DIYHiFi.org

For the sake of audio
It is currently Sun Dec 17, 2017 2:14 am

All times are UTC




Post new topic Reply to topic  [ 19 posts ]  Go to page 1, 2  Next
Author Message
PostPosted: Thu Jun 26, 2014 1:15 am 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
DSD upsampling. One of the topics I find interesting. Recently, Nativedsd.com has provided us with some upsampled DSD files for listening comparison. The upsampling from DSD64 was done with the Merging Pyramix workstation, and one example was done with the Signalyst software.

The process for DSD upsampling, to the best of my understanding is as follows.. 1 bit DSD, low pass-filtered to remove the high level of ultrasonic noise as to not overload the delta-sigma modulator, then re-modulation at new higher rate. Of course, the low pass filtering requires a conversion to multi-bit, something that DSD purists don't want to know. Trust me, I was/am one of them.

I have downloaded these files, and converted to DXD so I can take a look at the spectrum in Audacity. I would convert to something higher than DXD, but that is the limit of my software. The results are interesting.

The Merging Pyramix upconversion from native 64 to 128 is impressive. Very little difference in the two. Obviously a lot of care has gone into the filtering algorithms, to produce as identical a file as possible.

The same can't be said for Signalyst. That is, the result looks nothing like the original. The filtering looks to have been much more severe, and the noise curve looks nothing like the original.

Now, I have yet to listen to these files, but at least on paper Merging wins. You can see for yourself...



===========================


Original DSD64

Image


DSD64 converted to DSD128 with Merging Pyramix

Image


DSD64 converted to DSD128 with Signalyst

Image


Top
 Profile  
Reply with quote  
PostPosted: Thu Jun 26, 2014 1:33 am 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
Here is one more. Same original file.. This with Korg Audiogate.


DSD64 to DSD128, Korg Audiogate

Image


Top
 Profile  
Reply with quote  
PostPosted: Thu Jun 26, 2014 5:42 pm 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1167
Hi,

ball3901 wrote:
Here is one more. Same original file.. This with Korg Audiogate.


Try wav2dff too.

And Asio-Proxy. If you can figure out how to capture that datastream. I tried anything I could find, Jack, $hit whatever else...

What I know, on the 'sopce converting a single sample pulse at 44.1KHz to DSD512 using ASIO-Proxy looks ALMOST the same as the PCM... So DSD512 is at least as good as CD PCM for the first 20 or 30dB (no chance to see much lower on a 'scope - but you can look at a lot higher frequencies that way).

For fun, I tried MP3 (Softspace on Musictrade, kind of tranced out techno electronica) converted to DSD up to DSD512 (ASIO Proxy).

Make mine the original MP3, no oversampling, on the same DAC, same Headphones (Fostex T50RP). Even at DSD 512 it has a fair share of what I have sometimes called jokingly Matsushita's Delta-Sigma modulator system. The MP3 is more like the surname of that atonal composer I'd call Philip should I meet him...

I am sure many will disagree.

I can tell why some people like DSD. To my ears, a violin via DSD does not sound what a Violin sounds like to a Microphone suspended above the Conductor (that is where they usually are - I'd use a modified Decca tree myself -maybe 5 meters behind the conductor, but I usually do not get asked), it sounds like a Violin sounds like in the private boxes at the back... To me something that does that is "BAD" (including the concert hall).

I do consider the opposing view...

Attachment:
grumy cat.jpg


I usually sit centre, rows 3-6 at the Royal Festival Hall.

The Barbican usually has better Orchestras and conductors, but I hate the building (Stalin had better taste) and acoustics (think concrete box, yup...).

Plus the Southbank has my favourite watering holes for an after Concert Ale... And sometimes lesser known Ensembles and Conductors are surprising in a good way.

Ciao T


You do not have the required permissions to view the files attached to this post.

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Thu Jun 26, 2014 6:03 pm 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
After experimenting with DSD to DSD upsampling, and PCM to DSD upsampling, the sound is GREATLY influenced by the type of modulator.

I am not exactly sure of the differences between SDM A, B, C, and D in ASIO-proxy, (If it did conversion to a file, that would be awesome.. alas)

But my ear can certainly hear them. And at least one of those options to my ear creates an incredibly syrupy, 'euphonic' sound. As everyone's ear is different, I will leave it up to the listener to decide if they hear the same, experimenting on their own. But, remodulate a DSD stream with at least one of those choices, and what you get almost sounds synthetic! As in, live orchestral strings now sound like they came from my Korg Oasis synth!!! And cymbals sound so smooth, it is almost un-nerving.

