Forums

60Hz Notch filter at 25kHz sampling rate?

Started by abradley1984 August 28, 2008
Hello, 

I have an ENG signal, sampled at 25kHz that has high mains interference
that I want to remove.  The problem is I need the surrounding frequencies,
maybe not 50-70, but I can't afford to sacrifice much more than this.  The
filter designs I've been looking at don't seem to be able to do this at
such a high sampling rate.  

Any suggestions?  Would downsampling help?

Thank you in advance for your help

Allison  


On Aug 28, 4:07&#2013266080;pm, "abradley1984" <abradley1...@gmail.com> wrote:
> Hello, > > I have an ENG signal, sampled at 25kHz that has high mains interference > that I want to remove. &#2013266080;The problem is I need the surrounding frequencies, > maybe not 50-70, but I can't afford to sacrifice much more than this. &#2013266080;The > filter designs I've been looking at don't seem to be able to do this at > such a high sampling rate. &#2013266080; > > Any suggestions? &#2013266080;Would downsampling help? > > Thank you in advance for your help > > Allison &#2013266080;
Use multirate methods for the notch filter, you can get the shape you are looking for at a very low computational price. The filter will have a long impulse response so a correspondingly long startup time. What about harmonics of 60? Dirk Dirk
>Hello, > >I have an ENG signal, sampled at 25kHz that has high mains interference >that I want to remove. The problem is I need the surrounding
frequencies,
>maybe not 50-70, but I can't afford to sacrifice much more than this.
The
>filter designs I've been looking at don't seem to be able to do this at >such a high sampling rate. > >Any suggestions? Would downsampling help? > >Thank you in advance for your help > >Allison > > >
Hi Allison, Here's a simple IIR notch filter that you can try. Feel free to e-mail me with questions. Steve http://www.dspguide.com/ch19/3.htm
abradley1984 wrote:
> Hello, > > I have an ENG signal, sampled at 25kHz that has high mains interference > that I want to remove. The problem is I need the surrounding frequencies, > maybe not 50-70, but I can't afford to sacrifice much more than this. The > filter designs I've been looking at don't seem to be able to do this at > such a high sampling rate. > > Any suggestions? Would downsampling help? > > Thank you in advance for your help
Any filter must have an impulse response comperable to the relevant signal's period. In general, IIR filters perform well at frequencies well below the sample rate because, well, infinite is long. The usual transversal FIR must have many stages for its impulse response to be many sample periods long. Moral: try an IIR filter. Question: mains interference usually involves the first few harmonics of the line frequency; have you characterized the noise you want to filter? Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
On Thu, 28 Aug 2008 17:20:45 -0500, SteveSmith wrote:

>>Hello, >> >>I have an ENG signal, sampled at 25kHz that has high mains interference >>that I want to remove. The problem is I need the surrounding > frequencies, >>maybe not 50-70, but I can't afford to sacrifice much more than this. > The >>filter designs I've been looking at don't seem to be able to do this at >>such a high sampling rate. >> >>Any suggestions? Would downsampling help? >> >>Thank you in advance for your help >> >>Allison >> >> >> >> > Hi Allison, > Here's a simple IIR notch filter that you can try. Feel free to e-mail > me with questions. > Steve > > http://www.dspguide.com/ch19/3.htm
Your expression for R may be correct when the fractional bandwidth is << 1, but it doesn't look at all right for BW = 1/2. Little matter in this case.
On Thu, 28 Aug 2008 15:07:53 -0500, abradley1984 wrote:

