Hi all,
The majority of all DAC are delta-sigma DACs. As I know - please correct me if I'm wrong - the process of the delta-sigma DAC is technically the same as "native DSD".
The "PCM to DSD" convertion is done in the DAC to fs 64, 128 or 256 anyway.
So, what's the point of "upsampling" PCM to DSD in software?
Hi all,
The majority of all DAC are delta-sigma DACs.
True.
As I know - please correct me if I'm wrong - the process of the delta-sigma DAC is technically the same as "native DSD".
False.
Native DSD is a special case of Delta-Sigma, more properly called Single Bit Delta Sigma. This special case has a number of built in severe technical problems that are described here: Free copy of Vanderkooy-Lipshitz paper explaining difference between Delta Sigma and DSD (http://sjeng.org/ftp/SACD.pdf)
The background behind this is that real world Delta Sigma converters are not all the same. One way that they differ among themselves is the number of bits in the DAC that is at their core. As the reference points out, the Single Bit Delta-Sigma is inherently and seriously flawed in a number of serious ways, but the multi-bit versions of it avoid these problems.
Shortly after the publication of paper referenced above it was revealed that in fact the some or all of the allegedly Native DSD converters used in certain professional DSD products were strictly speaking not Naive DSD but were actually multi-bit devices.
To me the most important take away is that allegations of superior sound quality due to DSD have to be false, because if you believe some sources Native DSD is the same as Single Bit Delta-Sigma which is inherently flawed, or actual implementations of it are actually the same as multi-bit Sigma Delta which is what it is supposed to be superior to.
Logic says that for A to be superior to B it must be different from B which in the case of DSD, it is not.
I am unaware that there have ever been any DBTs that compared Sigma-Delta and DSD DACs with similar performance in the normal audible range and found any audible differences.
I bet some people that upsample think to bypass the evil PCM reconstruction filter when it becomes dsd while in reality this very conversation applies one.
Others argue the offline reconstruction filter is better as the internal DAC one so they skip that and use the DSD part of the DAC chip.
Self proclaimed golden ears claim it sounds better or muddens the sound in a way it sounds better to them.
I did read forums over several years now. It is the same species "bellieving" in this pretty new audiophile fashion while in reality they are unable to do stand the most simple abx test.
I bet some people that upsample think to bypass the evil PCM reconstruction filter when it becomes DSDwhile in reality this very conversation applies one.
No doubt.
It all starts out with the naive audiophile false belief that their current DAC is the major barrier to obtaining good sound from their audio system.
Others argue the offline reconstruction filter is better as the internal DAC one so they skip that and use the DSD part of the DAC chip.
Ditto.
Self proclaimed golden ears claim it sounds better or muddens the sound in a way it sounds better to them.
If you watch these guys on the audiophlie forums, that phase may only last for a little while, and then the siren song of the new improved DAC they don't have becomes the dominant force in their decision making process.
I did read forums over several years now. It is the same species "bellieving" in this pretty new audiophile fashion while in reality they are unable to do stand the most simple abx test.
Interesting how these self-proclaimed eggspurts lack the expertise to download and proprely use FOOBAR2000 and the ABX plug-in. ]
Interesting how these self-proclaimed eggspurts lack the expertise to download and proprely use FOOBAR2000 and the ABX plug-in. ]
It looks like there are at least a couple forum members who've at least figured out how to download foobar2000 (to be stylized in all lowercase).
Native DSD is a special case of Delta-Sigma, more properly called Single Bit Delta Sigma.
[...]
The background behind this is that real world Delta Sigma converters are not all the same. [...] the Single Bit Delta-Sigma is inherently and seriously flawed in a number of serious ways, but the multi-bit versions of it avoid these problems.
Shortly after the publication of paper referenced above it was revealed that in fact the some or all of the allegedly Native DSD converters used in certain professional DSD products were strictly speaking not Naive DSD but were actually multi-bit devices.
Thanks for pointing me to the multi-bit delta-sigma DACs.
My Conclusion: DSD/SACD is a hype. It's inferior to PCM, need more bandwith/storage. In addition any editing, sound manipulation etc. is not possible. DSD is a pain.
It's inferior to PCM, need more bandwith/storage.
I say the same thing about samplerates >48kHz and/or resolution >16-bits (especially for noisy media or content) but it doesn't seem to be very persuasive. Storage is cheap, they say.
Storage is cheap, yes but when i see people meanwhile use 5GB for a dsd128 vinyl rip i feel a bit baffled and wonder where this nonsense ends.
The question is where does the nonsense start. Queue up the arbitrary and not based on sound quality hand waving in 3, 2, 1. . .
