Converting DSD to PCM

[Miska]

3 hours ago, audiventory said:

I never heard about "PCM-style" of sound as measurable instance. And I never stumbled about ringing research results.

 

It is not hard, I recorded for example glockenspiel at 192k and 44.1k sampling rates (and also did typical software conversion of 192k to 44.1k) and the ringing of initial attack is clearly visible. Which is of course clear because of band limited nature of the wide-band signal.

 

3 hours ago, audiventory said:

High resolution PCM allow to use wider range of filter transient band steepness. But slope transient band can cause audible noise by utrasound products via intermodulations.

 

Yes, you need to make up your mind. That's why I also offer multiple choices. With DSD the nice thing is that in case of a good modulator, any intermodulation products are random and not correlated with the baseband signal. So at most you get reduced SNR. Unlike the case of PCM with leaky filters where images/aliasing is directly correlated.

 

[Miska]

2 hours ago, Ron Scubadiver said:

I don't know if it is right, but I see a lot of posts saying 88.2x24 has the same amount of information as  DSD64.  Then again, those post were not written by someone who develops resampling and conversion software.  On the filters being at too low a frequency, anything below 30k, and possibly higher, I agree.  The part I don't understand is why signal way above the audible range is important.  It certainly is important to the MQA crowd since their secret sauce is lossy compression of frequencies above 24k.

 

If I have to stay at 176 with a 50k (SACD specifcation IIRC) filter there will not be enough of a reduction in file size to make the task worthwhile.

 

From the looks of the output spectrogram the JRiver 24k filter is steeper than the specified 48db/octave slope.  Perhaps 60 or 72.

 

24-bit is more of the limitation than 88.2 kHz sampling rate. Although for content from analog sources 24-bit is plenty enough. From time domain perspective I made some calculations in the past and the DXD (352.8k) is pretty good because it matches capabilities of DSD64.

 

My choice of 176.4k for DSD64 was sort of "most practically sensible", because most PCM capable DACs support that and it sits right between the two rates.

 

Just for sake of it, I tested long ago how Schiit Loki DAC (the old product, not the new namesake) would produce 250 kHz full level sine at DSD64 (only thing it supports):

 

And yeah, there was still detectable output, although attenuated quite a lot due to the analog filters.

 

[Ron Scubadiver]

This is turning out to be more involved than I first thought.  Looks like I will keep doing on the fly conversions for the time being sending 352.8k output directly to my DAC.

[Ron Scubadiver]

I am starting to think I don't have to be concerned about low pass filtering if sampling down to 88.2k as Nyquist takes out everything over 44.1.  There are some lower SQ files that could be taken down to 48k.

[audiventory]

At some instruments with steep attack we can see ringing. It is not good, of course. But we don't know real impact the ringing to listener perception, because there are need researches.

 

On 26.01.2018 at 12:12 AM, Miska said:

intermodulation products are random and not correlated with the baseband signal

 

If significant high frequency noise intermodulated (non-filtered DSD converted to PCM, a.k.a. DXD) we have audible noise due intermodulations (if they significant, of course). Also slope playback filtering can cause aliases after downsampling.

 

 

[Miska]

6 hours ago, audiventory said:

If significant high frequency noise intermodulated (non-filtered DSD converted to PCM, a.k.a. DXD) we have audible noise due intermodulations (if they significant, of course). Also slope playback filtering can cause aliases after downsampling.

 

Yes, but because those are uncorrelated, that still sounds like white noise, so essentially just SNR reduction. DXD is not "non-filtered" though. But as I said I one can use a filter that matches slope of the DSD noise. No need to be any steeper, because that'll keep the noise floor pretty much flat throughout from 0 Hz to Nyquist frequency.

 

Unlike with leaky PCM filters where the intermodulation products are correlated and sound like distorted music. Leaky PCM filters typically also produce much higher levels than what the leftover DSD noise is.

 

Whether that intermodulation is happening, is easy to test.

[audiventory]

28 minutes ago, Miska said:

DXD is not "non-filtered" though.

 

Agree. Wideband filtered.

 

29 minutes ago, Miska said:

Unlike with leaky PCM filters where the intermodulation products are correlated and sound like distorted music

 

It is interesting conclusion. I don't think about it from that side before.

[Ron Scubadiver]

Various minor discoveries:

S-Audio has 2 FIR low pass filters on it's site.  The Short MP filter uses fewer CPU resources, but does not do a very good job of removing high frequency noise.  The other one is very close to the 30khz FIR filter supplied with foo_input_sacd in terms of CPU usage and the size of a downsampled to 88.2 PCM file.  It's a do your own listening and make your own choice kind of thing.  If there is a difference in the results it's subtle.

 

I converted a few DSD's to 48/24 using the 64fp multistage in foobar.  It has no lowpass filter.  These were chosen due to their hot and somewhat abrasive sound.  They sure didn't sound worse, and might have sounded a little better.  48/24 has roughly twice the bitrate as Redbook.

 

foo_input_sacd has been updated for compatibility with Foobar2000 1.4 beta.  It will now play ISO's.

 

 

[mansr]

5 minutes ago, Ron Scubadiver said:

I converted a few DSD's to 48/24 using the 64fp multistage in foobar.  It has no lowpass filter.

Conversion to 48 kHz by definition has a lowpass filter at 24 kHz. There is no need for a DSD-specific filter.

[Ron Scubadiver]

3 hours ago, mansr said:

Conversion to 48 kHz by definition has a lowpass filter at 24 kHz. There is no need for a DSD-specific filter.

Yeah, I understand that, Nyquist.  On these particular recordings the result was good.  I don't think it is a solution for everything, unless file size is of paramount importance and one does not want to go all the way to Redbook.

 

[jaynyc]

any updating thinking/learning here since January?   What's the best most transparent way to convert this

https://store.acousticsounds.com/d/120764/The_Doors-The_Soft_Parade-DSD_Single_Rate_28MHz64fs_Download

  to PCM for playback on PCM only DACs

 

thanks