PCM to DSD conversion

[JamesD]

Ahh - I should properly read the waffle doc - then I won't miss the source of your points

Doesn't DSD do its digital filtering upstream rather than downstream? Must read up on it.

BTW this is, in part, why NICAM sounds so good...

I getting really annoyed by the increasing level of waffle and miss direction in what purports to be technical documents - I'm finding it more and more in my work as well as in our hobby and my tolerance level for it is getting lower and lower... Ah well...

Back to records/single ended valve amplifiers and full range speakers - at least I know why it measures like it does and why it sounds nice

[Ray P]

If anyone is interested in trying DSD recordings (and has a suitable DAC) NativeDSD is offering a short, but free, sampler download that's available at various data rates;

https://www.nativedsd.com/catalogue/alb ... rter-pack/

[Ray P]

If thats in response to

Nick, I've had a quick look but can't find it at the moment but

I didn't see any of that info about DSD encoding, just a lot of articles. Ifs there one I should look for?

I had a look for the original article on encoding that I recall but haven't seen it so far Nick, sorry; maybe that it's been removed but it could also be entirely possible that I've just lost a few too many grey cells.

Moving on to the more general topic, as the main, and perhaps only, consistent user of DSD playback on this forum I just want to emphasise that I'm not really that bothered by arguments about whether, technically, PCM is better/worse than DSD - I listen to DSD simply because I have consistently found that I prefer it though I can't explain that preference, which might be due to all sorts of influences.

In fact it goes beyond not being able to explain the preference as I don't even understand some of what I hear; for example most of my listening is done at DSD256, derived from upsampled 44.1KHz/16bit CD rips using HQ Player. I know that there cannot be any more information than is contained in the original file and yet with the DSD256 playback I hear 'more' - by more I mean things like being able to clearly hear each instrument/performer in their own space and the sounds of the performers/instruments are more natural (by that I mean more like what I know things sound like from direct acoustic experience) but the thing I really don't get is the level of additional 'detail' that I perceive. As well as the 'more' I also find the sound more natural and 'organic' - it has a 'smoothness' that makes it very easy to listen to. Don't get me wrong, I'm not anti-PCM and it can sound exteremely good too - my Soekris DAM1121 R2R DAC sounds fantastically good.

My DSD preference has been so clear to me that I now have a couple of 'DACs' that can only decode DSD and I recognise that could be some sort of self-reinforcement so I will continue to use and explore both technologies (I'm currently working an an AD1862 based R2R DAC project for example). If people want to dismiss DSD as a cul-de-sac or ask 'what's the point' that's fine, or maybe even their loss, but the only justification I need for using it is the pleasure of enjoying listening to music.

I hope that doesn't come across as me being defensive, I prefer to think that I'm just being open-minded and gaining pleasure exploring possibilities around subjective enjoyment regardless of the views that others may have about different technologies.

[ed]

Agree entirely with the philosophy Ray. At this point I can't agree on the 'which I find better' front because so far I haven't heard the difference. This may be spurious because there are too many variables at play, eg your HiFi may be better at displaying the difference, or your ears may be better at detecting the differences, or your environment may be better in-tune with your equipment etc etc.

I didn't think I had anything to contribute on the techy front, but, in acknowledging that many recording engineers are arguing about this very same thing at present then anything I do say will be worthless. Suffice to say that at present the word is all in favour of dxd(hi res pcm) throughout the recording/mixing chain and then mastering down to dsd.

the jury is out, the exploration continues, tomorrow may be different.

One argument that stood out for me was the fact that our brain can process sound(for change) at 7us, 48k sample is 22us and 192k is 10us. Surely anything above 384k(DXD) is only of interest to aliens.

[Nick]

Yep, not seen as defensive here, also I am not as I say against ot for anything.

I am well behind the curve with what I have heard, and ed's experiments are interesting in that it may be the encoding and data rate used is making them transparent. Or it may (just to try and consider as much as possible) that his DAC's remove any difference. My experience is from having a few SACD players in to repair and the DAC I made that would play DSD over PCM. In all those cases it was the "smoothness" I found the problem. It seemed to give everything the same sound. That's why I was interested in finding the encoding spec as I think I may have been hearing the filters in that spec. We know that the filter setting in a DAC will change what we hear, so viewing DSD as a "long" DAC it would seem likely the filters would also matter. Of course unlike the local filters in a DAC, DSD has them built in to the delivered data so can not provide any option for tuning to a particular preference.

Of course the data rate on the recordings I am referring to was much lower than what you have so that's likely to also be significant.

The cynic in me wonders if the extra separation and detail you find may be a artifact of the data reduction. Are there any open source encoders out there? That may be a source of the information.

