DSD or FLAC

In a nutshell, my belief is that the DSD file is likely to be 'high resolution', and so should be better SQ than a standard FLAC or WAV file. I assume that like most Naim users, you transcode FLAC to WAV on the fly when playing your FLAC files. Now that NAIM streamers support DSD, the DSD file should (if produced properly) give you a better sound quality than an equivalent FLAC. You should certainly not transcode the DSD to WAV.

Most of the few DSD files I have sound vey good indeed through my NAIM/Hugo streamer/DAC. 

     Hi Bert,  you might try a search of the forum as there are a lot of threads on the subject.  Some posts explain the technicalities if you are interested,  & many are concerned with which sounds best - worth the time to search.

I've been using DSD since it was first intro'd into the NDX beta firmware & quickly came to an opinion on it.  I believe that generally with DSD, the people who record with it tend to be very serious about producing a high quality recording & this is probably the main reason I hear something better/different with DSD.   It seems well suited when DSD (DSX) is used throughout the whole recording process with small acoustic groups.  Whereas I understand larger groups such as full orchestra are easier/better recorded in PCM & changed to DSD at some point during mixing/mastering.  

In my system DSD sounds very different to PCM (I am all WAV) Its very clear to the point its not really debatable. At first it seems the volume is too low, but the dynamic transients are so much more …. startling & loud is probably the most descriptive. These dynamic peaks have no hint of distortion or compression & all very true to life. The sound in general has more of a feeling of analogue sound than anything I've heard from other digital, a higher level of realism with both instruments & voice. This definition is present through all the frequencies, bass wood & string timbre is clear & you can hear the cymbal brass ring & the wooden stick hitting it. Sound stage is noticeably better with a very tangible three dimensional presentation, fixed positions & space between & a great sense of distance. 

I would be interested in your opinion once you've tried it. 

A well written summary from Mike.

A dynamic range of a DSD file can be significantly better than PCM. Also the upper frequency range can go up to 50 KHz. Whilst of course we cannot hear that 'high' we can nevertheles 'feel' the upper harmonics.

Hence the perception of improved quality.

Mike-B posted:

     Hi Bert,  you might try a search of the forum as there are a lot of threads on the subject.  Some posts explain the technicalities if you are interested,  & many are concerned with which sounds best - worth the time to search.

I've been using DSD since it was first intro'd into the NDX beta firmware & quickly came to an opinion on it.  I believe that generally with DSD, the people who record with it tend to be very serious about producing a high quality recording & this is probably the main reason I hear something better/different with DSD.   It seems well suited when DSD (DSX) is used throughout the whole recording process with small acoustic groups.  Whereas I understand larger groups such as full orchestra are easier/better recorded in PCM & changed to DSD at some point during mixing/mastering.  

In my system DSD sounds very different to PCM (I am all WAV) Its very clear to the point its not really debatable. At first it seems the volume is too low, but the dynamic transients are so much more …. startling & loud is probably the most descriptive. These dynamic peaks have no hint of distortion or compression & all very true to life. The sound in general has more of a feeling of analogue sound than anything I've heard from other digital, a higher level of realism with both instruments & voice. This definition is present through all the frequencies, bass wood & string timbre is clear & you can hear the cymbal brass ring & the wooden stick hitting it. Sound stage is noticeably better with a very tangible three dimensional presentation, fixed positions & space between & a great sense of distance. 

I would be interested in your opinion once you've tried it. 

Thanks for your impression - I will have a good listen over the weekend to the one I have got.

Adam Zielinski posted:

A dynamic range of a DSD file can be significantly better than PCM. Also the upper frequency range can go up to 50 KHz. Whilst of course we cannot hear that 'high' we can nevertheles 'feel' the upper harmonics.

........   errr    not so Adam,  as I understand it DSD can achieve 120dB  -  16-bit PCM is 96dB & 24-bit 144dB.   But this is theory (math) & in practice they all fall short of those numbers.   However (again as I understand it)  DSD "appears" to have higher dynamics because it operates in a linear form, whereas PCM has noise shaping. 

Yes DSD does (or maybe better to say "can")  go to 50kHz,  but so does 24-bit PCM that has been mastered as a HD download.  Problem is most all 16-bit files originate from CD masters & will have the 21kHz "brick wall" filter limit applied    

Bert - as Mike tends to suggest - PCM (FLAC/ALAC/WAV) and DSD are quite different formats - they encode the analogue data in quite different ways.

