interference and noise question

martin

That's a huge amount of gain. If the load cell has full-scale output of 5 mv/v, and you excite it from 5 volts, you should get +-25 mv full-scale, so a gain more like 200 might be reasonable.

Lots of instrumentation amps are sensitive to ambient RF pickup. The bipolar input junctions make good RF detectors, and rectify ambient radio stations and such into DC offsets. Offsets caused by hand-waving, or grabbing leadwires, suggests RF. Your board should have a solid ground plane, be in a sturdy box (for thermal reasons, too) and all signal and power paths should be well EMI filtered.

You can expect some hum pickup, and lots of vibration noise, so a slow integrating a/d converter, or lots of signal averaging, is usually needed for load cells.

What's your physical construction like?

I could post a known-good load-cell conditioner front-end schematic to abse if there were huge demand.

John

I'd love to see the schematic, as I havent really done much with modulated DC :-)

martin

Use twisted shielded pairs both to and from the load cell, if necessary drive the shield from the load cell with the common mode signal. Use separate shielded twisted pairs. Also why such high gain (10,000x)? (100x to 1000x total gain is usually enough)

You should be using shielded cable for all wires going to the load cell.

What kind of bandwidth do you need? If it's less than 60Hz, you can roll off the response of the amplifiers with some capacitors, then you'll lose all the power-line and higher related noise.

You'll also need really quiet power supplies and some filtering right at the amps.

OK, it's up.

John

Oh, with all that gain, make sure the amp chain isn't oscillating.

John

John Larkin escreveu:

Hello John,

First of all, thanks for helping.My first experiment using this load cell had a gain of approximately 500. The reason I increased the gain is because I could only notice a small change in the signal when I tested it. It is for use with a impact machine, so, the load cell is made up of iron with some other hard material so that it almost doesn=B4t deform.After increasing the gain I could notice the signal as the mass impacts the object.

I don=B4t have much experience with these projects so I=B4ll try to answer your questions the way i=B4m doing and the way I believe is right(but it might not be).

The signal caused by the impact lasts approximately 50ms, so I assume I need a large bandwidth. I put a filter after the AMP in a 40Khz cutoff frequency. I have a 60Hz noise disturbing my system, thats the main problem regarding filtering in my opinion. If I get rid of that noise, It=B4ll be amost good. I=B4m afraid I lose the signal(it gets filtered) if I=B4d put a 30HZ cutoff frequency, thats why I haven=B4t decided what filter to use.

My physical construction is a breadboard, but I=B4ll change to a printed circuit board as soon as possible.

You said your schematic is up, how can I see it? Could you send me the link?

Thanks a lot!

Boris

It's in the alt.binaries.schematics.electronics newsgroup. If you don't get the binary groups, email me. jjlarkin atsign highlandtechnology dotty com.

A more sensitive load cell, better matched to the mechanical load, would help. But with proper shielding and signal conditioning, 60 Hz pickup should be far below the thermal (random) and vibration noise floor.

John

I'm just a techie, but would it make any sense to swap the gain stages? i.e., the 20X first, then the 500X?

Thanks, Rich

Ideally, about equal gain per stage gives best bandwidth and stability... 100x100 maybe. But low frequency stuff like this isn't all that picky.

But with 10k gain and 10's of KHz bandwidth, overall GBW is in the hundreds of MHz, so a clean layout and a good ground plane are prudent.

John