I use 16/44.1 for everything. I buy higher bitrates when i know it is better sounding as some other version, less compressed or mostly the same price. Even then i dither 24/44.1 and high samplerates to 16/44.1 for use on my server and for portability. The originals go to backup drives. In all these years i have never found anyone convincing me it needs more. Seems that i am a default mortal. Recent AES papers don't really convince me. I could believe if someone with the caliber of a Guruboolez and his detailed private listening tests evidently finds some tiny advantages here and there but that never happened...
The quantization adds lots of noise. This noise needs to go somewhere. Ironically, it is pushed to higher frequencies.
If you look at music (what DSD is used for) then you will notice that the energy of the signal drops especially beyond 20 kHz. So up there what's left of the signal (deliberately not calling it "music") is very low level and mostly noise.
Now audiophiles should want to reproduce that accurately but DSD's quantization noise will more and more drown this part of the signal with increasing frequency.
That's why it's also funny when these people use plots that show how impulses look "better" with DSD. Such an impulse will contain energy up to several hundred kHz, theoretically MHz... You don't see that in music.
I have no idea why anyone would use DSD unless he/she is forced to. :P
I use 16/44.1 for everything. I buy higher bitrates when i know it is better sounding as some other version, less compressed or mostly the same price. Even then i dither 24/44.1 and high samplerates to 16/44.1 for use on my server and for portability. The originals go to backup drives. In all these years i have never found anyone convincing me it needs more. Seems that i am a default mortal. Recent AES papers don't really convince me. I could believe if someone with the caliber of a Guruboolez and his detailed private listening tests evidently finds some tiny advantages here and there but that never happened...
Because I was raising issues with 24 bit here in the past, I would like to add to this valuable experience also the fact that during this summer I have digitized/recorded some audio at home with good DACs (ALC 898 and CS 4398) and passed them to my friends in both 24 bit and 16 bit FLAC. Unfortunately nobody of them was able to hear/claim the difference on playback (I chose either TPDF dither or very low noise shaped dither). I personally still keep some albums/tracks in 24 bit even for playback, but I have to admit I did not get objective evidence for that practice. To be honest, I still have some "good feelings" about 24 bit tracks (without dither), but I cannot prove them.
The only new hypothesis I was thinking about with 24 bit playback recently is if the oversampling/upsampling done in some DACs does not work better (theoretically or empirically) in case the DAC is being fed with 24 bit "original" content instead of 16 bit "dithered". But it may be a nonsense.
I personally still keep some albums/tracks in 24 bit even for playback, but I have to admit I did not get objective evidence for that practice. To be honest, I still have some "good feelings" about 24 bit tracks (without dither), but I cannot prove them.
Easy to prove. Use very silent music and play them at unrealistic loud levels. The music will be the same but with noise. Reality could be it is pretty hard to find anything recorded with such a low noisefloor while music plays.
The question is where does the nonsense start. Queue up the arbitrary and not based on sound quality hand waving in 3, 2, 1. . .
"good feelings" about 24 bit tracks
...I swear it's like a placebophile dog whistle, and some fool comes running.
I cannot prove them.
Of course you can't and TOS8 is in place to spare the rest of us from your delusions, but you just can't help yourself.
Unfortunately nobody of them was able to hear/claim the difference on playback
Unfortunately?!? Being aware of your limitations should be a
good thing.
(without dither)
So all this wanking in other discussions over optimal noise shaping was nothing but hot air?
new hypothesis I was thinking about
Will you ever be done grasping at straws?
(theoretically or empirically)
More hand waving. Try neither.
But it may be a nonsense.
Bingo.
You should wait to tell us about it when you're in a position to demonstrate what you have to offer isn't nonsense.
8. All members that put forth a statement concerning subjective sound quality, must -- to the best of their ability -- provide objective support for their claims. Acceptable means of support are double blind listening tests (ABX or ABC/HR) demonstrating that the member can discern a difference perceptually, together with a test sample to allow others to reproduce their findings. Graphs, non-blind listening tests, waveform difference comparisons, and so on, are not acceptable means of providing support. (https://hydrogenaud.io/index.php/topic,3974.msg149481.html#msg149481)
It's inferior to PCM, need more bandwith/storage.
I say the same thing about samplerates >48kHz and/or resolution >16-bits (especially for noisy media or content) but it doesn't seem to be very persuasive. Storage is cheap, they say.
Yes it thankfully is and it could/should be used for storing the content the user can make use of. But when we agree/know that >48 kHz is complete overkill and >16 bit is practically irrelevant for playback, then I fully support that we should not waste the space we have, e.g. for common playback or mobile use. The argument that storage is cheap is relevant only when we store something that has some value for us - it only says that the cost of storage is so low that we do not have to judge what we will store and what not as it was in the days of 128,192,256 and 320 MP3s.
the days of 128,192,256 and 320 MP3s.