The way analogue audio is encoded and reconstructed is a compromise - and essentially PCM and DSD are about which different compromises to focus on. So talk of bandwidth and dynamic range really fails to capture the key audible differences (as much of this becomes marketing mumbo jumbo with respect to most quality audio replay equipment) and there are other aspects like aliasing and noise - frequency distribution that can have a marked effect ... IMO.. and recent research (I referred to some of the published research on this on my recent Hidef thread) has shown that frequency response with hearing is not limited to pitch response but to timing response of our auditory nervous system - and to capture the higher resolution timing one needs to a wider sampling bandwidth - but again Hidef PCM and DSD and its higher multiples can provide this - albeit most speaker / room couplings would blur and distort this additional info.

Therefore the bottom line treat DSD and FLAC as different source types - and the better your replay equipment the more apparent the virtues and benefits of each will become apparent. Some newer DAC c hips accept both PCM and DSD streams without requiring the DSP to convert from one top the other which otherwise can compromise the performance a little.

I tend to use DSD for orchestral and choral work as it brings a quality that I typically find elusive on PCM for those music types. However well recorded DSD is almost as rare as hen's teeth and so for me it is not a mainstream format - but a little treat to dip into now and again

Simon

The only DSD album I have so far scored is also available in HiRes in other formats so one should not automatically assume that  more love and care has gone into it because it is available in DSD, unless it is from a specialist label. I have the album in WAV and to my ears the DSD sounds less forced and more natural. It might be beginner's luck for me but I will be looking to increase my DSD catalogue when stuff finally comes out that I'm willing to pay for. As I'm sure it will. And if sourced from the SACDs already out there, some of them will sound awful. 

Mike-B posted:

Adam Zielinski posted:

A dynamic range of a DSD file can be significantly better than PCM. Also the upper frequency range can go up to 50 KHz. Whilst of course we cannot hear that 'high' we can nevertheles 'feel' the upper harmonics.

........   errr    not so Adam,  as I understand it DSD can achieve 120dB  -  16-bit PCM is 96dB & 24-bit 144dB.   But this is theory (math) & in practice they all fall short of those numbers.   However (again as I understand it)  DSD "appears" to have higher dynamics because it operates in a linear form, whereas PCM has noise shaping. 

Yes DSD does (or maybe better to say "can")  go to 50kHz,  but so does 24-bit PCM that has been mastered as a HD download.  Problem is most all 16-bit files originate from CD masters & will have the 21kHz "brick wall" filter limit applied    

That's why I used the word 'CAN' Mike

As to the upper frequency range limitation - most of the streamers that I know and CD players will indeed have a cut off around 20kHz. If my memory does not fail me, Super Audio CD players should go up to 50kHz (to handle the theoretical upper limit of the DSD files which are carried on the SACD).

NAIM streamers seem to be limited to 20kHz, with an exception of.... SuperUniti - the brochure states it's range as 50kHz. Curious that one...

Bert Schurink posted:

[...] If you look at the pure size [...]

Hi Bert, a FLAC file will always be smaller because it is compressed (think .zip here). This is very much like WAV/AIFF (uncompressed) on one hand and FLAC/ALAC (compressed), all of them PCM files, on the other.

If you take a look at the downloads at Hifistatement, there is a piece by  Michel Godard in 44/16 PCM, 24/192 PCM and 2.8 MHz DSD. The respective file sizes are 70.6 MB, 460.8 MB and 283.5 MB.

Mike-B posted:

[...]   However (again as I understand it)  DSD "appears" to have higher dynamics because it operates in a linear form, whereas PCM has noise shaping. 

Hi Mike, I think it is just the other way around. DSD has a quite aggressive noise shaping towards the upper frequencies. The dynamic range is very high at its lower end (which can extend into  the +20 kHz range), but virtually noexistant above that. PCM has a linear range that can/will be lower in some parts of the frequency spectrum.

Adam Zielinski posted:

As to the upper frequency range limitation - most of the streamers that I know and CD players will indeed have a cut off around 20kHz. If my memory does not fail me, Super Audio CD players should go up to 50kHz (to handle the theoretical upper limit of the DSD files which are carried on the SACD).

 

Adam you are talking pitch here - that is only one part - and arguably small part of the sound scape. There is also timing resolution as well - and that will be determined by the analogue recording resolution, sampling bandwidth of the original conversion, the media, and the DAC converter at the end... in the analogue domain pitch and timing are different - in the digital domain they are limited amongst other things by the Nyquist sampling theory.

Therefore a 96kHz Nyquist bound sample stream, bound by 20kHz pitch in the analogue domain can resolve to 2.08 x 10^-5  seconds

S

Mulberry posted:

Hi Mike, I think it is just the other way around. DSD has a quite aggressive noise shaping towards the upper frequencies. The dynamic range is very high at its lower end (which can extend into  the +20 kHz range), but virtually noexistant above that. PCM has a linear range that can/will be lower in some parts of the frequency spectrum.