Now, the thing is, on some material, I actually find that enjoyable!! It is unrealistic, but strangely likeable!!

Same can be said for DSD in general, I think. The example above is an extreme one, but the same effect I described is present in all DSD recordings, just a bit more moderated.

Also, I really didn't notice a lot of detail loss in DSD to DSD re-modulation. Of course there is information loss, but it wasn't as bad as I expected. The quality of the end result rests in the choice of modulator, and of course, the performance of the filter used.


Top
 Profile  
Reply with quote  
PostPosted: Thu Jun 26, 2014 6:56 pm 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
Kuei Yang Wang wrote:
I usually sit centre, rows 3-6 at the Royal Festival Hall.

The Barbican usually has better Orchestras and conductors, but I hate the building (Stalin had better taste) and acoustics (think concrete box, yup...).

Ciao T



You have WAY too much taste to be designing audio equipment. After all, ears aren't even required. :grin:

I DO kid, I Do kid. LOL.

Seriously, it is nice to know that some audio designers actually listen to good music, and know how it should sound, in addition to how it should 'look' on a scope...


Top
 Profile  
Reply with quote  
PostPosted: Fri Jun 27, 2014 11:08 am 
Offline
Dog
User avatar

Joined: Tue Dec 14, 2004 12:26 pm
Posts: 4981
Location: Lisbon, Portugal
ball3901 wrote:
Same can be said for DSD in general, I think. The example above is an extreme one, but the same effect I described is present in all DSD recordings, just a bit more moderated.


Be careful when you say "all DSD recordings" because very few are real DSD, 1 bit recordings. Most are derived from PCM.
What I find curious about DSD, is that the soundstage is noticeably wider, probably larger than life, but fun to listen to.
Thinking out loud, I guess it is the higher sample rate that does this effect, but I have no way to make extensive tests and put on paper what can't be put on paper - soundstage.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein


Top
 Profile  
Reply with quote  
PostPosted: Fri Jun 27, 2014 1:45 pm 
Offline
Hen

Joined: Sat Sep 22, 2012 5:05 pm
Posts: 1
Kuei Yang Wang wrote:
And Asio-Proxy. If you can figure out how to capture that datastream. I tried anything I could find, Jack, $hit whatever else...


ASIO-Proxy is based on Philips ProTech format converter, if you have access to that app you can get the same conversion though, unlike Maxim's ASIO-Proxy, limited to DSD64 output.


Top
 Profile  
Reply with quote  
PostPosted: Fri Jun 27, 2014 2:41 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4333
Location: Somewhere other than here.
Not hard to believe. Back when we were building CDPs, we found 0.1 - 0.2 dB, at 20 kHz, made a noticeable difference in soundstage width. So, if you have less aggressive low pass filtering, it is easy to understand it will sound wider.

Plot the response, at 20 kHz, for a second-order LPF. Especially if you play with alignment, and/or moving the - 3 dB point a small amount. You will see why DSD could easily sound that way. It doesn't have a 22 kHz brickwall, to start with. And probably different post-filters.


Top
 Profile  
Reply with quote  
PostPosted: Fri Jun 27, 2014 4:13 pm 
Offline
Dog
User avatar

Joined: Tue Dec 14, 2004 12:26 pm
Posts: 4981
Location: Lisbon, Portugal
Well, I also notice that PCM at 24/192 sounds "wider" and with more "space" around the instruments, even comparing to 24/96.
But DSD opens up in a way that the listening room seems to be larger.
Yeah, it could be the damn PCM brickwall filters, Jocko.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein


Top
 Profile  
Reply with quote  
PostPosted: Fri Jun 27, 2014 7:34 pm 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
carlosfm wrote:
Well, I also notice that PCM at 24/192 sounds "wider" and with more "space" around the instruments, even comparing to 24/96.
But DSD opens up in a way that the listening room seems to be larger.
Yeah, it could be the damn PCM brickwall filters, Jocko.



Yes, I think you are onto something. The filtering is key. The higher the sample rate, the less audibly invasive the filter is. DSD doesn't need brickwall filtering, and I believe that is a major part of its sound. It is why you get DSD purists denouncing any decimation to a lower sample rate.

The filtering, combined with the sound of the delta sigma modulator give DSD its distinct sound. And as you alluded to, if the source is actually PCM, the filter advantage is essentially lost.

One thing I am interested in trying, when the software becomes available, is downsampling DSD to 768khz PCM and playing back on Thorsten's soon to be released DAC. The decimation filter for such a high rate of PCM can be very gentle, also it would mean the megahertz level noise is no longer an issue on the DAC level.