When my main portable media player is capped at 16 or 32 GB, lossy compression is still a very beautiful thing.
~170kbps mp3 delivers near if not full transparency; even lower with aac and vorbis; and here you are "craving" several times that amount in order to satisfy your placebo. I'm happy being satisfied in wearing more sensible shoes.
To the previous reaction: I really do not think that I am on the "placebophile" side now and my posts are not meant in "pushing" those things anymore. The hypothesis I mentioned was just a thought, I have no problems with it being claimed as false.
the days of 128,192,256 and 320 MP3s.
When my main portable media player is capped at 16 or 32 GB, lossy compression is still a very beautiful thing.
~170kbps mp3 delivers near if not full transparency; even lower with aac and vorbis; and here you are "craving" several times that amount in order to satisfy your placebo. I'm happy to be satisfied in my sensible shoes.
No problem with that, as lossy compression works nicely with practical limits of human hearing and still may satisfy it in many situations. I also recently made some 320 kbps MP3 for my friend, whose minitower does not support FLAC on USB playback and received no complaints from him.
I really do not think that I am on the "placebophile" side now
Your previous post about "unfortunate" results tells an entirely different tale.
I really do not think that I am on the "placebophile" side now
Your previous post about "unfortunate" results tells an entirely different tale.
Yes I have to admit that in the summer I when I was giving those home recorded tracks to my friends was kind of unhappy that they judged 24 bit and 16 bit as equal when listening. But it is a fact, so I have accepted it.
320 kbps MP3
Did you try to find your friend's threshold of transparency? What about your own?
But it is a fact, so I have accepted it.
Why do you judge it today as unfortunate?
But it is a fact, so I have accepted it.
Why do you judge it today as unfortunate?
When I wrote that post I remembered that experience I described form the past. And still if somebody could demonstrate (e.g. by ABX) the benefit of 24 bit on some real common track, it would positively surprise me. Nothing wrong about it, I think.
Nothing wrong about it at all! That's why I have a difficult time seeing an increase in self-awareness being assessed as "unfortunate".
...but hey, I want to make the most of my 384kHz fucktard DAC because that's what it can do and it's hopefully better than super-human hearing even though I know fuck all about what that means.
I have no 384kbps DAC. For recording I use 48 kbps. I have tried in the past to find my listening threshold about 256 kbps MP3 with some sample tracks but I do not think that is the only relevant thing for audio storage. I mentioned my hypothesis just because tjis topic was about upsampling. The rest is hard to discuss when the opponent attacks the messenger.
I have no 384kbps DAC.
Sorry, I meant 384kHz. EDIT: I also explicitly mentioned DSD in the same breath. Sigh. I can't manage to get this nonsense straight anymore.
The rest is hard to discuss when the opponent attacks the messenger.
I'm attacking your inability to operate within the confines of reality. Honestly, I'd prefer to be your proponent, but you keep clinging to the idea that sonic bliss lies beyond the horizon.
As for the topic, the OP seemed pretty clear:
https://hydrogenaud.io/index.php/topic,112954.msg930001.html#msg930001
Ok. I have no other input, if my hypothesis about oversapmling/upsampling source is false then the rest was just an appreciation that i have not find any objective evidence since our debate in the spring.
I have no 384kbps DAC. For recording I use 48 kbps. I have tried in the past to find my listening threshold about 256 kbps MP3 with some sample tracks but I do not think that is the only relevant thing for audio storage. I mentioned my hypothesis just because tjis topic was about upsampling. The rest is hard to discuss when the opponent attacks the messenger.
And I by mistake wrote about kbps (while thinking about those MP3s ... ) where should be kHz :) :)
You quoted my dumb mistake which I fixed. I should have gone to bed.
To your earlier point, sure, I won't dismiss the possibility that there are DACs which require upsampling in order to be transparent because they are incompetently designed.
From PCM to DSD? I have no idea. Good luck finding a single advocate for this method who actually performed a proper listening test.
OK. I was talking about PCM upsampling only (from 16bit/24bit source), as I have no experience with DSD (and from what I have read I do not miss anything by leaving DSD/SACD aside).
I was talking about PCM upsampling only (from 16bit/24bit source)
That is not upsampling. Regardless, that has already been discussed to death, and this topic is not the place to have another round at beating the horse.
Wow! Reading greynol's comments in this thread reads exactly the same as most of the comments I've been making on a couple of 'audiofool' forums I've been stupid enough to join whilst taking a hiatus from here. It's good to be back! :))
I want 384kHz. ADC. For higher work precision when recovering analog media with physical damage. Of course, chances are, even with the RX 5 suite of tools, it's totally pointless, and serves nothing better than the 44100/16 recordings I already made and cleaned up. They even satisfy me, except for the fact that this record seems to be damaged or something, as the first 25-30 seconds seem to have some sort of saturation going on, without actually hitting the recorder's clipping threshold. Oh well, I'm satisfied with what I've got.
