- Joined
- Feb 23, 2016
- Messages
- 20,706
- Likes
- 37,449
Ugh! So I think we need to close this to comments again now.
Now, that is a smart suggestion! Here are the results:Is it possible to send a properly dithered 20-bit signal at high and low level (e.g. -1dB and -90dB) and compare with the same 24-bit signal and show the FFT? Basically I would like to see how it performs without truncation artifacts.
Have you tried the 1 khz measurement to see if it is still off by that 0.00008 number using a 20 bit signal? What is the linearity of the -90 db test using 20 bit? What kind of probloem with the signal processing do you think is causing this 24 bit distortion?Now, that is a smart suggestion! Here are the results:
View attachment 13692
As suspected, there is signal processing error in conversion of 24 bit to a 20 bit PCM DAC used in Yggdrasil. The 20 bit samples definitely look cleaner and only have harmonic distortion and at lower amplitude. The 24-bit reproduction is clearly inferior with a lot more anharmonic content and higher distortion spikes.
Yes, the noise floor is higher for 20 bit but that is benign and comers with territory (of using lower resolution).
Now, that is a smart suggestion! Here are the results:
View attachment 13692
As suspected, there is signal processing error in conversion of 24 bit to a 20 bit PCM DAC used in Yggdrasil. The 20 bit samples definitely look cleaner and only have harmonic distortion and at lower amplitude. The 24-bit reproduction is clearly inferior with a lot more anharmonic content and higher distortion spikes.
Yes, the noise floor is higher for 20 bit but that is benign and comers with territory (of using lower resolution).
I haven't but I am pretty sure that is not changed.Have you tried the 1 khz measurement to see if it is still off by that 0.00008 number using a 20 bit signal?
I measured linearity but did not save it. It definitely changes from those up/down owings to a more gentle error.What is the linearity of the -90 db test using 20 bit?
They need to dither (add noise) prior to conversion to 20 bits. Simple truncation causes problems. It is shocking to see a digital DAC designer not know about this core principle of signal processing but that is what the facts point to.What kind of probloem with the signal processing do you think is causing this 24 bit distortion?
Reply ban issued to member @Thomas savageUgh! So I think we need to close this to comments again now.
Now, that is a smart suggestion! Here are the results:
View attachment 13692
As suspected, there is signal processing error in conversion of 24 bit to a 20 bit PCM DAC used in Yggdrasil. The 20 bit samples definitely look cleaner and only have harmonic distortion and at lower amplitude. The 24-bit reproduction is clearly inferior with a lot more anharmonic content and higher distortion spikes.
Yes, the noise floor is higher for 20 bit but that is benign and comers with territory (of using lower resolution).
I'd be pretty up front that for more than $2000 I would avoid the pain in the a** and buy a superior RME DAC for less than half price. A modern DAC of less than 24 bits that truncates is a total design failure. Completely unnecessary unless it is SCHIIT!If memory serves they were pretty upfront about the truncation. Sounds like a recommendation upgrade is in order for CD playback? The bad news for hi-res is that 20 bit real time dither is not always available in players. I guess one could run their hi-res library through SoX dithering to 20 bits, but it's a pita.
We need measurements with dithered signal to better understand redbook performance. Personally I don't care about hires.
I see no reason to test whether this expensive DAC might perform adequately in some limited use case. Since the existing measurements have already established that it is poorly engineered, no one should buy it. People with $2,399 to spend on a DAC should be giving their money to reputable companies, like Benchmark, that consistently produce top-quality products. That's the kind of company that we want to survive in this market space.We need measurements with dithered signal to better understand redbook performance. Personally I don't care about hires.
And no, the designers aren't ignorant of the fact. I tell you they are rubbing everyone's face into it and laughing all the way to the bank.
???I want to at least see linearity results using dithered signal, this would provide some insight into how the DACs cope with interpolated samples, assuming Schiit digital filter produces 20bits on its output from 16 bits.
It's a variation on the 'infinite monkeys' theorem. If you conduct an infinite number of tests on an infinite number of Schiit DACs, eventually one of the tests will show the DAC accurately converting the digital input to analog.Here is the comparison of 16 bits versus 24 bits:
{plot deleted for brevity}
The noise floor jumps up by 30 dB when using 16 bit samples. Is this what you were asking about?