> Hello, > > I have an ENG signal, sampled at 25kHz that has high mains interference > that I want to remove. The problem is I need the surrounding > frequencies, maybe not 50-70, but I can't afford to sacrifice much more > than this. The filter designs I've been looking at don't seem to be > able to do this at such a high sampling rate. > > Any suggestions? Would downsampling help? > > Thank you in advance for your help > > Allison
What leads you to believe that the filter designs you've been looking at can't hack it? For that matter, what filter designs have you been looking at? An IIR notch filter should work fine on a pure 60Hz tone, but you need a filter with a Q of 5 or so, and a fractional bandwidth of .0024. So you're going to need data paths with at least 22 more bits than your incoming data. Assuming 16-bit input data, you're talking more than 32 bit data paths, which is a bit silly. If you don't really need 25kHz samples, then yes, downsampling will help -- the necessary precision goes down as the square of sampling frequency for a resonant filter. But heed Jerry: mains interference is often much more than a pure sinusoid at the mains frequency. It can have harmonics extending up into the MHz regions on bad days; even 'routine' mains energy can extend into the 100s of Hz. So you either need to implement a bunch of resonant IIR filters, or you need to cobble up a nice long FIR comb filter (if you want to discriminate between 60 and 70Hz you should figure on it being well over 1/10th of a second long). Either of those tasks will be eased by downsampling. --- Tim Wescott http://www.wescottdesign.com Control and communication systems consulting.
On Aug 28, 6:20&#2013266080;pm, "SteveSmith" <Steve.Smi...@SpectrumSDI.com> wrote:
> >Hello, > > >I have an ENG signal, sampled at 25kHz that has high mains interference > >that I want to remove. &#2013266080;The problem is I need the surrounding > frequencies, > >maybe not 50-70, but I can't afford to sacrifice much more than this. > The > >filter designs I've been looking at don't seem to be able to do this at > >such a high sampling rate. &#2013266080; > > >Any suggestions? &#2013266080;Would downsampling help? > > >Thank you in advance for your help > > >Allison &#2013266080; > > Hi Allison, > Here's a simple IIR notch filter that you can try. Feel free to e-mail me > with questions. > Steve > > http://www.dspguide.com/ch19/3.htm- Hide quoted text - > > - Show quoted text -
Hi Steve, Some considerations for using this filter for the OP. It would seem that the biggest possible problem using this filter at a narrow bandwidth is that the stopband width (defined as frequency range for gain below some attenuation threshold) for a reasonable attenuation is quite narrow. The frequency plots (I didn't check the equations) look like they have a zero at the notch frequency and the gain rises rapidly to either side. That means if the frequency is not almost EXACTLY at the notch frequency, or if there is a small variation in the frequency (unintentional frequency modulation, amplitude modulation, frequency drift), the attenuation may decrease dramatically. If any of these cases regarding the frequency is the case, and a lot of attenuation is required, the OP would need to consider a wider stopband width filter, hopefully one that doesn't have to have an excessive increase in bandwidth. I am not saying the OP can't use the filter for his application, just some things to consider. Given the ENG application it would seem that power line frequency harmonics may also be a problem the OP has to deal with. Dirk

abradley1984 wrote:

> Hello, > > I have an ENG signal, sampled at 25kHz that has high mains interference > that I want to remove. The problem is I need the surrounding frequencies, > maybe not 50-70, but I can't afford to sacrifice much more than this. The > filter designs I've been looking at don't seem to be able to do this at > such a high sampling rate. > > Any suggestions? Would downsampling help?
As mentioned by many, it is not a big deal to build the narrowband notch at power frequency with or without downsampling. However if you are dealing with the recorded signal (as opposed to the processing in the real time) the better approach would be to estimate the parameters of the interference and then subtract it from the signal. Assuming the interference is stationary or slowly varying, this will allow canceling it without making damage to the useful data. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
"abradley1984" <abradley1984@gmail.com> wrote in
news:OIydnZkXT-cEmCrVnZ2dnUVZ_gGdnZ2d@giganews.com: 

> Hello, > > I have an ENG signal, sampled at 25kHz that has high mains > interference that I want to remove. The problem is I need the > surrounding frequencies, maybe not 50-70, but I can't afford to > sacrifice much more than this. The filter designs I've been looking > at don't seem to be able to do this at such a high sampling rate. > > Any suggestions? Would downsampling help? > > Thank you in advance for your help > > Allison > >
If ENG is electronystagraphy, you're sampling an order of magnitude faster than you need to, even conservatively. Downsampling would decrease the steepness of your filter, and help you avoid some ringing. The best suggestion, as always, is to figure out why you're having such a big problem with mains noise, and fix it. Increase your CMRR, shield your cables, etc. -- Scott Reverse name to reply
&#2013266080;Downsampling would decrease the
> steepness of your filter, and help you avoid some ringing. > >
Is someone going to challange these two statements? Mark