Dear comp.dsp

I'm a ECE undergrad student. I'm using Matlab to clean up unipolar cardiac electrogram
recordings. I expect signals between 30 and 500 Hz. However, I have noticed significant noise at
60 Hz and its harmonics (you can see peaks in the PSD at 60,120,180,... Hz). We are sampling at
25 kHz.

My question is: since I'm expecting signals in the 30 to 500 Hz range, how should I filter out
the harmonics? I could do a comb filter after decimating the  original signal, but I don't want
to cancel out the actual signal.
Is there a way to get rid of these peaks by, for example, replacing the harmonic peaks by
interpolated values from neighboring frequencies? I've read that interpolation in the frequency
is tricky.

Thank you for your time.

- Fernando

f...@gmail.com wrote:
> Dear comp.dsp
>
> I'm a ECE undergrad student. I'm using Matlab to clean up unipolar cardiac electrogram
recordings. I expect signals between 30 and 500 Hz. However, I have noticed significant noise at
60 Hz and its harmonics (you can see peaks in the PSD at 60,120,180,... Hz). We are sampling at
25 kHz.
>
> My question is: since I'm expecting signals in the 30 to 500 Hz range, how should I filter
out the harmonics? I could do a comb filter after decimating the  original signal, but I don't
want to cancel out the actual signal.
> Is there a way to get rid of these peaks by, for example, replacing the harmonic peaks by
interpolated values from neighboring frequencies? I've read that interpolation in the frequency
is tricky.
>
> Thank you for your time.
>
> - Fernando

I'm an antique. What is "ECE"?

However, back in the 60's I was a BSEE student working as an 
electronics tech for the university's veterinary physiology 
department. We had lousy EEG and EKG signals that I could 
never filter successfully. Understood years later why it is 
better to "prevent interference" than to "ATTEMPT to REMOVE 
interference".

Filtering after the fact is not a good answer.
Do you have the possibility of getting better data?
Or are you stuck trying to ... ... ... ;/

> I'm an antique. What is "ECE"?
ECE means Electrical and Computer Engineering. It's a dual major program :)

> However, back in the 60's I was a BSEE student working as an 
> electronics tech for the university's veterinary physiology 
> department. We had lousy EEG and EKG signals that I could 
> never filter successfully. Understood years later why it is 
> better to "prevent interference" than to "ATTEMPT to REMOVE 
> interference". 

What do you mean by preventing interference? Does that mean having shielded wires going from the
electrodes to the data acquisition system? Or does that mean to have hardware analog filters
before sampling is done?

> Filtering after the fact is not a good answer.
> Do you have the possibility of getting better data?
> Or are you stuck trying to ... ... ... ;/

Right now, the acquisition box has a 60 Hz notch filter, but it doesn't get rid of the
harmonics. Therefore, I'm stuck trying to do this :/

What I'd try is
- take a piece of signal with an exact integer length of 60 Hz cycles
- fft
- zero out the n * 60 Hz bins at positive and negative frequencies
- ifft (and discard the very small imaginary part, it's roundoff noise)

This is more or less ideal, as long as your powerline hum and the
acquisition clock are free of phasenoise.
More advanced (nonlinear) methods would use a PLL, for example.
If you search for this, you'll probably find more info than you can read in
a lifetime.

forgot, there can also be amplitude variations in the hum, leaking it out
of an FFT bin.

On Tue, 21 Aug 2012 10:43:35 -0700, f.quivira wrote:

>> I'm an antique. What is "ECE"?
> ECE means Electrical and Computer Engineering. It's a dual major program
> :)
> 
>> However, back in the 60's I was a BSEE student working as an
>> electronics tech for the university's veterinary physiology department.
>> We had lousy EEG and EKG signals that I could never filter
>> successfully. Understood years later why it is better to "prevent
>> interference" than to "ATTEMPT to REMOVE interference".
> 
> What do you mean by preventing interference? Does that mean having
> shielded wires going from the electrodes to the data acquisition system?
> Or does that mean to have hardware analog filters before sampling is
> done?

Hardware analog filtering is just attempting to remove interference in 
the analog domain rather than digital.

He means shielded wires, or balanced wires, or whatever it takes to keep 
the 60Hz from getting into your electronics in the first place.

