From: Paul Khoury (pkhoury_at_loop.com)
Date: Wed Feb 11 1998 - 01:18:00 EST
On Tue, 10 Feb 1998 11:56:18 -0500 (EST), John Kim wrote:
>On Mon, 9 Feb 1998, Paul Khoury wrote:
>> On Mon, 09 Feb 1998 17:06:15 -0700 (MST), billy_at_MIX.COM wrote:
>>
>> >> >22kHz is a bit high for a 44.1kHz sampling rate - the high frequency
>> >> >limit is one half the sampling rate...
>> >> >
>> >> I don't quite understand (for a change. <g> ). Could you explain that
>> >> again, please?
>> >
>> >In the world of digital sound recording and transmission there is
>> >what's called (I believe) Shannon's law (many years since I was in
>
>I always thought it was Nyquist's theorem. Shannon is usually
>more associated with coding information inside a signal.
>
>> >school) which says the highest frequency that can be digitized is
>> >one-half the digital sampling rate. Thus a 44.1kHz sample rate
>> >system can only go up to 22.05kHz. Then there are anti-aliasing
>> >low-pass filters to enforce this, but they are not perfect so there's
>> >some gradual decline in frequency response before hitting the actual
>> >cutoff point. Since in this case you are only 50Hz away in practice
>> >there will not be much left of your signal after going through this
>> >filter.
>> >
>> >For what it's worth any decent analog tape recorder running at
>> >15 ips (inches per second) or faster would record this quite well.
>> >Or if you could sample at 48kHz you might have a chance of getting
>> >it that way too. I don't know what software might be available
>> >for the Thinkpad at that sample rate though.
>
>I believe the default DOS audio software that came with the 701
>is capable of 48kHz. It's been ages since I booted to DOS so I
>could be wrong.
>
Same here. I'm only using OS/2, so that's currently my only option for recording.
>> >The next step in commercial systems looks like a 96kHz sample
>> >rate - if this eventually finds its way into notebook computers
>> >you'd have no trouble at all, but that's probably a year or
>> >two away.
>>
>> I'm just trying to record a train passing by for a small project,
>> so if I choose 22KHz, is that the sample rate, or the other rate?
>>
>> Or do you think 11KHz might work fine as well?
>
>That depends on what you want to record. If it's just an audio
>effect or looking at a doppler shift, 22kHz (44kHz sampling
>rate) is above the hearing range for most people (depending on
>age and gender, most people limit out at around 14kHz to 20kHz).
>If you've never heard a 16-20kHz sound, it is *very* high
>pitched. 22kHz sampling should be fine. You may even be able
>to get away with 11kHz (depending on if you're listening for the
>whistle or the engine roar).
>
>OTOH, if you want to do a sophisticated analysis of the sound of
>a train passing, you're going to need the higher frequencies a
>44, 48, or 96kHZ sampling rate will provide. The fundamental
>tone of any sound (e.g. a whistle) is its lowest frequency. All
>the harmonics and overtones that give each different sound its
>individual character are higher in frequency, often several
>octaves (one octave is twice the frequency). Also oversampling
>makes it easier to run a Fourier transform (to get it into
>frequency space so you can tell which "notes" make up the whole
>thing).
>
>Boy does this bring back memories... :-)
Just wondering, what do you do for a living? This is getting quite interesting. :)
(Of course, I'm only 18, so I have a long ways to go to get MORE knowledge :) ).
And the project is mainly to get a sound for an alarm clock I'd like to program
(as soon as I figgure out how to program in C).
-- Paul Khoury <pkhoury_at_loop.com> http://pkhoury.dyn.ml.org (If you get an error, then the server is not online)Current system statistics: The Operating System/2 Version is 4.00 Revision 9.029 There are 22 Processes with 83 Threads. This machine's uptime is 0d 22h 35m 54s 640ms.
This archive was generated by hypermail 2.1.3 : Thu Jan 23 2003 - 09:53:56 EST