standard sampling rate and increasing sampling rate

Hi when we sample a analog signal we follow the nyquist theorem where the sampling rate has to be twice the highest frequency component. I read that sampling rates even higher than twice the highest freq is utilized. But can anyone tell me for what reason (for eg say audio wise) is the sampling rate chosen to be twice above highest frequency. I mean does this make the converted data in the audible frequency range or for wat. And what difference would it make in the audio if the sampling rate is increased. I am learning so pardon me if i'm wrong abt anything.
Also i read sampling is nothing but measuring the amplitude at equal intervals and if the sampling of the analog signal is very close to each other then we obtain higher resolution. So what is the disadvantage of having increased sampling rate. Don't we get better resolution??? THank you!!!

According to sampling theorem, the sampling rate must be greater than twice the highest frequency component in oder to be able to reconstruct the original signal from its samples.
Higher sampling frequency is feasible. This facilitates the reconstruction, The only disadvantage is an increase in the storage requirements or transmission rate needed for store or transmit the samples.

i'm not an expert in audio but from my understanding there is not much advantage in sampling audio signals with frequency higher than nyquist (2x Highest frequency you would like to record), you will not gain from it.
However when you measuring RF signals (IF or Baseband after down-converting it) you would like to filter the replicas of he desired signal replicated at the sampling frequency (fs, 2xfs, 3xfs etc.). For the nyquist sampling this will fall into Adjacent channel and will required a very good filter, linearity and Phase Noise for your system (complicates everything). In this case it is better to use oversampling.

RS

Hi,

Nyquist is theoretical somehow.

You need a sampling rate with more than twice the highest frequency component.

But how do you know the "highest frequency component" .. let's say in an audio signal.
Somewhere it is said that audio frequencies go up to 20000Hz.
So you need a sample frequence higher than 40kHz. For the audio CD standard 44.1kHz is chosen.

Now imagine you do some interview recordings jn a factory with machines generating higher frequencies than the 20kHz. An ultrasonic cleaner for example.
Let's assume it's frequency is 38kHz pure sine.

If there are no filters then the (inaudible) 38kHz become to audible (44.1kHz - 38.0kHz) = 6.1kHz.

To avoid this you need anti_alias filters.
The filters need to pass ALL audio frequencies without (much) attenuation. But it needs to attenuate ALL above fs/2 with high attenuation.
It means pass up to 20kHz, then attenuate all above 22.05kHz
Such sharp filters are about impossible to build as analog filters.

To ease the filter design one uses oversampling technique with much higher sampling rate.
Lets say 4 x 44.1kHz = 176.4k. Then the analog anti alias filter still needs to pass frequencies up to 20kHz, but attenuate above 88.2kHz.

The digital values from the ADC additionally are digitally filtered to attenuate frequencies above 22.05kHz (digital filters are more flexible and more accurate) and only every 4th output value is "recorded" to get the desired data_rate of 44.100Hz.

Klaus

KlausST said:

.................
The digital values from the ADC additionally are digitally filtered to attenuate frequencies above 22.05kHz (digital filters are more flexible and more accurate) and only every 4th output value is "recorded" to get the desired data_rate of 44.100Hz.

Click to expand...

To clarify, that "4th output" is the the decimated output from the digital filter which has reduced the sample rate by 1/4 as part of the filtering process.

Thank you so much for the answer!!! Understood it really well. But one question. So the anti alias filter passes freq upto 20k and attenuates freq above 88.2k. What happens to the frequecies inbetween 20k and 88k???

Hi,

What is between white and black? gray.

A filter is not ON (= pass) or OFF (= attenuate), it is a curve
If you simply look at any filter chart you will see this.
There is a curve between... but the shape and steep depends on filter_order and filter_characteristic.


Klaus