Top
 Profile  
Reply with quote  
PostPosted: Sat Jun 28, 2014 1:58 am 
Offline
Pig

Joined: Wed Jun 29, 2005 6:34 pm
Posts: 1068
Location: East Coast USA
Or... The large amount of pseudo white noise above 20 KHz beats down into the generally recognized audible band (aka IMD), where it modulates all sorts of system components based on the content. Power supply crap gets into this as well. (Ever check out what kind of PSRR and CMRR you get from a typical audio oriented opamp at 100+ KHz?)

Or... The same noise like distortion products appears to the hearing system to be similar to certain types of 1/f modulation sidebands, which are generally recognized to affect the perceived soundstage.

Or... These noise products effectively mask some of the bad effects of 1/f noise sidebands.

Or... The high frequency noise adds Doppler-like distortion to the tweeter frequencies.

Or... Some of the above.

Or... All of the above.

Or... None of the above.

The human aural system is not as straightforward as the DBT guys like to portray. At least, the acousticians don't seem to think so.

Most analog filters really kind of stink when it comes to delay characteristics and distortion (especially in the case of active filters). I imagine that changing the characteristics just a little could make a big change, as Jocko 'splains. Those Sallen-Key and similar active filters are nightmares when implemented using opamps unless you really take serious measures to minimize so-called common mode distortion.

The list goes on.

And - just how wide do you guys expect the soundstage to be? My own observations - ravings of a lunatic - is that the room acoustics and speakers placement determine about half of that, while power supply influenced noise and "acoustically stimulated" audio component characteristics pretty much the rest. The amount of masking that normally takes place is really amazing.


Top
 Profile  
Reply with quote  
PostPosted: Sat Jun 28, 2014 8:38 am 
Offline
Cow
User avatar

Joined: Tue Aug 26, 2008 7:14 pm
Posts: 206
Yes, injecting an inverted phase signal from other channel will cause wider soundstage. Maybe similar coupling is happening here, HF noise is modulating opposing channel?


Top
 Profile  
Reply with quote  
PostPosted: Sat Jun 28, 2014 11:50 am 
Offline
Dog
User avatar

Joined: Tue Dec 14, 2004 12:26 pm
Posts: 4981
Location: Lisbon, Portugal
I seem to have started a discussion... but I think a positive one.
My experiences are based on careful listening tests on my own system and yes, I know that 1cm aside can make a noticeable change on speaker placement - but I don't move the speakers, it has taken me hours to place them, many years ago.
The wider soundstage and more "air" and sense of space I also notice with PCM, the higher the sample rate. As I've said before, 24/192 is audibly "wider", more "airy" and even sounds faster than 24/96, which is already very good.

And I do know some stuff about audio production. You can turn a mono signal into sounding like "stereo" by delaying that signal by a few ms on one channel.
But it's not these artifacts that I'm talking about.
I'm talking about the sound getting "out" of the speakers, not so much left, center (if well positioned) and right, but wide, deep, tall.
And IME with PCM, lower jitter also sounds like tighter bass, more precise treble, better location of the instruments and... wider soundstage.
I can't consider wider soundstage a bad thing, or some kind of artifact.

PS: all this may pass unnoticed for those that only listen to music with headphones. :music:

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein


Top
 Profile  
Reply with quote  
PostPosted: Sun Jun 29, 2014 4:52 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1167
Hi,

CG wrote:
check out what kind of PSRR and CMRR you get from a typical audio oriented opamp at 100+ KHz?)


Try 80-100MHz (for ESS DAC running at that clock speed), if you like a real nightmare...

CG wrote:
Or...


CG wrote:
The human aural system is not as straightforward as the DBT guys like to portray. At least, the acousticians don't seem to think so.


Not by a long stretch. In fact the almost complete lack of referencing audio measurements/test (including ABX) to what is known about human hearing and the lack of validation of these tests makes modern audio (I am talking about the objectivist camp) a bunch of Cargo Cult (Science) Practitioners.

JA at stereophile is a prime example. A degreed EE, practiced for along time, also long time editor of "subjectivist" magazines. And after all these years he gets incredibly frustrated when the listening impressions of his reviewers and his measurements refuse to line up, again and again and again.

By now we (as a "sub culture" - not specific individuals) should have learned this. But it is much easier to pretend it matters and follow empty rituals and mutter empty incantations and somehow hope the magic will happen. Good thing I got away from that cargo cult early enough...

CG wrote:
Most analog filters really kind of stink when it comes to delay characteristics and distortion (especially in the case of active filters).


Yes, but does it matter? A speaker is an analogue filter. As is a Headphone. Any offers as delay characteristics and distortion of the best?

