Non-Oversampling DAC

[lucpes]

Originally posted by Bernhard
...the dynamics picture, does it show a low level (-67dB) 1 kHz sine ?

Yep, I used lower value gain resistors in the I/V stage than specified by the designer - what I had around - so the voltage out is a tad bit low (7dB lower than the sound card input level - I guess around 1V) but I won't change them now as both measurements and sound (on either CDP transport or soundcard) are spectacular.

I wouldn't pay that much attention to the freq graph, those values do not correlate with the sound quality;

I also guess that the soundcard also uses 24 bit internal for recording no matter what the output bitrate in order to improve the accuracy.

Regarding the DAC itself, I have to close the box, put a lock on it and throw away the key in case I get the tinkering itch - it's a finished item from this POV

[lucpes]

Originally posted by Terry Demol


The AD1892 will impose a digital brickwall filter even when
resampling at 1:1, such is the nature of how ASRC's work.

The idea of 0 x OS is to get rid of brickwall filter.

Cheers,

Terry

Well, I hope I get it right, as far I as I understand it:

The AD1892 is the SPDIF receiver chip, the 1:1 resampling (recklocking) here would be needed to insure that the DAC is not keen at all on source jitter and uses it's own low jitter master clock - I tested this by accident - at first my SPDIF cable had the ground connection broken and the thingie stilll sounded pretty good. I had some reserves too on this, but the AD1892 specs + the very low jitter clock are good enough to ensure that the 1:1 resampling does not degrade the sound at all - see the measurements.

So the AD1892 spits out plain 1x44.1kHz I2S (which should be in 90% of the cases better than directly derived from the incoming jittery SPDIF/TTL signal) to the AD1865 dac; AD1865 I out and the I/V stage have no filtering of their own, just a miller comp cap in the I/V stage to decrease its bandwith to 1-2MHz and prevent self oscillation as the components used here will happily go to around 40Mhz and pick up any RF present in the area.

Hope I did not make any mistakes, I have only a MT degree (Master Tinkerer)

[lucpes]

Looks like the brickwall filtering appearance on the freq graph is due to the soundcard's ADC conversion, here's the both the soundcard/DAC freq test using RMAA:

[Bernhard]

Originally posted by lucpes


Yep, I used lower value gain resistors in the I/V stage than specified by the designer - what I had around - so the voltage out is a tad bit low (7dB lower than the sound card input level - I guess around 1V) but I won't change them now as both measurements and sound (on either CDP transport or soundcard) are spectacular.

The thd graph shows -10 dB input level and the dynamics graph -67 dB input level,

so is it a low level DIGITAL signal in the dynamics graph ?

Or what does the dynamics graph show ?

[lucpes]

Yep, in the dynamics graph there's a low level digital signal.

THD: 1kHz -3dB 'digital'

dynamics: 1kHz -60dB digital

IMD: -4dB 60Hz and -17dB 7000hz

The DAC's analog out at 0dB digital in is 7dB lower than the 0dB volume of the soundcard's line in, and all the graphs show the soundcard SPDIF->DAC->DAC's (analog) out->soundcard line in chain.

[Bernhard]

Ok, because Harmonics then start at -50dB which is not so very good.

Is it a "good" dithered signal like the one from Pedja's homepage ?

[lucpes]

I really don't think that for a -60dB signal a -110dB harmonic is that bad. We're getting into thermal or resistor noise issues in this range, not to mention that I used a $100 soundcard to test with its inherent deficiencies

Here's a PDF regarding some basic concept of measurements for an older version the program used (RMAA 5.3 - http://audio.rightmark.org/). The basic concepts stayed the same, but the frequencies/amplitudes are a bit changed in new version.

Don't know about what you want to say by 'good dithered' signal... I'm not that technical, and I'm not the one who designed the DAC, just a happy customer...

[Bernhard]

Originally posted by lucpes
I really don't think that for a -60dB signal a -110dB harmonic is that bad.

Don't know about what you want to say by 'good dithered' signal... I'm not that technical, and I'm not the one who designed the DAC, just a happy customer...

The -110dB is absolute scale, the signal is -60dB also absolte scale, so harmonics are -50dB.

Just shift the signal to 0 d<B and the harmonics by the same. They will be at -50 dB absolute scale.

[lucpes]

That's really not true, as you can see in the THD graph (-3dB FS digital sine and resulting noise/harmonics);

There are lots of co-acting factors that result different harmonic spectrae when using 0dB FS or -60dB FS signals on every audio equipment one may wish to test which I won't bother to cover in this particular thread.

Based on your reasoning one would test a piece of equipment with a -90dB FS sine and conclude that a device has only 15dB of dynamic range if the most proeminent harmonic is -105dB FS

[Bernhard]

We are talking about different things.
DAC distortion gets worse with smaller signals.