This is probably a good place to start

https://audiophilestyle.com/forums/topi ... /#comments

[Nick]

48k sample is 22us and 192k is 10us

You may mean 96k is 10us or 192k is 5us.

[Nick]

And it gets even more confusing, this now describes the higher rates as PDM

http://help.nativedsd.com/en/articles/3 ... ed-and-why

[Nick]

Ok, so its all PDM...

[ed]

48k sample is 22us and 192k is 10us

You may mean 96k is 10us or 192k is 5us.

Doh! any fule no that 192 is twice 48!
nearly proof by induction

but seriously

If I master my tracks directly to dsd then the resolution will only be as good as my original tracks were recorded, which has always been 44.1 or 48. So I will get the same resulting DSD file as using the AUI conversion. It may be that some of Ray's recordings started life at a much higher sample rate in which case the end result may be much more telling.......I don't think I'll pursue this any further.....I would have to do some more recordings at 192 to make anything like a reasonable comparison...maybe later.

edit: if this is the case:

In fact it goes beyond not being able to explain the preference as I don't even understand some of what I hear; for example most of my listening is done at DSD256, derived from upsampled 44.1KHz/16bit CD rips using HQ Player.

then my previous statement about comparisons is valid

[ed]

I’ve been thinking a bit more about this dsd thing. I’ve been messing with the studio(first time in ages) in the hope of laying down a couple of new songs, tracked at 192khz, to enable mastering directly to dsd for comparison.

It occurred to me that I usually record the piano via midi, using one of the synth voices I have. It then further occurred that whenever I add trumpet or violin or sax parts I always use midi via the piano keyboard. Basically I can neither play a violin or a trumpet or a saxophone. One thing leads to another and then I thought...what about all those modern recordings we hear that have instruments beyond the usual 4 part band?...how often are synth voices put on modern high visibility releases, where the band or studio don’t want to fork out for a musician???
So the question comes round to what resolution do the synths pass the tracks to the pcm recording and hence to the dsd.

At the moment I’m thinking that all is not as it seems and I might do some more exploring.

[Nick]

So the question comes round to what resolution do the synths pass the tracks to the pcm recording and hence to the dsd.

Not sure its helpful, but maybe is that one of the things we are hearing when we prefer analogue synths to more modern digital ones?

[ed]

it may be possible to open a can of worms here......

To the best of my knowledge(it's limited) analog synths are restricted to generating the sound with oscillators and as far as mimicking orchestral instruments is concerned, a bit of a nono. The purists, I beleive, are all in the modernist camp. There is an old argument that goes along the lines of 'no 2 identical instruments sound the same. I do remember Knebworth in about 1973 when Pink Floyd kept us waiting for ages because they couldn't tune the synth due to voltage variations....In truth I think they were waiting for the spitfire to flyover and commence the gig.

Digi synths have long been into the realm of using samples, i.e they reproduce real instrument sounds, which is why the question is 'is it realistic'. Certainly some of the softsynths I have are outstanding for realism. As I say, my knowledge is limited, but I do beleive there's a third method called analog modelling which may address the variations in electronic analog sounds to some extent but I'm pretty sure it's not accurate orchestraly.

I don't think it's possible to compare analog and digi in the context of preference, not objectively anyway. My take is that an analog synth is an instrument in it's own right, whereas a digi synth mimics other instruments. YMMV.

The last time I played an analog synth digital ones didn't exist, so what do I know. Well, I do know that when I bought my Yamaha SY77 back in the day I was blown away by the orchestral voices. I really haven't thought about it since.

[Ray P]

Not relevant to Ed's conundrum but just for general information, having made some minor tweaks to the clock 'matching' in my Valve DAC I'm now listening consistently at DSD512 rate - maybe sounds just a tiny bit more 'refined' compared with DSD256 but I could just be deluding myself.

I've been listening at my work desk via headphones and really need to put the Valve DAC into the main system again for a listen - will be done just as soon as I complete the finishing touches to the 13E1 SE-OTL amp.

[rowuk]

Any upsampling is done in the digital realm and is arguably lossless. If however, a digital to analog conversion is less intrusive at higher sampling rates, then playback at those rates could sound better. I can well imagine that a DSD playback “sounds” different than a PCM at 1/5th the sampling rate.


Anyone wanting to find out more about the state of the art sampling of live instrument sounds, should check this out:
https://www.vsl.co.at/en
This is as close to real as it gets. They even have devices for crescendo/decrescendo, articulation and phrasing.

[Nick]

This is as close to real as it gets

Well, there is real. That could well be closer.