1
\$\begingroup\$

I am looking to acquire data from several sensors on a machine, each of which are part of a control system featuring several grounds linked at a single central point. I need each sensor's signal at the point of measurement compared to its own ground at that point, while avoiding ground loop issues caused by the common ground in the DAQ.

The sensors operate on a 5 V supply and generate an analog output between 0-5 V, with an output from DC to 20 kHz. I would like the DAQ to log the sensor signals, while isolating grounds, with an accuracy of 0.2 %.

What component or circuit exists that performs this function?

\$\endgroup\$
8
  • 1
    \$\begingroup\$ Be specific - what DAQ are you using and what level of DC and AC accuracy are you after? It's going to be a compromise DAQ if it doesn't have proper differential inputs. Please also state what DC power rails are available to convert a single ended DAQ input to a diff input. \$\endgroup\$
    – Andy aka
    Apr 28 '20 at 11:48
  • \$\begingroup\$ @Andyaka Hi Andy. Be more specific on the accuracy you're looking for. The DAQ does not have proper differential inputs, I have made my peace with this fact. As mentioned, the sensor(s) are generally three wire, ground, +5 V and 0-5 V signal return so 0 V and 5 V power is available. From the DAQ side I can generate any voltages necessary. \$\endgroup\$
    – J Collins
    Apr 28 '20 at 12:02
  • \$\begingroup\$ What is the DAQ - is it a purchased item? \$\endgroup\$
    – Andy aka
    Apr 28 '20 at 12:04
  • \$\begingroup\$ What budget are you looking at per channel? \$\endgroup\$
    – Andy aka
    Apr 28 '20 at 12:27
  • \$\begingroup\$ Is your current DAQ capable of delivering 0.2% accuracy? \$\endgroup\$
    – Andy aka
    Apr 28 '20 at 12:35
1
\$\begingroup\$

What component or circuit exists that performs this function?

Given that the OP has revealed that the DAQ is the DT9816-S, in order to get the most from converting its very good single ended input to a differential input, I would suggest using the AD8221 instrumentation amplifier as a decent choice. I've used them myself when designing DAQs for test beds and found them to be pretty good: -

enter image description here

Because of the need to protect sensitive inputs I would use series resistors with each input. They will also allow for a little RFI filtering: -

enter image description here

The gain set resistor can be left open circuit for G = 1 operation.

\$\endgroup\$
4
  • \$\begingroup\$ Great answer. Is there a limit to the common mode across -IN and +IN? \$\endgroup\$
    – J Collins
    Apr 30 '20 at 10:11
  • \$\begingroup\$ Running on +/- 15 volt rails should hopefully allow several volts of CM noise or offset to be tolerated. Adding the filter means CM high frequency noise would have less effect. Having a lower cut-off frequency for the filter improves this more. See also fig 17 in the data sheet for common mode rejection characteristics. Circa 90 dB rejection at 20 kHz. \$\endgroup\$
    – Andy aka
    Apr 30 '20 at 10:21
  • \$\begingroup\$ Okay so the key is that the two circuits need to be at the same ballpark ground within the supply range of the amplifier. There can't be arbitrary thousands of volts different. Does this suggest say an oscillocope handles the situation differently? \$\endgroup\$
    – J Collins
    Apr 30 '20 at 11:20
  • \$\begingroup\$ An oscilloscope (by default) does not have a differential input and is prone to all sorts of common mode noise problems. \$\endgroup\$
    – Andy aka
    Apr 30 '20 at 11:25
0
\$\begingroup\$

Based on the accuracy levels, use differential amplifiers at the inputs, this way each input to the ADC is signal - signal ground, so no matter how its offset or moves, the difference between them is consistant, you would have then a seperate "sense" lead back to each sensors ground pin

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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