4

What shielding techniques would be best suited for a digital sensor (3 wires - Pwr, Gnd, Sig) in a noisy environment (specifically fluorescent lighting)?

Twisted pair (trio in this case) works well for differential voltages, but I suspect it won't be effective here even though the current in the wires is balanced.

Does shielded wire require a separate ground wire, or is it acceptable to use the shielding as ground? Should the shielding be ground at both ends, or only at the PCB?

Edit: Sensor is connected to PCB via ~1m of cable.

flag
Are you sure you even need it? – endolith♦ Apr 22 at 3:57
What kind of sensor are you using? What kind of protocol (if it's a complex sensor) does it use? Where are you seeing the noise - in the digital output or the analog front-end of the sensor? – W5VO Apr 23 at 4:31
We are having trouble with sites using a particular fluro ballast, other sites are fine. The sites are remote so diagnosis is hard. The sensors (hall effect) are in a light box with the fluros. – Peter Gibson Apr 23 at 5:27

6 Answers

3

I suggest you purchase / check out a copy of Henry Ott's "Noise Reduction Techniques in Electronic Systems". It talks about these things.

Don't ground the shield at both ends, as this creates a ground loop. The PCB end is probably the best place to ground the shield.

link|flag
Can a ground loop be a problem if the other circuit (sensor) is isolated from the ground? – jpc Apr 23 at 6:28
Ott is definitely the reference to have for this. – Clint Lawrence♦ Apr 23 at 8:32
@jpc: yes. The shield is not intended to carry current, so that it is of uniform potential. So you need to use another conductor to carry the return current at ground. But if you tie the shield to the ground conductor at both ends, then you have a loop. – Jason S♦ Apr 23 at 11:36
1

You can use the shielding as the ground wire. Given that the sensor current is probably tiny, the impedance of the shield will be very low.

You should ground the shield at least on the PCB. It doesn't matter whether you ground it at the sensor or not. What you're trying to do is reduce the size and impedance of loops in noise can be induced.

With a digital signal, you should be relatively impervious to noise, anyway.

link|flag
Gack! Don't use the shielding as the ground wire. – Jason S♦ Apr 22 at 12:33
I'm not saying it's a good idea, but for a sensor where the current will be a few mA, isn't the impedance of most shielding low enough? – pingswept♦ Apr 22 at 14:12
it's less that, than the idea that the shielding is not supposed to supply current to/from a circuit. Otherwise it can act as an antenna and radiate noise. You hide the circuit currents from the external world inside the shield. – Jason S♦ Apr 23 at 11:39
1

Try using some ferrite beads. They're the big lump at the end of some cords, like USB cords and power cords. They filter noise pretty well.

link|flag
1

You could probably try filtering out the 60Hz noise using a simple RC filter.

link|flag
Florescents aren't 60 Hz (well, maybe some are). Ballasts are used to ramp that 60 Hz up to 20 kHz or so. – ajs410 Apr 22 at 19:13
0

If your question is one of signal integrity, are you sure you have a problem? If so by how much and at what frequencies? Are you doing impedance matching between the driver and the transmission line? If so how?

If your worried about EM emissions from the cable, which i bet are quite high with a single ended digital signal then the best solution is to create a small pcb for the sensor to live on along with a differential line driver.

General question, whats the clock frequency of the signal?

link|flag
0

Are you sure the problem is noise coupling onto the cable? Is there any chance the ballast is producing a big enough magnetic field to saturate the sensor itself?

link|flag

Your Answer

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