Tweet
I have a query about the amount of phase distortion or
lag that occurs when using low pass filters of the
first or low-order to produce the linear envelope in
EMG signal processing. Specifically how much will this
effect EMG onset times.
EMG studies often contain descriptive phrases such as
'a sixth order low pass filter was used to prevent
temporal signal distortions', but I have been unable
to find any literature that specifies by how much a
signal can or will be distorted temporally when low
order filters are used.
In discussions with a signal processing expert, I was
advised of the following:
"When a non-sinusoidal signal goes through a filter
its waveshape is distorted, so defining "delay" is not
straightforward. You will be familiar with the effect
of an RC on a step waveform - if you define delay as
the time to reach 50% amplitude you get a different
answer than if you look for 90% amplitude.
For a 50Hz first-order Butterworth (=RC) the
time-constant is 3.183ms. For a sinusoid at the
cut-off frequency of a first-order Butterworth, the
phase shift is 45°. For a 50Hz filter this is a delay
of 2.5ms. For lower frequencies the delay is slightly
more, 3.0ms at 22Hz rising to 3.183ms at 0.1Hz. For
higher requencies it's less, 2ms at 81Hz, 1.5ms at
127Hz. For a higher-order filter these times are
simply multiplied by the order.
So the overall picture is that for a non-sinusoidal
waveform our 50Hz first-order Butterworth will delay
gradual slopes by about 3ms but steep edges will be
delayed only about 2 ms. This applies equally to both
channels of course, except to the extent that their
waveforms are different so, as I stated earlier, it
all depends on how you define delay."
To me, the above calculations appear to show that the
temporal delay using a first order filter is only 2 to
3 ms, and it applies to all channels or muscle sites.
For those interested, I am using a digital 50Hz
lowpass filter, first order Butterworth, looking at
the difference in onset times between muscles.
Can anyone confirm or refute this, or supply any
literature references to back this up? I will of
course collate and publish all responses.
Thank you
John
Dr John Dixon
Research Fellow and Laboratory Manager
Teesside Centre for Rehabilitation Sciences
University of Teesside
------------------------------------------------
The James Cook University Hospital
Marton Road
Middlesbrough
TS4 3BW
UK
Telephone/Fax: 01642 854324
E-mail: john.dixon@tees.ac.uk
__________________________________________________ _________
ALL-NEW Yahoo! Messenger - all new features - even more fun! http://uk.messenger.yahoo.com
-----------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
-----------------------------------------------------------------
lag that occurs when using low pass filters of the
first or low-order to produce the linear envelope in
EMG signal processing. Specifically how much will this
effect EMG onset times.
EMG studies often contain descriptive phrases such as
'a sixth order low pass filter was used to prevent
temporal signal distortions', but I have been unable
to find any literature that specifies by how much a
signal can or will be distorted temporally when low
order filters are used.
In discussions with a signal processing expert, I was
advised of the following:
"When a non-sinusoidal signal goes through a filter
its waveshape is distorted, so defining "delay" is not
straightforward. You will be familiar with the effect
of an RC on a step waveform - if you define delay as
the time to reach 50% amplitude you get a different
answer than if you look for 90% amplitude.
For a 50Hz first-order Butterworth (=RC) the
time-constant is 3.183ms. For a sinusoid at the
cut-off frequency of a first-order Butterworth, the
phase shift is 45°. For a 50Hz filter this is a delay
of 2.5ms. For lower frequencies the delay is slightly
more, 3.0ms at 22Hz rising to 3.183ms at 0.1Hz. For
higher requencies it's less, 2ms at 81Hz, 1.5ms at
127Hz. For a higher-order filter these times are
simply multiplied by the order.
So the overall picture is that for a non-sinusoidal
waveform our 50Hz first-order Butterworth will delay
gradual slopes by about 3ms but steep edges will be
delayed only about 2 ms. This applies equally to both
channels of course, except to the extent that their
waveforms are different so, as I stated earlier, it
all depends on how you define delay."
To me, the above calculations appear to show that the
temporal delay using a first order filter is only 2 to
3 ms, and it applies to all channels or muscle sites.
For those interested, I am using a digital 50Hz
lowpass filter, first order Butterworth, looking at
the difference in onset times between muscles.
Can anyone confirm or refute this, or supply any
literature references to back this up? I will of
course collate and publish all responses.
Thank you
John
Dr John Dixon
Research Fellow and Laboratory Manager
Teesside Centre for Rehabilitation Sciences
University of Teesside
------------------------------------------------
The James Cook University Hospital
Marton Road
Middlesbrough
TS4 3BW
UK
Telephone/Fax: 01642 854324
E-mail: john.dixon@tees.ac.uk
__________________________________________________ _________
ALL-NEW Yahoo! Messenger - all new features - even more fun! http://uk.messenger.yahoo.com
-----------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
-----------------------------------------------------------------