DSD: Any Takers?

[johnb]

johnb

re msg 36, the second example sounds pretty bad, but it doesn't sound bad in the way I would expect. I wonder if there's an interaction with the mp3 encoding. If the same process is done to produce a wav file do you get similar audible results?

Dave,

The original was a 16/44.1 wav file and all the editing and processing was done with wav files. Afterwards I encoded the wav files as mp3 256 kbps for convenience.

What you hear with example 2 is the quantization error generated by the reduction from 16 bit (65536 possible levels) to 8 bit (256 possible levels) without any dither being applied. As a check, I have also done the same bit depth reduction (without dither) using SOX which confirms the results from Sound Forge.

The quantization noise is more extreme than you might have expected because the original (as ripped) was recorded at a fairly low level. It peaked at -16dB and had an RMS level of approx -37dB.

If I take the original, normalise it (say -1dB peak), do the fade out and then reduce the bit depth from 16 to 8 without dither (as before) - the quantization noise is much, much lower (compared to the music) at the beginning of the 60 seconds clip but reaches similar levels towards the end.

To allay your suspicions about the effects of mp3 - this is the original wav file for example 2, from which the mp3 was encoded: https://app.box.com/s/sw53fxpt0neb2yawia41

There is a touch of magic in how noise shaped dither can transform the quantization noise and extend the perceived dynamic range!

[Dave2002]

johnb

OK - you got me there. I think there was a slight difference between the wav and the mp3 - but both sounded pretty bad, and the signal dependent noise is present in both. The gaps for the silent sections make sense, too. It is surprising how much better things sound with both forms of dither.

Who needs hi-res at all? In this case 8 bits seems OK.

[Dave2002]

Just a quickie - if the recording is stereo, should the dither be identical on both channels for best effect? I'm guessing that down at low bit levels it won't make much difference, but just a thought .... Might give a slightly odder effect in 8 bit recordings.

If the dither is the same it will correspond to a central noise source - if anti-phase - who knows, and if random it will be spread around. Won't be possible to make sense of this unless only a low resolution signal is replayed.

[richardfinegold]

Posting on this thread has cured me of the desire to waste cash on a DSD DAC, but I notice in the current issue of Hi Fi News that Denon UK is selling a DSD DAC
for 329 pounds. It has the "same advanced AL32 processing and 32-bit/192 khz converter used in Denon's flagship CD/SACD players, has asynchronous USB and a headphone jack. It does DSD at 2.8 Mhz and 5.6MHz and supports all usual file formats.
I have Denon SACD/CD player that does 24 bit/192 and it's a great player.
It looks like the price of the DSD DACs have become realistic; now if only the price ofthe downloads would follow suit.

[Dave2002]

I found this about playing DSD files on a computer using foobar - http://www.audiostream.com/content/h...ing-foobar2000

Also this link to an SACD decoder - http://sourceforge.net/projects/sacddecoder/

How useful that information is I'll leave up to the reader.