David Smith

10-31-2006, 11:02 PM

Dear Kieran

Kieran

My thoughts are;

Perhaps the authors mean the frequency range of the amplifier should be

twice that of the input signal after the gain is applied so that if there

are unexpectedly high input signals they are not clipped at their peaks and

result in loss of precision.

Also:

The gain applied to the input signal should ideally be equal across all

frequencies EG from 10 to 300HZ. If it is not the extreme frequencies may

experience relative distortion. If at midrange the amplifier, with bandwidth

of 0-1000Hz, gain was a nominal 100% then at 1000HZ it may only be 70%

therefore to ensure that gain is equal across the frequencies of interest

(i.e. the midrange) then it may be necessary to use a band with up to

2000Hz.

Or;

Maybe it is because by experience these authors have found that the higher

sampling rate characterises the signal of interest better than the Nyqvist

minimum of x 2. (I think Nyqvist meant this a a minimum standard since a x2

sampling rate will only capture the peaks of the signal if you are lucky and

if one uses this x2 standard then it is neccesary to repeat the recording

several times and look at the range of signal characterisation. In this case

it may only be possible to characteraise the signal of interest with a low

confidence value in +/- 1sd)

In my studies on sampling rates and EMG I found that a sampling rate of x2

was not very accurate in characterising the signal. I believe that a more

reasonable sampling rate is 4-7 times the signal of interest. However there

needs to be a balance between successful characterisation of the signal and

capture of unwanted higher frequency signals.

Does this make sense to anyone?

Regards Dave Smith

----- Original Message -----

From: "Kieran Moran"

To:

Sent: Tuesday, October 31, 2006 6:31 PM

Subject: [BIOMCH-L] EMG filtering and sampling rates

> Hi

>

> On the issue of sampling and filtering, Peter Konrad (The ABC of EMG

> published by Noraxon) states (p13)

>

> "In order to accurately “translate” the complete frequency spectrum of a

> signal, the sampling rate at which the A/D board determines the voltage of

> the input signal must be at least twice as high as the maximum expected

> frequency of the signal. This relationship is described by the sampling

> theorem of Nyquist: sampling a signal at a frequency which is too low

> results in aliasing effects (Fig. 18). For EMG almost all of the signal

> power is located between 10 and 250 Hz and scientific recommendations

> (SENIAM, ISEK) require an amplifier band setting of 10 to 500 Hz. This

> would result in a sampling frequency of at least 1000 Hz *(double band of

> EMG)* or even 1500 Hz to avoid signal loss".

>

> This is also suggested by ISEK

> (http://isek-online.org/pdf/ISEK_EMG-Standards.pdf) on page 2.

>

> However, I am unsure why the sampling rate must be more than twice the

> highest amplifier *band width* setting and not simply more than twice the

> *"maximum frequency of interest in the signal"*. In other words, if we use

> a band setting of 10 to 1000 Hz, and we are interested only in signals up

> to 400 Hz, would it not be appropriate to sample at 1000Hz. The

> recommendations from Peter Konrad would suggest the sampling rate should

> be more than 2000Hz (twice the highest *band width*).

>

> A number of published experimental studies would seem to agree that the

> sampling rate does not have to be twice the highest band width, but twice

> the highest maximum frequency of interest

>

> J.R. Potvin & S. Brown (2004) Less is more: high pass filtering, to remove

> up to 99% of the surface EMG signal power, improves EMG-based biceps

> brachii muscle

> force estimates. Journal of Electromyography and Kinesiology 14 (2004)

> 389–399:

> (gain = 1000, input impedance = 10 GX, 10–1000 Hz, CMRR = 115 dB at 60 Hz,

> Bortec, Octopus AMT-8, Calgary, Canada) and sampling rate was 1024 Hz

>

>

> M.G. Feltham et al. (2006) Changes in joint stability with muscle

> contraction measured from

> transmission of mechanical vibration

> Journal of Biomechanics 39 (2006) 2850–2856

> (gain = 1000, band-pass filtered between 10 Hz and 1000 Hz (NL820,

> Neurolog System, Welwyn Garden City, UK), and sampled at 1000 Hz

>

> Does any one have any thoughts on this issue?

>

> Many thanks

>

> Kieran

>

> --

> Dr Kieran Moran

> Biomechanics Research Group

> School of Health and Human Performance

> Faculty of Science and Health

> Dublin City University

> Collins Avenue

> Dublin 9

>

> tel; 00 353 1 700 8011

> fax: 00 353 1 700 8888

>

> kieran.moran@dcu.ie

>

> ---------------------------------------------------------------

> Information about BIOMCH-L: http://www.Biomch-L.org

> Archives: http://listserv.surfnet.nl/archives/Biomch-L.html

> ---------------------------------------------------------------

>

Kieran

My thoughts are;

Perhaps the authors mean the frequency range of the amplifier should be

twice that of the input signal after the gain is applied so that if there

are unexpectedly high input signals they are not clipped at their peaks and

result in loss of precision.

