So everyone knows that higher sampling rates in a digital audio system == more better. But what happens if we use a low sampling rate...say, 22 kHz or even 11 kHz to record some audio? Describe in as much detail what will happen to the recorded audio and what it will sound like. Is it possible to fix it later? Also, what is the generally accepted lowest sampling rate one can use for full audio fidelity and why?
Low Sampling Rates
- Thread startermbenonis
- Start date
TimmyP1955
Well-Known Member
At first the highest frequencies will roll off or start to sound "bad". This will increase as the sample rate decreases. Eventually the entire spectrum will start to sound "choppy" like a bad cellular connection (or like when someone talks through a fan).
Lost information is lost forever. You can fake it, but you can never recover it. (Although for reasons even the equipment manufacturers don't understand, upsampling can sometimes make it sound better.)
Lost information is lost forever. You can fake it, but you can never recover it. (Although for reasons even the equipment manufacturers don't understand, upsampling can sometimes make it sound better.)
Soxred93
Active Member
44.1 khz is the generally accepted standard for sampling rates. It is also CD quality. If you go a bit lower that, it may sound normal to the untrained ear, but professions will notice a difference. When you get to around 11 khz or lower, it starts to sound choppy and if someone is stuttering. As was said earlier, a bad connection. It would be like if you could plug your ears 11,000 times a second when someone was talking, it would sound like that.
On the other side, if you have a higher rate, the closer to accurate sound you have. It sounds better, but it also takes up a huge amount of space. 44.1 khz is generally ysed because it's a good compromise between quality and size.
On the other side, if you have a higher rate, the closer to accurate sound you have. It sounds better, but it also takes up a huge amount of space. 44.1 khz is generally ysed because it's a good compromise between quality and size.
waynehoskins
Active Member
Hint continued: "y"
Well, you have to consider that the average frequency range of human hearing is 20Hz-20kHz. The sampling rate of of digital audio is directly related to the highest frequency that can be recorded cleanly. In fact, the sampling rate needs to be at least double the highest frequency you wish to record. So, in order to record a 20kHz sound you need at least a 40kHz sampling rate. This is why, as you lower the sampling rate you start to lose the higher frequencies.
The other thing that starts to happen when you lower your sampling rate is everything starts to sound more square. This is because you have more time between samples so your waveform becomes step-like.
There are some interesting things that you can do when you have the option to work with multiple sample rates. Consider this situation that I once found myself in: I had to archive to CD a stack of old 1/4" tape. All the tapes had been recorded at 7 ips, but the only deck we still had could only play at 14 ips. So, I set up the computer to record at a 48kHz sampling rate and then changed it to playback at 44.1kHz, essentially bringing the tape back into the correct speed. It was fun to listen to the chipmunks sing opera!
The other thing that starts to happen when you lower your sampling rate is everything starts to sound more square. This is because you have more time between samples so your waveform becomes step-like.
There are some interesting things that you can do when you have the option to work with multiple sample rates. Consider this situation that I once found myself in: I had to archive to CD a stack of old 1/4" tape. All the tapes had been recorded at 7 ips, but the only deck we still had could only play at 14 ips. So, I set up the computer to record at a 48kHz sampling rate and then changed it to playback at 44.1kHz, essentially bringing the tape back into the correct speed. It was fun to listen to the chipmunks sing opera!
Should that not have been record at 96k and playback at 48k? Then you have doubled the length of the file and hence are back to the original timing?
The answer to minimum is DXD isn't it?
Well, that would have been more exact, but I don't think that our equipment at the time supported recording at a 96k sampling rate. When you consider all the variables working with tape payed back at double speed, that extra 0.1k doesn't really amount to much.
Nope, sorry! Strictly speaking, you don't "lose" anything (though things *will* distort)! In fact, it would theoretically be possible to record a 20 kHz wave with a 5 kHz sampling rate (though some post processing would be needed). Anyone care to explain how that would work?
Um my maths would put half of 48KHz at 24 kHz, a little more than 0.1k out... I suspect we may be talking about different things here?
Nope, sorry! Strictly speaking, you don't "lose" anything (though things *will* distort)! In fact, it would theoretically be possible to record a 20 kHz wave with a 5 kHz sampling rate (though some post processing would be needed). Anyone care to explain how that would work?
Huh? Me thinks I shall be learning something in this...
Sorry, but I must disagree with you.Nope, sorry! Strictly speaking, you don't "lose" anything (though things *will* distort)! In fact, it would theoretically be possible to record a 20 kHz wave with a 5 kHz sampling rate (though some post processing would be needed). Anyone care to explain how that would work?
If I take a 20 Khz sine wave, and sample it every 0.2 mS (5 KHz rate) then all samples will be at the same point on the wave, and the result will be a flat line. 100% of the information would be lost. It would also be impossible to distinguish between a 20 KHz, 10 KHz, and a 5 KHz wave, as they would all be flat lines. Similarly, if sampled at 20 KHz the result for a 20 KHz wave would be the same. Which is why Mr. Nyquist (along with Mr. Shannon) say that you need to sample at 2x the frequency.
Now, are you perhaps confusing this with the encoding rate? That can certainly be (and these days often is) considerably less than the sampling rate.
Edit: Also, your typical sampling systems have a Nyquist (there he is again) filter before the sampling stage, so your 5k sampling should have a 2.5k filter... to prevent things like a 6 kHz signal sounding like a 1 kHz signal.
-Fred
Last edited:
I stand corrected. Let's say I put in a 21 kHz signal and sample at 5 kHz. I argue that I can recover it with some prior information and processing. That said, you are exactly on the right track here. with regard to higher frequencies than 1/2 the sample rate being recorded. Let's see where this goes.
PS - definitely not thinking about encoding rate here...in fact, I'm ignoring quantization.
Last edited:
Ah, there's the missing piece... and additional information outside of the actual sampling changes things considerably. In a sense, that's how low-rate/lossy compression algorithms work. But, exactly what additional information is needed, how is it collected, stored, transmitted, etc.? Typically, in order to get that additional information, you'd need to sample at a higher rate. Certanly, if I tell you that I'm giving you a signal between, say, 20 and 22 KHz, you will be able to reproduce it with a 5 KHz sample rate. But, if I say it's between 10K and 50K, you won't get enough information from just your 5kHz sample to be able to reproduce it.
-Fred
Nope, you are right, I am just tired, in tech, had a headache... My math was bad.
Similar threads