............  oh err that did not read as I was intending ....... trying to keep it simple.   What I was trying to say ........  as I understand it .......... DSD noise shaping has to be done post recording - it has to be to reduce the >25kHz noise,  whereas PCM can be part of the recording process

Mike - noise shaping is part of the ADC and DAC process for DSD (ie its part of the reconstruction filtering). PCM and DSD have different reconstruction filter methods, but the analogue signal simply doesn't care how it is encoded or reconstructed... Filtering for PCM is typically done outside of the recording process too so as to avoid aliasing noise. (which sounds horrible) Ideally construction and reconstruction filters should match for optimum accuracy and quality - whether they be DSD or PCM

S

Mike-B posted:

     Hi Bert,  you might try a search of the forum as there are a lot of threads on the subject.  Some posts explain the technicalities if you are interested,  & many are concerned with which sounds best - worth the time to search.

I've been using DSD since it was first intro'd into the NDX beta firmware & quickly came to an opinion on it.  I believe that generally with DSD, the people who record with it tend to be very serious about producing a high quality recording & this is probably the main reason I hear something better/different with DSD.   It seems well suited when DSD (DSX) is used throughout the whole recording process with small acoustic groups.  Whereas I understand larger groups such as full orchestra are easier/better recorded in PCM & changed to DSD at some point during mixing/mastering.  

In my system DSD sounds very different to PCM (I am all WAV) Its very clear to the point its not really debatable. At first it seems the volume is too low, but the dynamic transients are so much more …. startling & loud is probably the most descriptive. These dynamic peaks have no hint of distortion or compression & all very true to life. The sound in general has more of a feeling of analogue sound than anything I've heard from other digital, a higher level of realism with both instruments & voice. This definition is present through all the frequencies, bass wood & string timbre is clear & you can hear the cymbal brass ring & the wooden stick hitting it. Sound stage is noticeably better with a very tangible three dimensional presentation, fixed positions & space between & a great sense of distance. 

I would be interested in your opinion once you've tried it. 

Mike, Thanks for sharing your listening/hearing experiences.

Agree that a difference in sound quality between DSD and wav is not really debatable.

There is a lot of discussion on the technical aspects of different formats and mixing mastering but in my listening experiences flac, wav and dsf files of the same recording (read: master = flac file, and the wav and dsf are transcoded from this master) all have other sound qualities.

Independent whether the source is on a USB memory stick or UPnP, the three file formats have there own "character of sound".  Think I prefer the "*.dsf format" because it's the more delicate, detailed and stresless sound (to my ears).

Considering to transcode my most appreciated music collection to *.dsf...................

Interesting contributions, thanks for this, it gives a good insight. I will have a listen myself and hope that one of the tracks which I have in DSD is also available on my NAS in FLAC format - that would give me a good comparison.

While the concensus seems to be that DSD is better.

You should pick whatever sounds best to your ears in your system, and it may be either DSD, PCM 44.1, or  PCM 192. In reality though we mostly choosing between different masterings rather formats themselves. Sure, there may be sites where you can download same song from exactly same mastering just in different formats and make fair comparison.

Andrius posted:

You should pick whatever sounds best to your ears in your system, and it may be either DSD, PCM 44.1, or  PCM 192. In reality though we mostly choosing between different masterings rather formats themselves. Sure, there may be sites where you can download same song from exactly same mastering just in different formats and make fair comparison.

Hi Andrius,

Fully agree that we should pick whatever sounds best to our ears.

The choice of a mastering however is random; We cannot know which one we like the best until after we got them downloaded and give ourselves the opportunity to compare one mastering to another.

We can stream the master in its native format butthe choice on another format is easily and quickly provided by transcoding the master ourselves to whatever (mp3, flac, wav, etc.) format we like best.

For me this was wav but dsf format now. Trying to find an affordable software trancoder.................

Mulberry posted:

Bert Schurink posted:

[...] If you look at the pure size [...]

Hi Bert, a FLAC file will always be smaller because it is compressed (think .zip here). This is very much like WAV/AIFF (uncompressed) on one hand and FLAC/ALAC (compressed), all of them PCM files, on the other.

If you take a look at the downloads at Hifistatement, there is a piece by  Michel Godard in 44/16 PCM, 24/192 PCM and 2.8 MHz DSD. The respective file sizes are 70.6 MB, 460.8 MB and 283.5 MB.