(I'd be even more satisfied if the original studio still had some master tapes and produced a CD, so I didn't have to hit up Discogs for an aging record release. Not like I could expect K-Tel, the company that brought us "As Seen On TV" product branding, to find a 1980 children's record to be marketable to today's generation. I wouldn't expect them to even bother putting it on streaming services. Got my nostalgia on, and moved on with my life.)
chances are [...] it's totally pointless
This can be easily tested.
I can't actually test whether the signal processing software has a better chance of processing pops and clicks at that sample rate, as I don't have the hardware to record over 48KHz from a turntable. I suppose it can easily be tested by someone with such insane hardware, but that same person is probably likely to have their opinion colored by their hardware choices and budget.
According to Cookie Marenco, Blue Coast Music will create DSD from PCM "for your convenience" and they note it on the top of the page where you buy the stuff. I take it that "for your convenience" means just that, convenience only and that there is no reason to do the conversion otherwise. What's the point?
I want 384kHz. ADC. For higher work precision when recovering analog media with physical damage.
Why?
Of course, chances are, even with the RX 5 suite of tools, it's totally pointless, and serves nothing better than the 44100/16 recordings I already made and cleaned up. They even satisfy me, except for the fact that this record seems to be damaged or something, as the first 25-30 seconds seem to have some sort of saturation going on, without actually hitting the recorder's clipping threshold. Oh well, I'm satisfied with what I've got.
It is possible that certain tools that are sold for the purpose of recovering analog media might be themselves poorly made. and might benefit from addition of huge amounts of excessive bandwidth.
It is likely that well-designed products of this nature can be highly effective, even if operating with less bandwidth and lower resolution than 44/16, given that 44/16 is itself sonically an overkill format.
Sounds to me like the product you are using has serious problems that even more bandwidth couldn't help.
http://www.soundonsound.com/sound-advice/q-are-wow-and-flutter-key-analogue-tape-sound
someone had invented a de-wow-and-flutter system that tracked variations in the pitch of the bias signal to correct for wow and flutter, and he said the result sounded 'just like digital'
How about this? Is it real? I heard about this technique before but I don't know if an actual product exists or not. From what I know about analog tape, bias signals can be as high as 100khz or more right?
But there is also a product claims to correct wow and flutter without using the bias signal.
https://youtu.be/qqK6wgsh3QA
http://www.soundonsound.com/sound-advice/q-are-wow-and-flutter-key-analogue-tape-sound
someone had invented a de-wow-and-flutter system that tracked variations in the pitch of the bias signal to correct for wow and flutter, and he said the result sounded 'just like digital'
How about this? Is it real? I heard about this technique before but I don't know if an actual product exists or not. From what I know about analog tape, bias signals can be as high as 100khz or more right?
But there is also a product claims to correct wow and flutter without using the bias signal.
https://youtu.be/qqK6wgsh3QA
There are several products that claim to do this.
One is Celemony Capstan: Celemony Capstan (http://www.celemony.com/en/capstan)
The other is Plangent: Plangent Processes (http://plangentprocesses.wixsite.com/plangent)
There may be others
Thanks Arnold. So the Plangent Processes is a hardware system capable of capturing the bias signal and this piece of specifically designed hardware should be much more expensive than a 384khz ADC. I ran some 96k RMAA tests on a cassette deck and it started to lowpass at 15khz and after 20khz there is only noise, so I suppose the bias signal cannot be reproduced by a typical cassette deck or open reel tape recorder at all.
Thanks Arnold. So the Plangent Processes is a hardware system capable of capturing the bias signal and this piece of specifically designed hardware should be much more expensive than a 384khz ADC. I ran some 96k RMAA tests on a cassette deck and it started to lowpass at 15khz and after 20khz there is only noise, so I suppose the bias signal cannot be reproduced by a typical cassette deck or open reel tape recorder at all.
Yes, either process needs some kind of a reference tone that "in the wild" can reasonably expected to be relatively free of jitter at the frequencies that are present in the recording to be de-jittered.
I believe that Plangent Processes has some more detailed white papers (AES? IEEE?) about their procedures. I seem to recall their needing to use custom reproduce heads with exceedingly narrow gaps.
There are also papers with some details about Celemony's process, which I believe uses the components of the music itself as the source of a jitter-free reference.
Given that we can measure jitter to far lower levels than we can hear, this method might work as well as Planget's without special hardware.