4
\$\begingroup\$

I'm looking into isolating an analogue input coming from field equipment and interfacing to a PLC analogue input module. The input will be of the 4-20mA range (1-5V).

There are a few ways to approach this but I am currently pursuing the solution as seen below:

Proposed solution to analogue input isolation

ADC and DAC are being proposed as Microchip MCP3221 and MCP4725 respectively.

A digital isolator would give a smaller footprint than an opto-isolator with its accompanying components and would provide much greater performance. Possible components include TI ISO7420FEDR and Analog Devices ADUM1200ARZ.

Now for the actual question! Keeping both sides isolated is pretty vital and I'm wondering how this will effect the ADC and DAC linking.

I'm assuming I can send the data line (SDA) through the digital isolator but to synchronise the ADC and DAC they would need a common clock signal. Would sending the clock over the digital isolator be a bad idea? If the digital isolator was many times faster than the clock would it have a negligible effect in terms of delay?

Missing info:

  • Not tied to I2C, SPI would be appropriate too
  • Isolation to a minimum of 2.5kV
  • The ADC will be powered on the 'dirty' side so there is complete isolation, only the data/conversion control signals need to pass through the isolation barrier
  • Resolution should be between 10 and 16 bit, no specific value is required
  • Conversion throughput of around 100ksps would be sufficient
\$\endgroup\$
4
  • \$\begingroup\$ How much isolation? i.e. what stand off voltages must you be able to support. It can't be too much as you have to supply power to the ADC which presumably will be closer to the DAC power supply than the original signal. Signalling rate? bit depth? Etc. etc. you are lacking some very key information here. \$\endgroup\$ Jul 16, 2014 at 14:52
  • 1
    \$\begingroup\$ Beware that if your "digital isolator" is a sampled device which introduces jitter, it may introduce noise into the conversion of devices where that is driven by the data clock, though perhaps not enough to matter in your application. Have you considered using voltage-to-frequency and frequency-to-voltage converters? \$\endgroup\$ Jul 16, 2014 at 15:21
  • \$\begingroup\$ Have a look here: silabs.com/products/power/isolators/Pages/… \$\endgroup\$
    – user25093
    Jul 16, 2014 at 15:33
  • 1
    \$\begingroup\$ You can check this link, I asked similar question some 9 months ago. electronics.stackexchange.com/questions/84091/… \$\endgroup\$
    – AKR
    Jul 16, 2014 at 16:15

2 Answers 2

3
\$\begingroup\$

Frankly, SPI interfaces are better suited for isolation than I2C. You can set the master clock to a frequency where the timing works out reliably. I2C is a bidirectional bus which makes isolation a pain.

If you insist on using isolated I2C, AN-913 from Analog Devices is an application note that illustrates how:

\$\endgroup\$
0
\$\begingroup\$

If I needed an analogue signal isolator like this I'd choose a serial output ADC that could run continuously, possibly from a local clock generator that also generated a conversion pulse. The output I would aim to get is: -

< 12 bit conversion >< gap >< 12 bit conversion >< gap >< 12 bit conversion >

This could be transmitted thru a simple magnetic coupler (as per a lot of the ADI isolation devices) and you could decode the serial stream using the "gap" like the stop bits in an async serial stream with a small cheap microprocessor. This would then feed a DAC.

I think that trying to do it without a micro controlling to the DAC could hit problems. At least start off with a methodology that stands a chance of working and is easily modified to suit different ADCs and DACs.

I'd also be interested in the isolated ADC part driving a decent length cable so the DAC part can sit at some distance to it - this would allow you to digitize at source rather than degrade the analogue signal by having this transmitted to the "converter".

Just a few ideas.

\$\endgroup\$
2
  • \$\begingroup\$ I suppose the microprocessor would only need the most basic of functionality to control the conversion so it's definitely an option. It would make this a lot easier but the requirement for a very small footprint means that IC count needs to be kept pretty low. It's definitely something to keep in mind though. \$\endgroup\$
    – JoshGreen1
    Jul 16, 2014 at 19:55
  • \$\begingroup\$ @JoshGreen1 you'd be surprised what can be crammed into a small place - my company make a 16ch, 16 bit, 200kSps DAQ module that transmits serially up to 300m and measures 4" x 1" x 0.5" \$\endgroup\$
    – Andy aka
    Jul 16, 2014 at 20:17

Your Answer

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

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