Just thought some more and realised that a non-linear DAC would also create IM between signal and images. In fact, this might be a good test for non-linearity. An 18kHz tone (for example) would give a first image at 26.1kHz. Send the output from the DAC chip into a low pass filter to get rid of signal plus image. Look for 8.1kHz 2nd order IM, or 9.9kHz 3rd order IM.
If you want a 12kHz sampling frequency then all you need to do is use a 5-6kHz brickwall antialiasing filter. You will lose all the higher overtones but it will still sound roughly like music. People who have never heard hi-fi or live music might not notice the difference. To improve the sound balance you could also raise the LF rolloff to around 50Hz.
I believe Shannon proposed this easy way to record music some years ago.
I believe Shannon proposed this easy way to record music some years ago.
Do you know if this is something I could do with Audacity?
Can I simply browse through the menus until I find something called 'brickwall antialiasing filter' and set that to about 5 kHz, then press 'record'?
Easier than what way of recording music? Shannon? -You mean this man?:
Claude Shannon - Wikipedia, the free encyclopedia
I just thougt it would be interesting to hear if music recorded with an audible sample frequency sounds like music played on speakers with the tweeters missing or if it sounds more like it is distorted. I remember that some, but far from all, CDs from when they were a brand-new thing in retail and I was a teenager with better hearing sounded a bit synthetic with some of the high-frequency sounds, especially cymbals. Probably just bad sound engineering, I guess?
Last edited:
The former, assuming you follow the sampling theorem (i.e., filter out all components higher than fs/2) and don't do something stupid like leaving out the anti-imaging filter.
Perhaps it is not generally realised that if you have used a landline telephone then you have almost certainly listened to sounds sampled at 8kHz, and probably coded using 8 bits. 16 bit coding would reduce the quantisation distortion.
The basic story is that nothing different happens if the sampling frequency is within the audible range. There are no tones, noise etc. as the filters removes all these. The only change is that imperfect filters may allow through audible artifacts instead of ultrasonic ones.
The basic story is that nothing different happens if the sampling frequency is within the audible range. There are no tones, noise etc. as the filters removes all these. The only change is that imperfect filters may allow through audible artifacts instead of ultrasonic ones.
I'll trust your useful anwer, SY, and live happily with the knowledge that the music would sound muffled and not distorted. I also better refrain from fiddling around in the menus of Audacity and try to make sense out of strange English technical terms like 'anti-aliasing' (who is using false names?) and 'anti-imaging' (to erase what pictures? -Aren't we dealing with sound here?). I probably would end up doing something stupid! 
Many of the older speech synthesizer chips used 8, 10 or 12 KHz for the sample rate since memory was precious then. They used 8 bit samples too. There was an old music sampling keyboard the Casio SK1 that ran 8 bit samples at about 20 KHz. The built in piano sample sounded reasonably believable too.
Anti-aliasing filter prevents aliasing. An alias is something pretending to be something else, so aliasing is the process by which a high frequency sound (above the Nyquist limit fs/2) ends up looking exactly like a low frequency sound. Aliasing happens at the ADC stage in the recorder.Drumhead said:
Anti-imaging filter prevents imaging. An image is something reflected in a mirror. It may be a different size or inverted. A DAC image is a reflection (above fs/2) of a sound below fs/2. Imaging happens at the DAC stage in the player.
People often say 'alias' when they mean 'image'. This both creates, and is perhaps derived from, confusion.
DF96, let me se if I understood this correct:
So aliasing is kind of an unwanred 'undertone' the A/D converter creates , sort of like what a bass octaver pedal is doing? An octaver as the ones I know adds an artificial tone one octave below the input and sounds pretty scary if you sneak it into the line of a mike when a girl speaks or sings :-D
And imaging is like an unwanted overtone, created in the D/A converter?
So aliasing is kind of an unwanred 'undertone' the A/D converter creates , sort of like what a bass octaver pedal is doing? An octaver as the ones I know adds an artificial tone one octave below the input and sounds pretty scary if you sneak it into the line of a mike when a girl speaks or sings :-D
And imaging is like an unwanted overtone, created in the D/A converter?
Yes, sort of. An alias will always be lower in frequency than the original signal. An image will always be higher in frequency than the original signal. Neither will be harmonically related to the original signal (except sometimes by numerical accident), so in that respect different from undertones or overtones (which are usually approximately sub-harmonics or harmonics).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.