Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: What Is Quantization (Read 4126 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

What Is Quantization

I've searched trough this forum but seem that i cant' find the definition of quantization.
Can please someone explain what is quantization?
[ Commodore 64 Forever...! ]

What Is Quantization

Reply #2 –
Thank you for the rapid answer!
[ Commodore 64 Forever...! ]

What Is Quantization

Reply #3 –
Quantisation occurs when you need to store data into a digital form which can only take certain values. For example in most cases audio has 16bit quantisation which means each sample can only take 65536 values. Let's say the scale we sample audio at goes from -32768 to 32767 and the actual sample value measured is 12345.6789, this will be quantised by the ADC to 12346 before being stored in digital form.

During perceptual encoding, samples are usually transformed and filtered and the output of this filters give you values with fractions again, so when these output is stored, the values have to be quantised again.

The challenge in perceptual encoding is to assign more bits to more audible portions of the audio (e.g. mid frequencies) and less bits to less audible portions of the audio (e.g. very low and high frequencies), so you will see lower quantiser values/scale factors for bands which are more audible. For example, at 5kHz, coefficients may be sampled at 0 1 2 3 4 5 6 7 8 9 10 (quantiser = 1) while at 15kHz, coefficients may be sampled at 0 5 10 15 20 25 30 35 (quantiser = 5).

Enough rumbling.. heh

What Is Quantization

Reply #4 –
Or if you want a very basic answer, it's like the number of digits after the point.

What Is Quantization

Reply #5 –
How about this 'interpritation'.

Say you have two axis: one is time and other is signal level.

So you 'split' (to make it more visual) both axis with some step. It gives you somewhat of 'grid'. Then you watch which point of that grid is the closest to your current signal value. (If linear quantisation used.)

So it'll be sampling for time axis and quantisation for signal axis.

If the whole signal range is splitted to 2^16=65536 parts you need to determine which one of 65536 possible values to associate with every sample (so 16 bits per sample needed then to store information about signal). The more bits you use the less rounding (quantisation noise) appears.

Damn it was fun to write. Hope my teachers wouldn't laugh at me if they'll happen to read this. Altough I'd never use such language while doing some exams.

What Is Quantization

Reply #6 –
Quote
so you will see lower quantiser values/scale factors for bands which are more audible


Even though I have continously read over research documention from time to time I am still quite puzzled as to where exactly the mantissa and the exponent fit into this category with the scale factors. I know that in some form or another they all related. I am just not sure in to what category they fit in? what is actually being quantized if we where to speak in terms of the mantissa?
budding I.T professional

What Is Quantization

Reply #7 –
I think the simplest way to define it is rounding to the nearest applicable value in a certain range.

Ruairi
rc55.com - nothing going on
 

What Is Quantization

Reply #8 –
Quote
Even though I have continously read over research documention from time to time I am still quite puzzled as to where exactly the mantissa and the exponent fit into this category with the scale factors. I know that in some form or another they all related. I am just not sure in to what category they fit in? what is actually being quantized if we where to speak in terms of the mantissa?

Not sure about mantissa but exponent may refer to MP3 quantising with a 4/3 exponent, meaning instead of the scale going 0 1 2 3 4 5 6 7 8 9 10, it goes something like: 0 1 2^(4/3) 3^(4/3) 4^(4/3) etc...