Hi
I get very concerned when I see statements like 'As the bandwidth of your
microphone is 50 kHz, the 100 kHz sample rate is adequate'. If a signal is
limited so that there are no components of detectable amplitude above 50kHz,
then a 100kHz (Nyquist) sampling rate is adequate, bu this is a theorteical
limit which implies that the signal and restoration filters have perfect
cut-off. In the real world this is not true, it is certainly not true for a
microphone which tends to have a 12dB/octave drop above the specified 3dB
frequency. Aliasing can be a very insidious effect(1) which can give rise
to invalid conclusions about signal structure and it is important to
understand the limitations of equipment, especially with digital techniques.
As a rule of thumb, a sampling frequency of 3 times the maximum signal
frequency is a safer bet, but if you need slower rates then you really have
to consider the rates of all the filters in the system, including the
microphone and amplifier responses, and both on the sampling and the
restoration sides of the system.
Justin Halls
1) Susan C. Silver and Justin A. T. Halls 1980
"Recording the Sounds of Hydropsychid Larvae - A Cauionary Tale"
J COMP PHYSIOL A, 140 159-161 1980
-----Original Message-----
From:
On Behalf Of Cornell III,
Howard M
Sent: 29 August 2006 18:51
To: Anna Hall;
Subject: RE: Recording Question
Anna,
I looked up the USB Oscilloscope specs. The Oscilloscope function takes a
million 12-bit samples per second, and the datalogger takes 100,000 samples
per second. As the bandwidth of your microphone is 50 kHz, the 100 kHz
sample rate is adequate. Although the USB 2.0 has a high data rate, there
is some unpredictable bus overhead that may interfere with
using the bus continuously. Yet many audio devices, with several
tracks of 96 kHz data, have solved this problem. Will this bother you?
Perhaps you can place a waveform generator at 50 kHz as an input and see how
much of the data survives. Also see if your datalogging software has a
large enough sample storage buffer to overcome times that Windows takes over
the computer (a few milliseconds every second): that could be part of the
same test, I guess.
I have sent an email to USB Instruments to see if their datalogging has an
antialiasing filter; but they are in Europe and may not have been up waiting
for my inquiry. You will need to prevent aliasing, as any data component
with a frequency so high that it does not allow two samples per sampling
period will show up at a lower frequency; but you probably knew that.
Howard Cornell
-----Original Message-----
From:
Behalf Of Anna Hall
Sent: 25 August 2006 16:54
To:
Subject: recording question
Hello all,
I am recording harbour and Dall's porpoise echolocation clicks to my
computer using a Cetacean Research Technology C54XRS hydrophone connected
through a DS1M12 USB Instruments Oscilloscope.
I had trouble with my Toshiba Satellite which dropped 2.4% of data at a 2us
sampling rate.
Has anyone had experience recording acoustic data using a Toshiba Tecra
A7 unit?
The Tecra spec's are 2.0 GHz Intel Core Duo processor, 100 GB hard drive,
512 MB RAM, but can be expanded to 4 GB.
Thanks for any advice.
Sincerely,
Anna Hall, PhD candidate
Marine Mammal Research Unit
University of British Columbia
|