>> Filtering after the fact is not a good answer. Do you have the
>> possibility of getting better data? Or are you stuck trying to ... ...
>> ... ;/
> 
> Right now, the acquisition box has a 60 Hz notch filter, but it doesn't
> get rid of the harmonics. Therefore, I'm stuck trying to do this :/





-- 
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com

On Tuesday, August 21, 2012 11:47:25 AM UTC-5, f.qu...@gmail.com wrote:
> Dear comp.dsp I'm a ECE undergrad student. I'm using Matlab to clean up unipolar cardiac
electrogram recordings. I expect signals between 30 and 500 Hz. However, I have noticed
significant noise at 60 Hz and its harmonics (you can see peaks in the PSD at 60,120,180,...
Hz). We are sampling at 25 kHz. My question is: since I'm expecting signals in the 30 to 500 Hz
range, how should I filter out the harmonics? I could do a comb filter after decimating the
original signal, but I don't want to cancel out the actual signal. Is there a way to get rid of
these peaks by, for example, replacing the harmonic peaks by interpolated values from
neighboring frequencies? I've read that interpolation in the frequency is tricky. Thank you for
your time. - Fernando

Get a copy of a book by Widrow and Stearns titled Adaptive Signal Processing. In there you will
find a section titled "Cancelleing 60-Hz Interference in Electrocardiography". Your
library should have a copy, or can get one easily. Be sure to check with your prof to see if
(s)he wants you to go this route to clean up the signal.


Maurice

On 21.8.12 9:43 , Tim Wescott wrote:
> On Tue, 21 Aug 2012 10:43:35 -0700, f.quivira wrote:
>
>>> I'm an antique. What is "ECE"?
>> ECE means Electrical and Computer Engineering. It's a dual major program
>> :)
>>
>>> However, back in the 60's I was a BSEE student working as an
>>> electronics tech for the university's veterinary physiology department.
>>> We had lousy EEG and EKG signals that I could never filter
>>> successfully. Understood years later why it is better to "prevent
>>> interference" than to "ATTEMPT to REMOVE interference".
>>
>> What do you mean by preventing interference? Does that mean having
>> shielded wires going from the electrodes to the data acquisition system?
>> Or does that mean to have hardware analog filters before sampling is
>> done?
>
> Hardware analog filtering is just attempting to remove interference in
> the analog domain rather than digital.
>
> He means shielded wires, or balanced wires, or whatever it takes to keep
> the 60Hz from getting into your electronics in the first place.
>

That ain't so simple:

The capacitance of a person to the active phase of the mains power is
of the order of 20 pF, and the capacitance to ground is of the order
of 200 pF, which creates a high-impedance 1:10 voltage divider feed from
the 60 Hz (or 50 Hz on this side of the pond) power signal.

The trick for a successful capture of a biosignal (EEG, EKG, EMG etc) is
to use a differential input amplifier. The contact impedances to the
actual signal source are quite high, and very probably unbalanced. To
keep an useful CMRR, the input impedance to the pre-amplifier must be
very high, including a very small effective capacitance. The cables have
to be shielded to keep most of the crud out, but the shields usually
need bootstrap feedback from the amplifier, to keep the input impedance
high enough.

For patient safety, the input amplifier should be totally floating, with
pretty small leakage capacitances to the ground.

I'm of the same opinion as the other posters:

  - the main way to handle with the hum is to prevent it from entering,

  - the way to filter the hum with harmonics is a comb filter, but
    you have to be prepared for appreciable signal distorsion.

Been there, done that - I built EKG equipment for intensive care
in the 70's.

-- 

Tauno Voipio

On 8/21/12 3:03 PM, maury wrote:
>
> Get a copy of a book by Widrow and Stearns titled Adaptive Signal Processing. In there you
will find a section titled "Cancelleing 60-Hz Interference in Electrocardiography".
Your library should have a copy, or can get one easily. Be sure to check with your prof to see
if (s)he wants you to go this route to clean up the signal.
>

ECG and speaker-phones were, i believe, the original two applications of 
Widrow's LMS filter.  back in the '90s (maybe '92), the AES convention 
in SF had Widrow as an invited speaker and he brought up both of these 
applications.  it was a good talk.

you use the actual line voltage as the "desired signal" that you're 
trying to cancel.  if the line voltage is too smooth (not enough 
harmonic energy), you might want to jazz it up a little by running it 
through a slight non-linearity to generate harmonics so the LMS has 
something to cancel.

but, if a 60 Hz notch filter works to get the fundamental, try a sharp 
comb filter to get it and the harmonics.  if you sample the line voltage 
anyway, you can use that to "pitch track" to make sure the notches of 
your comb are positioned right-on-the-money.  you can make the notches 
as slim as you want.


-- 

r b-j                  r...@audioimagination.com

"Imagination is more important than knowledge."

Richard Owlett <r...@pcnetinc.com> wrote:

(snip)

> I'm an antique. What is "ECE"?

Around here (Seattle) it is Electrical and Computer Engineering.

What many people think of as Computer Science is more Engineering,
and belongs in an Engineering department. Computer Science, then,
is closer to Math, maybe part of the Math department.

> However, back in the 60's I was a BSEE student working as an 
> electronics tech for the university's veterinary physiology 
> department. We had lousy EEG and EKG signals that I could 
> never filter successfully. Understood years later why it is 
> better to "prevent interference" than to "ATTEMPT to REMOVE 
> interference".

-- glen