20bit DAC - where does the resolution come from ?

[Bernhard]

Hi,

from where does a 20bit DAC chip take those 4 extra bits ?

Interpolation from oversampling in the digital filter ?

Isn't resolution lost at high frequencies ?


Bernhard

[borges]

You are right about the oversampler. The input signal to an oversampler may be 16bits/44.1 while its output may be 20bits/352.8.

In one implementation, the digital oversampler keeps the original samples and fills in 7 interpolation values between two originals. With 16bit input, the originals are simply shifted 4 bits up while the generated interpolation samples are truncated or rounded down to 20 bits.

If the DAC doesn't feature a built-in analog filter, it shouldn't care about the data source, and thus be capable of reproducing any 20-bit signal you throw at it. (As long as you don't exceede its maximum update frequency etc.)

Greetings,
Børge

[peranders]

If you have 16 bit in you can't get more out in total quality but you can get better filtering, more ideal with some interpolation tricks. CD-qualilty stays CD-quality no matter what you do but you can come closer to the theoretical limit.

[Bricolo]

it's like reencoding a 64kbps MP3 in 320kbps...

[Bernhard]

So a 20 bit DAC like PCM63 converts lots of samples that have values which are somewhere between original 16bit values.

Beside positive effects regarding filtering etc. , doesn't that reduce resolution / loose detail ?

If a picture is interpolated, quality does not get better.
It can be blown up without blowing up pixels, but some detail is lost when 2 pixels were black and white and the space between is filled with shades of grey after interpolation.
Also after resizing to the original , detail is lost.

[Werner]

In a digital filter (i.e. the oversampler) the actual 16-bit signal is convolved with the filter coefficients. These coefficients themselves are described with an accuracy of , say, 16 up to 32 bit. The resultant stream is meaningful in 32-48 bits, and these are prior to output reduced to 20 bits, because, of course, no DAC in our universe can reproduce these.

Doing digital filtering on a 16bit signal while restricting the output to 16 bit would indeed harm the information in the stream, i.e. the resolution.

Outputting an original 16 bit stream at 20 up to 24 bits is almost lossless.

[Bricolo]

does this mean that most of the cd players, that use 16 bits dacs with oversampling, are loosing resolution?

[CheffDeGaar]

Hi Bernhard,

Have you read this thread ?

Asynchronous Sample Rate Conversion

It should answer your questions

[Bernhard]

From the above thread:

I know how disturbing this sounds. But we must remember a couple things : One, a bandlimited signal is NOT free to do whatever it pleases between Nyquist samples ... in fact, it's pretty smoothly varying between those samples. And two, the more samples we can calculate beyond the Nyquist rate, through a very accurate interpolation process, the more adjacent samples will look the same. These are simple, indisputable facts of life.

But no matter what, grabbing the most recently calculated interpolated sample, instead of the exact sample required at Fs_out, will introduce error. The question is ... How far do I have to interpolate, by what integer, so that grabbing the most recent sample will give me acceptably LOW error ??? We can never achieve perfection here, but we can get ARBITRARILY CLOSE ... simply because the higher I interpolate, the lower the error

So, there is error and in case of 8 x os the error will be 1/8 LSB if a nearest neighbour sample is taken instead of the right one ?

[rfbrw]

Beware of comparing apples with oranges. That thread was about Asynchronous sample rate conversion. What takes place in an oversampling filter like the DF1704 is different.