KarlXII skrev:Bra jobbat, Matereo.
Själv tyar jag inte plöja fakta för tillfället.
Men ovanstående kanske det finns något i?
Jitter är sannolikt mer störande än man kan tro.
Paradisean lät faktiskt betydligt bättre när man körde usb direkt från datorn via ASIO - vilket ska ge väldigt lågt jitter - än när man matade den digitalt via RCA. Samma sak med DAX - den lyfte när den blev matad med ASL jämfört med vanlig koax.
Ja de som jämfört själva och pillar med DACar verkar tycka att klangen blir mörkare(svartare? kan ha med avsaknad av fasproblem att göra) och känns renare. Nu har ju ingen vettig människa pengar eller prioritet att köpa en Audionote Dac5, men många som testat säger att den låter bättre än nästan något annat.
Så här beskriver en kille som bygger Dac'ar om sin variant av NOS tillämpning.
"1) It runs at 44.1 KHz (NON-oversampling)
2) It uses NO (digital) brickwall filtering at all.
3) It causes NO "ringing" as a digital chebychev filter does.
4) The output signal is "clean" enoug to leave out the "external filter" that IS necessary with a standard NON-oversampling DAC (check your NOS-DAC diagram).
5) The phase response is LINEAR
6) It uses NO decimation (no addittional noise)
7) Amount of samples is increased (smoother signal, mirror image is moved upwards to a higher frequency region, ultrasonic interference is greatly reduced)

Analog resolution increases (smoother steps as there are intermediate steps now, theoretically 19 bit resolution with an octal D-I DAC version)
>The aim of this tread is to build an ultimate NON-oversampling DAC (improve the concept to the max) I think the D-I DAC is a good start . I am not claiming this DI-DAC is perfect but what is?
If you look at the oscillograms I posted, it's difficult to admit there is no improvement made (consider that the oscillograms show the quad DI-DAC so resolution can go even Higher).
Drawbacks:
As poobah already indicated, (linear) interpolation has it's drawbacks. You try to theoretically reconstruct something based on 16 bit samples. I agree, you never achieve the original high resolution studio quality. But you can try to come close.
A reply to the 18 KHz "issue":
A 18 KHz (20KHz) "sinewave" looks as a squarewave in the standard NOS-dac as there are only about 2 samples. And squarewaves don't sound nice.
The D-I dac outputs a triangle wave using linear interpolation and a sinewave with non-linear interpolation. You must admit that a triangle wave or sinewave sound better.
By using binary coded non linear steps (different output gain at each I/V stage) the very high frequencies could be shaped like perfect sinewaves. As the audio signal consists of numerous sinewaves (fourier theorema) 20KHz being about the highest recorded with 44.1 KHz samplerate. Using non linear amplitude steps seems logical and worth investigating.
But here is an important question to you all: Who can hear the difference between a 20 KHz sinewave and a 20KHz triangle wave as they sound almost the same.
"