I'm using an instrumentation amplifier to amplify the signal from a load cell. You can see a zoom-in of the amplifier output in the two attached files. The general 50mV noise doesn't bother me so much, but every 2 - 3 msec. there's an alternating +/- spike that's around 120mV peak-to-peak. scope_1.jpg zooms in on one of these spikes. Maybe it's so fast I don't have to worry about it (I'm averaging 50 samples over 5 msec. to get my value), but I was wondering if there's a way to filter it out anyway. Thanks for your help.

 

Low-pass filter?

John

 

What is the frequency range of the signal you want to see? If the "spikes" are far enough away from those frequencies, then a low-pass filter (RC, or similar) should work. If you are sampling the signal with an ADC, then you have to be sure that the spikes are not aliased such that they look like part of the REAL signal of interest.

You may need an analog active filter if the spike is close to your required signal frequencies.

Do you have any idea what is causing the spikes?

 

I'm using an instrumentation amplifier to amplify the signal from a load cell. You can see a zoom-in of the amplifier output in the two attached files. The general 50mV noise doesn't bother me so much, but every 2 - 3 msec. there's an alternating +/- spike that's around 120mV peak-to-peak. scope_1.jpg zooms in on one of these spikes. Maybe it's so fast I don't have to worry about it (I'm averaging 50 samples over 5 msec. to get my value), but I was wondering if there's a way to filter it out anyway. Thanks for your help.

Post your schematic.

 

An interesting way to reduce spikes without a filter, even if the spike frequency is comparable to the signal frequency is to use a slew-rate limiter circuit. This works as long as the spike amplitude is larger than the signal amplitude.
The limiter reduces the voltage and thus the energy of the spikes without otherwise affecting the bandwidth.
Here's a discussion of an op amp slew rate limiter circuit.
You just select the maximum slew rate to pass the highest frequency signal at its maximum peak value (slew-rate = 2πfVp where f is the highest frequency and Vp is the maximum peak signal voltage), and any signal that has a higher slew rate will be suppressed.
You could also combine the slew rate circuit with a low-pass filter for additional suppression of the spike if it is at a higher frequency than the signal.

 

If I read your scope picture correct the event is over in 10 us and you sample every 5 ms/50 = 100 us. So in your sample you may get one single event in your sample.

Why not just toss the largest and smallest of your samples and use the remaining 48 to do your averaging? I've done similar things before so I toss away any glitches I happen to capture.

Zero hardware, tiny bit of software.

 

So where are the spikes coming from?

 

If its external noise then a differential amplifier on the front end should get rid of it.

Otherwise a filter is needed to keep wanted signal but reject glitches and noise.