sample frequency and upper limit

[Gaston]

the isd voicechips come with a variable sample rate. when you use the higher sample rate ,lets say 12k instead od 4k, does that mean that you can record a higher frequncy on it than the 4k ? or does it just give you better quality of sound? by the way i finaly got my recorder to work. it took about two weeks of spare time

 

[Nigel Goodwin]

A higher sampling frequency gives you higher frequencies, this is what makes the sound higher quality.

 

[Gaston]

so does that mean i can record a 12k hz sound on it? were as at 4k i can record only a 4k hz?

 

[Nigel Goodwin]

so does that mean i can record a 12k hz sound on it? were as at 4k i can record only a 4k hz?

Try googing for 'nyquist theorem', that explains it - but essentially your maximum frequency must be less than half the sampling frequency - it's essential to ensure this with a suitable low pass filter.

 

[ericgibbs]

so does that mean i can record a 12k hz sound on it? were as at 4k i can record only a 4k hz?

hi,
Your sampling frequency should be at least 4 times the highest frequency you want to sample. I would aim for 10 times greater sampling.

The easiest way to visualise this is to plot a single wave of say 4KHz and overdraw a 4KHz square wave,
then draw a 16KHz square wave, the difference becomes very clear.

Regards

 

[Nigel Goodwin]

hi,
Your sampling frequency should be at least 4 times the highest frequency you want to sample. I would aim for 10 times greater sampling.

It's a good job you weren't the designer for CD audio!

 

[Hero999]

4kHz would bearly produce voice bandwidth.

12kHz would sound crap but would be alright for voice.

If you really want good quality you need 40kHz to reproduce the full 20kHz audio bandwidth.

Some people don't realise this, you can buy cinema systems that sample at 192kHz but there's not point.

 

[ericgibbs]

It's a good job you weren't the designer for CD audio!

A common used standard for 'voice quality' audio digitising is 44/48KHZ, and often a range of 200/300Hz thru 5KHz is good enough for speech
[Rem: old phone bandwidth].

IIRC some of the CD quality is at 128KHz, so you are probably right "It's a good job you weren't the designer for CD audio!"

Regards

 

[Hero999]

What are those CDs aimed at cats and dogs?

48kHz is more than enough to account from the anti-aliasing filter's cut-off freqeuncy.

It's probably better to use high bit rates than smaple rates past a certain quality but onece you're sampling into the noise margin there's no point. 24-bit audio at 48kHz will sound just as good or bad as 32-bit audio sampled at 128kHz.

 

[Nigel Goodwin]

A common used standard for 'voice quality' audio digitising is 44/48KHZ, and often a range of 200/300Hz thru 5KHz is good enough for speech
[Rem: old phone bandwidth].

IIRC some of the CD quality is at 128KHz, so you are probably right "It's a good job you weren't the designer for CD audio!"

CD is sampled at 44.1KHz, for a 20KHz bandwidth - like I said previously, consult nyquist! - it's all his fault!

 

[Hero999]

Another thought, as audio has such a large dynamic range would storing it in a floating point format give good quality whilst not using too much space?

Also aren't sound waves symetrical? Therefore would it not be acceptable to only store one half of the waveforum then duplicate it on the other side when it's reconstructed?

 

[Nigel Goodwin]

Another thought, as audio has such a large dynamic range would storing it in a floating point format give good quality whilst not using too much space?

No, floating point takes more space and is less accurate.

Also aren't sound waves symetrical? Therefore would it not be acceptable to only store one half of the waveforum then duplicate it on the other side when it's reconstructed?

No they aren't, nothing like it!.

 

[ericgibbs]

the isd voicechips come with a variable sample rate. when you use the higher sample rate ,lets say 12k instead od 4k, does that mean that you can record a higher frequncy on it than the 4k ? or does it just give you better quality of sound? by the way i finaly got my recorder to work. it took about two weeks of spare time

Hi Gaston,

Came across this weblink that may help.
http://www.vias.org/simulations/simusoft_nykvist.html

Download and unzip.

As I said the practical sampling frequency is at least 4 times the frequency you want to sample, the author says 5 times.

Also look at the parent page.

Sorry Nigel,,, its another 'simulator' !!!!

Regards

 

[Hero999]

So if the author says it then it must be true.

That's nonsense, sampling more than double the bandwidth is pointless. Your ears can't tell the difference between a 20kHz sinewave and a 20kHz squarewave assuming you're young enough to hear 20kHz, even so the anti-aliasing filter will get rid of most of the harmonics anyway.

 

[ericgibbs]

So if the author says it then it must be true.

That's nonsense, sampling more than double the bandwidth is pointless. Your ears can't tell the difference between a 20kHz sinewave and a 20kHz squarewave assuming you're young enough to hear 20kHz, even so the anti-aliasing filter will get rid of most of the harmonics anyway.

Hi,
In order to check my thoughts and understanding of this point, have checked my text books and trawled the web.

And ALL say the same thing:
It should be understood that the Nyquist frequency is an absolute maximum frequency limit for an ADC, and does not represent the highest practical frequency measurable. To be safe, one shouldn't expect an ADC to successfully resolve any frequency greater than one-fifth to one-tenth of its sample frequency

I think the point we are at a difference is, I am not talking about audio reproduction but an engineering signal digitisation/measurement.
If I have a analog signal coming from say a transducer and I want to convert it to digital I would sample it at least 4 times the max frequency rate.

 

[audioguru]

Over-sampling is used in CD players so that the lowpass filter can have a more gradual slope.
When the filter has a very sharp slope then it rings badly.

 

[Hero999]

ericgibbs,
You're right, thinking about it more. . .

To an extent oversampling is good because sampling at 40kHz wouldn't give you enough resolution for you to be able to distinguish between 19kHz and 20kHz; also the maximum phase accuracy at 20kHz would be 180°. Not that any of this matters with audio, the distance between the ears is many wavelengths and the difference between 19kHz and 20kHz is largly irrelevant.

 

[ericgibbs]

hi hero,
At my age, the top end of my hearing range has long gone!

Also I have tinnitus in both ears since I damaged them in the 1970's playing around with a high powered SMPS. [take care if you are frequency tweaking one of those damn things]

I was talking in terms of electronic measurent rather than audio work

IIRC, I believe the CD -3db low pass roll off is around 44KHz for CD's.

Im sure AG will produce a graph.

 

[eblc1388]

There is a mix up on sampling an audio signal for CD recording and sampling the data on CD in order to produce an audio signal.

The data rate on CD pretty much control how fast one can sample the audio. If one uses higher rates than 44.1KHz, one simply have to reduce the data rate at the production stage anyway in order to fit them on the CD track.

I might be wrong but I'm pretty sure "oversampling" applies to how the decoder inside the CD player samples the data from the CD data stream. It will not give extra frequency coverage than is already recorded onto the CD track.

 

[audioguru]

The clock for a CD is at 44.1kHz so the lowpass filter must attenuate it by at least 100dB (one hundred-thousandth) but not attenuate 20kHz. Nearly impossible.
Oversampling multiplies the clock frequency during playback so that the lowpass filter can have a more gradual slope.

I couldn't find the frequency response curve for a CD. It is just a straight line.
The spec's for an ordinary Sony CD player lists its frequency response from 2Hz to 20kHz plus or minus only 0.5dB.