The human ear itself is a complex system with build in bandpass (what are the delay characteristics of that I wonder) and very high frequencies are sensed through bone conduction and very low ones through the entire body's skin. Plus it uses digital trasducers tuned to a specific SPL and Frequency and uses what amounts to analogue (D) signal processing to reconstruct the signal from interference patterns.

CG wrote:
The list goes on.


So does mine...

CG wrote:
And - just how wide do you guys expect the soundstage to be?


Here is the crux.

I expect realism.

That means the subjective soundstage of an orchestral work is as wide as that with a real orchestra in a hall, at least mid-hall or further up. So, wider than my living room, way deeper than the space behind the speakers, actually way past the walls of my office (it is behind the speakers) into open air, 19 floors up... ;-)

Back in the 80's I was working on minimalist microphone techniques that would allow this as a simple 2-channel capture (with some success, but I lacked a key ingredient and no I lack time and incentive to do much). Decca recordings often capture this well (given the right replay setup), at the expense of image specificity.

CG wrote:
My own observations - ravings of a lunatic - is that the room acoustics and speakers placement determine about half of that, while power supply influenced noise and "acoustically stimulated" audio component characteristics pretty much the rest. The amount of masking that normally takes place is really amazing.


I tend to go with that.

Get speaker / room setup right (and the speakers themselves, most are designed in my view "contra legem/principa naturae").

Control noise well (RFI, EMI, PSU, etc. - see why I have doubt about DSD?).

Avoid obvious ana-harmonic (or dissonant) distortion even at fairly low levels, but devil may care about 1% H2 at 90dB SPL and 3% H2at 100dB.

Correct for the spatial representation problems inherent to most panpot stereo recordings and/or "point mike" recordings (including ORTF, Jeklin et al) and you are 80% there.

Add some tubes for the voodoo they do, you are at 90%. My office system is like that, with a Custom build speakers, Gainclones with CRC filtered PSU and inexpensive electronics (my iTube and iDSD nano plus an old Notebook).

To get the last 10% gets challenging. I put my living room system at maybe 95% (> 10K US worth of AMR electronics moded with "voodoo capacitors" and all that jazz plus speakers that, were they not custom build, would cost serious money, probably > 10K US the way the market goes these days).

That said, the big system may be only 5% better, but it gets > 80% of serious listening,the office system only sees use as background noise machine...

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Sun Jun 29, 2014 5:19 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1167
Carlos,

carlosfm wrote:
The wider soundstage and more "air" and sense of space I also notice with PCM, the higher the sample rate. As I've said before, 24/192 is audibly "wider", more "airy" and even sounds faster than 24/96, which is already very good.


Have you ever tried digital filterless DAC's? I find on my DAC very little if any difference in sound-staging of the same recording at 44.1 and 192K, provided they are all played filterless or with the very minimal "organic" filter.

I would posit that the sound-stage effects from higher sample-rates have little to do with either analogue filtering nor the sample rate itself. Instead I find almost all DAC's and ADC's relax digital filtering substantially as sample rate goes up.

And digital FIR filters are like this:

Image

Actually, analogue FIR filters too are like this, but they can be made to have an impulse response sufficiently short that a normal analogue IIR filter at a lower frequency willswamp their effect.

To imagine what a FIR filter does think of it like this:

Image

(Incidentally, these images are from Kusonoki's article in MJ,reproduced at Sakurasystems: http://www.sakurasystems.com/articles/Kusunoki.html)

Is it any wonder that the soundstage improves if the filter length is reduced?

Is it any wonder that for DSD which in principle lacks any substantial filter of such kind (except when using some DAC Chip's of course, which can still have shorter filters for DSD) has improved soundstage?

Now if the filter length is reduced at the same sample rate while accepting the presence of larger levels of nyquist images or by raising the sampe-rate while keeping image levels the same, either way we reduced filter length.

Non-Oversampling with SINC compensation in the lowpass is the logical conclusion of making the filter response as short as possible.

I was very much vexed and bothered when it turned out I was unable to give the iDSD nano this feature because of multiple hardware limitations of the stuff we used. :doh:

Thankfully the iDSD micro overcomes these limitations and Iam going to snag one of the first of the line for my own use... :P

carlosfm wrote:
PS: all this may pass unnoticed for those that only listen to music with headphones. :music:


It should not. All you need to do is to correct for the spatial distortion introduced by playing recordings made for speakers on headphones. I worked on that in the 80's, with the East German radio research institute, I have revived some of this stuff recently, with good success it seems.

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 19 posts ]  Go to page 1, 2  Next

All times are UTC


Who is online

Users browsing this forum: Yahoo [Bot] and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group