Analyses on single point data signal

[Getwonder]

In Order to remove the noise of 50 Hz that is embeded in my origional signal of 5mV.

Learning from the project given on http://zone.ni.com/devzone/cda/tut/p/id/4851

I developed this circuit with the following parameters

Filter Type:- Bandstop Filter

Low Cutoff:- 45 Hz

High Cutoff:- 50 Hz

Sampling Frequency:- 188 Hz

Order:- 20

Windowing:- None

I am unable to remove the noise completely.

Kindly guide me some technique in order to remove this noise in an effective manner without reducing the amplitude of the original signal

[johnsold]

1. The upper cutoff frequency should be higher than 50 Hz, perhaps 55 Hz to be somewhat symmetrical with the lower cutoff. If the interference happens to be slightly higher than 50 Hz you would not filter very much of it.

2. The bandstop filter  will not remove higher harmonics.  If the interfering signal is not perfectly sinusoidal, that filter will not remove all of the interference.

Lynn

[Getwonder]

2- Sir Any solution do you have regarding it?

[johnsold]

The best way to get rid of an interfering signal is to prevent it from getting into the data.   Since you are dealing with 5 mV signals and a 50 Hz interference, it sounds like you may be monitoring an physiological signal or the output of a strain gage sensor.  In either case using a suitable signal conditioner (differential amplifier and filters) ahead of the A/D converter will often eliminate most of the interference.  Trying to eliminate it later in software is quite often much less successful.

Lynn

[Getwonder]

I have tried many analog filters but their response is not at all accurate, Also they eats the required signal to so i was trying LabVIEW to solve it.

So what do you suggest?

[Eric_L]

I have worked with many physiological signals in the 5-100 microvolt range. To remove the line noise in the signal, you will have to do it in hardware or oversample (much higher than 188Hz, 1-2kHz should do) the signal and apply a digital filter to cleanly remove the 50Hz line noise.

Eric

[Getwonder]

Thats great that you have also worked on the same.

Can you pls send me more detail on both.

[Getwonder]

To inform you what i have done so far.

I have built an analog filter design using MF10.

But not satisfied with its result. Do you have any other solution to design a bandstop filter in analog way.

So i was trying LabVIEW to achieve my task.

[Eric_L]

Building a precise analog filter is really difficult to do. I have built a couple but I would normally just lowpass at 60Hz since I was not interested in the signals above that frequency. The lowpass does not require the precision that a notch/bandstop filter requires. We then moved to amplifiers that oversampled in the MHz range and apply a digital filter in hardware before down sampling to my acquisition frequency of 256Hz. When I am unsure of what filters I want to run, I oversample at 1200Hz and apply filters in post processing. Sample at the highest possible frequency, then apply your filter and see if that makes you signal look any better. You can also resample your data after applying your filter if you need your data to be at 188Hz.

Eric

[johnsold]

"The MF10 is a switched capacitor (sampled data) filter."  From page 14 of the data sheet.  This is not an analog filter.  It is not an anti-aliasing filter.  It can be very useful when you need to tune a filter over a wide range of frequencies.  It is also a 2nd order filter which means that its cutoff is not very steep.

As Eric said, appropriate processing in the analog domain is best.  I do not believe that you have said what kind of signals you are measuring.  A good differential amplifier or instrumentation amplifier along with some shielding and proper grounding (earthing) will usually eliminate or significantly reduce common mode noise and interference.

Please give us an overview of your entire system from sensors/electrodes through amplifiers and signal conditioners to the data acquisition equipment along with what you want to do with the data after you get it.

Lynn