at 7:01
  • 1
    \$\begingroup\$ Are these signals shared to a common reference, as in all within 1-2 volt of each other, or that they have 1-2V difference but may be on top of a larger voltage, e.g. 20V - 19V would be a 1V differential signal, but needs to be treated differently equally the differential measurements, is that 40 wires, e.g. every input has its own return path, or is it less? \$\endgroup\$
    – Reroute
    Apr 18, 2020 at 7:34
  • \$\begingroup\$ Thanks @Reroute. The common voltage is also limited to 1 to 2 V. 3 V at most. Each signal has its own return wire, so yes in total 40 wires. \$\endgroup\$
    – Ken Grimes
    Apr 18, 2020 at 8:57
  • 1
    \$\begingroup\$ Mux's are expensive, high channel ADC's are expensive, Honestly, I would throw something like 5+ ATtiny1634's at the problem, Have them spit out over TTL UART, clocked off the same clock IC so they don't drift, the internal ADC can over sample so much that the samples are effectively the same time, you would just use a timer to trigger at roughly 1000Hz, the interrupt grabs 4 samples, and spits it out to a UART to USB adaptor, probably along with something like a 32 bit time code, and a micro ID (8 bits) to make it easier to deal with dropped or delayed packets, as well as post processing. \$\endgroup\$
    – Reroute
    Apr 18, 2020 at 9:21
  • \$\begingroup\$ Or if you go with the picoscope method, the pain will be juggling the mux changing, \$\endgroup\$
    – Reroute
    Apr 18, 2020 at 9:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.