Hi All,

I have to measure a set of voltages across some current shunts using an NI based A/D module on the 0-200mV range. The cable run will be a few meters. Would shielded twisted pair be the best kind of cable for this, or something else? And should I tie the shield to ground at the A/D end of the equipment?

Thanks,
Sean

 

It will depend on the signal frequency. But below 100Khz the shielded twisted pair should work best. Do not ground the shield in both ends. My favorite book on the this topic is Noise reduction techniques in electronic systems By Henry W. Ott. Even if it is somewhat old. Perhaps you can find on Google Books

 

t06afre is rigrt on- ground at AtoD end. Assuming A to D input is high Z, no buffering is needed.

 

Hello,

Beside noise problems, which could be eliminated with filters, if the noise is out of the band of interest; you should consider the voltage droop along the cable.

In precision measurement, if the ADC is far from the source or the buffer that feed the signal, it's good practice to close the loop of the buffer (operational amplifier at gain of 1) taking the signal from the ADC input rather than the buffer output. The feedback will compensate the voltage droop.
So in your case I would consider a buffer located next to the shunt. By the way this will depend on the resolution you want to achieve.

...remember to limit your bandwidth with a filter...
...if you have a DC put a filter anyway...this will limit the noise that you capture...
the design spec will drive you...!

Regards,

Mauro

 

Er... A curent shunt is pretty much guaranteed to be a lower source impedance than an op-amp buffer; all that would do is introduce complexity and reduce accuracy.

The direct, screened twisted pair connection is best.

 

As mentioned the design will guide the right solution and required complexity.

In professional designs, where precision is not a compromise, the solution I have proposed is used if the ADC is few inches from the source (on the same PCB). In his case who knows....

By the way I would recommend other approaches for high accuracy:

• Convert the signal to a current (current is sot affected by voltage droop problem). This is why some sensor have a current generator as output...
• Convert the signal to a signal with a frequency proportional to the voltage.

The common approaches is to make some kind of conversion next to the source. Many times designer will try to put the ADC next to the sensor, and use a serial ADC output version to transmit the signal to the DSP/Microntroller or whatever...

...beside that, I would recommend to put Schottky diodes to protect the ADC input (probably your board has already those). This is required because your source, including the cable is high inductive.


Ciao,

Mauro

 

Thanks guys, very informative responses. My sampling rate is very low about 10Hz.

And the reason for not grounding the cable at both ends is to avoid the mysterious ground loops, correct?