Butterworth filter will not remove 60Hz noise

[Lammers]

I am attempting to use a Butterworth filter to remove 60 Hz noise from a signal.  It's quite obviously 60 Hz - when I recorded a signal for one second and then looked at the signal in Excel, there were 60 waves.  When I filter the data using an 8th order bandstop Butterworth filter, with the low stop = 58Hz and the high stop = 62Hz, there is a huge bit of noise for the first 1/8 sec.  The noise then goes down, and there is considerably less 60Hz noise, but it is very clear that the 60Hz noise is still there.

Any idea what I'm doing wrong?  Attached is the VI, which currently tries to use three different ways of wiring a Butterworth filter.  Also attached is a pre-filtered one-second data sample and the same data sample after filtering.

[mikeporter]

Trying to filter 60Hz power-line noise is software is a bit like closing the barn door after the cows have escaped -- your response is too little, too late. If you are seeing 60Hz noise like that you have a hardware problem in the analog front-end of your acquisition system that needs to get fixed. Check shielding, input method (single-ended vs double-ended), wire routing, wiring errors, etc.

Mike...

Message Edited by mikeporter on 05-29-2007 12:45 AM

[Lammers]

Thanks for the reply - so the filter is probably working, but the noise is just way too much?  I'll also try grounding the instrument and the data acquisition box. 

What do you mean by single-ended vs. double-ended input method?  The data acquisition board is used to complete the "corners" of a Wheatstone bridge, but I'm not sure if that makes the input single-ended or double ended.

[johnsold]

A Butterworth filter has a relatively slow rolloff in frequency. An eliptical filter or a notch filter might be better choices if you have important signal components near the interference.

Another issue is the transient response of the filter. That is probably what is happening in the first 1/8 second.

As Mike said, it is always better to prevent the interferenece from entering your system than to try to remove it later.

After looking at your data, I think you will have some difficulties with simple filtering. It appears that five different types of data occur within that one second. Even within the subsets, some strong shifts or jumps change the phase of the interferring signal. What is your real signal supposed to look like and how is it being genreated and processed?

Lynn

[Lammers]

The physical instrument is designed to measure ground reaction forces from small animals moving across a force plate.  Basically it's a voltage change multiplied by a coefficient to get force, which should be zero in these sample.  But the data samples attached in this thread are a record of one second of data collection (sample frequency = 20,000 Hz) with no signal.  Thus these data samples show pure noise and no actual signal.  When force is actually being recorded, the signal is considerably stronger than the noise in some directions.  But in other directions (especially "mediolateral") the signal to noise ratio is too low to be very usable.  I would very much like to reduce the noise either during data collection, or afterward by filtering.

I'm moving my lab from the fourth floor to the first floor within a week or two.  The 60Hz noise may be substantially reduced by moving... the fourth floor lab often has a lot of background noise that is absent on the first floor of my building!

[RickH]

Single ended Vs Double-ended termination-  Double-ended is grounding both ends of the coax shield to two chassis grounds.  In many (if not all) cases, a common ground current is produced across the coax shield.  This is referred to as Common Mode Noise.  Single-ended termination, allows grounding only at one end of the coax.  In this case the coax shield prevents stays noise into the signal, but does not provide an electrical connection between the two chassis grounds.  This principal is applied also to twisted pair wires, which require impedance matching to cancel interferring noise.   Are we confused yet?

Regards,

Rick