Also:

The gain applied to the input signal should ideally be equal across all

frequencies EG from 10 to 300HZ. If it is not the extreme frequencies may

experience relative distortion. If at midrange the amplifier, with bandwidth

of 0-1000Hz, gain was a nominal 100% then at 1000HZ it may only be 70%

therefore to ensure that gain is equal across the frequencies of interest

(i.e. the midrange) then it may be necessary to use a band with up to

2000Hz.

Or;

Maybe it is because by experience these authors have found that the higher

sampling rate characterises the signal of interest better than the Nyqvist

minimum of x 2. (I think Nyqvist meant this a a minimum standard since a x2

sampling rate will only capture the peaks of the signal if you are lucky and

if one uses this x2 standard then it is neccesary to repeat the recording

several times and look at the range of signal characterisation. In this case

it may only be possible to characteraise the signal of interest with a low

confidence value in +/- 1sd)

In my studies on sampling rates and EMG I found that a sampling rate of x2

was not very accurate in characterising the signal. I believe that a more

reasonable sampling rate is 4-7 times the signal of interest. However there

needs to be a balance between successful characterisation of the signal and

capture of unwanted higher frequency signals.

Does this make sense to anyone?

Regards Dave Smith

----- Original Message -----

From: "Kieran Moran"

To:

Sent: Tuesday, October 31, 2006 6:31 PM

Subject: [BIOMCH-L] EMG filtering and sampling rates

> Hi

>

> On the issue of sampling and filtering, Peter Konrad (The ABC of EMG

> published by Noraxon) states (p13)

>

> "In order to accurately “translate” the complete frequency spectrum of a

> signal, the sampling rate at which the A/D board determines the voltage of

> the input signal must be at least twice as high as the maximum expected

> frequency of the signal. This relationship is described by the sampling

> theorem of Nyquist: sampling a signal at a frequency which is too low

> results in aliasing effects (Fig. 18). For EMG almost all of the signal

> power is located between 10 and 250 Hz and scientific recommendations

> (SENIAM, ISEK) require an amplifier band setting of 10 to 500 Hz. This

> would result in a sampling frequency of at least 1000 Hz *(double band of

> EMG)* or even 1500 Hz to avoid signal loss".

>

> This is also suggested by ISEK

> (http://isek-online.org/pdf/ISEK_EMG-Standards.pdf) on page 2.

>

> However, I am unsure why the sampling rate must be more than twice the

> highest amplifier *band width* setting and not simply more than twice the

> *"maximum frequency of interest in the signal"*. In other words, if we use

> a band setting of 10 to 1000 Hz, and we are interested only in signals up

> to 400 Hz, would it not be appropriate to sample at 1000Hz. The

> recommendations from Peter Konrad would suggest the sampling rate should

> be more than 2000Hz (twice the highest *band width*).

>

> A number of published experimental studies would seem to agree that the

> sampling rate does not have to be twice the highest band width, but twice

> the highest maximum frequency of interest

>

> J.R. Potvin & S. Brown (2004) Less is more: high pass filtering, to remove

> up to 99% of the surface EMG signal power, improves EMG-based biceps

> brachii muscle

> force estimates. Journal of Electromyography and Kinesiology 14 (2004)

> 389–399:

> (gain = 1000, input impedance = 10 GX, 10–1000 Hz, CMRR = 115 dB at 60 Hz,

> Bortec, Octopus AMT-8, Calgary, Canada) and sampling rate was 1024 Hz

>

>

> M.G. Feltham et al. (2006) Changes in joint stability with muscle

> contraction measured from

> transmission of mechanical vibration

> Journal of Biomechanics 39 (2006) 2850–2856

> (gain = 1000, band-pass filtered between 10 Hz and 1000 Hz (NL820,

> Neurolog System, Welwyn Garden City, UK), and sampled at 1000 Hz

>

> Does any one have any thoughts on this issue?

>

> Many thanks

>

> Kieran

>

> --

> Dr Kieran Moran

> Biomechanics Research Group

> School of Health and Human Performance

> Faculty of Science and Health

> Dublin City University

> Collins Avenue

> Dublin 9

>

> tel; 00 353 1 700 8011

> fax: 00 353 1 700 8888

>

> kieran.moran@dcu.ie

>

> ---------------------------------------------------------------

> Information about BIOMCH-L: http://www.Biomch-L.org

> Archives: http://listserv.surfnet.nl/archives/Biomch-L.html

> ---------------------------------------------------------------

>