Mulberry, not all FLAC files are compressed. You have the option of selecting one of ten "options" when encoding - 0 (75% compression) to 8 (70% compression) - but the last FLAC option is 'uncompressed', which will therefore be the same size as the WAV file.

If the DSD or PCM are from the same master ... Unless it's an analogue master which are rather rate these days I would stay well clear of unless you are limited on playout options with your replay equipment. Converting between PCM and DSD  is an inexact lossy process. A genuine  DSD recording will have been encoded as DSD (or more typically a DSD variant such as DSD-wide ) from the start. Like wise PCM is best remaining in PCM from recording/encoding to playout.

As I say PCM and DSD are two different sampling format types, it's not simply a digital encoding type and converting from one to the other lto varying degrees ooses information and/or adds artefacts.

Simon

not many will have many dsd files/high rez  compared with cd rips .so it is not a issue for most.   I cant detect any difference between 96/192 ,but dsd files are the best in my system.

The biggest con is the cost of all high rez files, period

Simon-in-Suffolk posted:

If the DSD or PCM are from the same master ... Unless it's an analogue master which are rather rate these days I would stay well clear of unless you are limited on playout options with your replay equipment. Converting between PCM and DSD  is an inexact lossy process. A genuine  DSD recording will have been encoded as DSD (or more typically a DSD variant such as DSD-wide ) from the start. Like wise PCM is best remaining in PCM from recording/encoding to playout.

As I say PCM and DSD are two different sampling format types, it's not simply a digital encoding type and converting from one to the other lto varying degrees ooses information and/or adds artefacts.

Simon

Simon, can confirm your technical explanation, converting *wav files to *dsf is creating (false) information that is not in the original file.

However, recently converted several flac files to dsf using a free (result = 3 seconds of silence in the middle of the converted track) test version of AuI audio converter.

Maybe information was lost during this conversion process or maybe artefacts are introduced.  As it comes to enjoying music I only trust my own ears.  In all cases I prefer the "*dsf sound". 

Peter

Bowers posted:

Simon, can confirm your technical explanation, converting *wav files to *dsf is creating (false) information that is not in the original file.

Simon is correct ... converting WAV to DSF (or more correctly converting PCM to DSD) is a lossy process in that you cannot take a PCM file, convert it to DSD and then convert back to PCM and have the second PCM file identical to the original.  There is always something added and something lost in the conversion.   Its like (in very simple terms) taking 1 and dividing it by 3; storing the result in decimal (0.3333) and then multiplying it by 3.  The final result is very close to 1 but not exactly 1.

However, recently converted several flac files to dsf using a free (result = 3 seconds of silence in the middle of the converted track) test version of AuI audio converter.

Maybe information was lost during this conversion process or maybe artefacts are introduced.  As it comes to enjoying music I only trust my own ears.  In all cases I prefer the "*dsf sound". 

At the end of the day, if you enjoy listening to the music more once converted to DSD thats one thing.  But (a) you have to understand that conversion is one way so you should keep a copy of the original and (b) that is a result of the way your DAC processes DSD vs PCM rather than any inherent superiority of DSD over PCM.  Its the same as preferring a DAC which does up sampling compared with one which doesn't.

Eloise

Hi Eloise,

You, Simon and I all agree that technically the PCM to DSD process of this coding software might not be reversable and that information is changed/added. No debate here (I think).

There is no statement that DSD is superior to PCM; In my set-up and through my ears I (very much) prefer the converted "dsf sound". ("a layer of distortion seems be peeled off")

This topic is about preferring a format and because the goal is to enjoy our beloved music, I think we can stop discussing the technical facts (there are more than enough statements about this on the internet). In the end it's what we enjoy the most and this might be very different for each of us. Think this might be interesting.........

Please share your own preferences and experiences on format.  

Hi Peter, no issue at all in modifying the audio for it to sound preferable on a given system.. but I guess for those less well informed they my be concerned about such modification to the audio.. It's clearly far from being bit perfect, just like some would be wary of using tone controls on their system.

Simon

Simon, it might have become clear that I am not a "bit perfect" addict.  

"Bit perfect" is all technical/digital matter that we can explain and understand but for me this is no holy grail.

My ears are not "bit perfect" and there are no "bit perfect" amplifiers or loudspeakers .

We pick our streamers, amplifiers, loudspeakers, cabling etc. etc. mostly by listening and  every component and process adds distortion. The listener decides whether he likes the distortion or finds it is irritating.

I have always heard irritating distortions as CD sources were involved while some of my friends fully enjoyed the sound, so our hearing must be personal/individual.

I would like to challenge you to transcode some of your own music to dsf, have a good listen in your set-up, and share your experiences.