usb power noise filtering

[bmx]

Greetings. I have a problem with noise induced in a device by the usb interface power. The device consists of a sensor and an analog to digital converter. The converter is powered by 5v from the usb and uses the usb for data transfer. When powering the sensor from the usb, there is a lot of noise. When powering the sensor from a battery the noise is much lower.
What kind of filter can I use for the sensor considering the sensor needs 5VDC and 5mA power input ?

[gootee]

Have you tried putting a 10uF-or-so electrolytic capacitor (not a low ESR type) in parallel with a 0.1uF (not NPO or C0G type), from 5V to gnd, right at the 5V input to the sensor and right at the 5V input to the converter? (i.e. One pair for each 5V input.) Use shortest possible leads and connect as closely as possible to 5V input pins. In particular, connect 0.1uF caps as closely as possible to the 5V input pins. You could vary the cap values and see what happens, too.

You might want to take the sensor's caps' grounds to the main USB ground, i.e. not to the one that comes through the converter. I would probably try it both ways, just to see what happens.

Depending on how much current is drawn, you might also be able to put a 10-to-100-or-so Ohms resistor (metal film type) in the 5V line, just before the caps to ground at the converter, to form a better low-pass filter. Try something like 33 Ohms, maybe.

If the current draw is small-enough, you could put another resistor in the 5V line just before the caps at the 5V input of the sensor. I would probably take the 5V for the sensor from _before_ the resistor in line with the converter's 5V input. But you should try it both ways.

You should also see what happens if you connect the sensor's ground directly to the USB ground, instead of having the sensor share the ground run with the converter.

I would probably try using the largest resistances I could use, vis-a-vis the current drawn vs the voltage drop, to get the most filtering.

You might also want to try using a cap pair to gnd right at the USB 5V/gnd OUTPUT, if you use a resistor before the converter's 5V input, so that you also low-pass filter the noise going back INTO the USB 5V output.

Cheers,

Tom Gootee

[wakibaki]

It looks to be low frequency, but try a ferrite on the cable. If you've got a high-mu toroid, try a dozen or so turns. There might be more than one cause so don't remove any fixes until you can see it's gone. Then you can try removing them to see if it comes back.

I had a really nice handheld Tektronix scope. It never really worked as good noise-wise on its mains supply as it did on batteries, so don't expect miracles.

w

[star882]

For that low current, try a simple RC filter. Something like 10 ohms and 100uF or so.

[gootee]

Originally Posted by star882

For that low current, try a simple RC filter. Something like 10 ohms and 100uF or so.

Yahhh mon! Dat prolly be goood (enough).

[stormsonic]

there must be some voltage headroom to spare, as regulator will always drop some voltage. You can not get "clean" 5V output from "dirty" 5V input.
Even with simple RC filter there will be voltage drop on resistor. Ohm law is working against you
And feritte beads will "kick in" only at MHz range, at audio frequencies they do nothing.

[bocka]

Originally Posted by star882

For that low current, try a simple RC filter. Something like 10 ohms and 100uF or so.

The usb spec requires not using more than 10uF as input cap. It's better to look for another solution.

[bmx]

Thank you very much for the solutions. I will try them all and update as soon as I'm done.

[bmx]

Back with an update. I've tried so far
A/D converter powered through a 22ohm resistor followed by a 4.7uF and a 100uF capacitor and the sensor powered separately through the same value resistor and capacitors. No change.
I've tried to power the sensor through a fast diode and still no change.

Switched PC's and the noise is the same.

I have a clip uploading to show the fluctuating noise in the 20-2000Hz spectrum. It looks like it's pulsating.

Tried battery power again and the noise magically drops and the pulsation stops.

What can I try next ?

[bmx]

video!
27042010074.flv video by DAdrianD - Photobucket