I have read that MP3 SQ can be improved by up-scaling it to DSD, and that this could be done within Foobar. I wondered if anyone had tried this, and if so how did you do it and was it worth it?
Mitch
Hi Mitch,
who are these guys believing that decoding from a lossy format to whatever lossless format could improve quality? They tell you some alternate truth. The only thing that rises will be the bitrate ;)
There's only one way to improve the SQ of MP3 (assuming it's non-transparent versus the original). You'll need to find a way to reverse time and add back the data that the encoder discarded. If you find such a technique, please don't keep it to yourself. Reversing time would be something a great many people would be interested in!
In the real world, compressing to MP3 is a one-way, lossy process that cannot be undone. Any quality reduction on encoding (audible or not) is impossible to reverse. If you've "read somewhere" to the contrary, it's safe to dismiss the notion as not based on reality.
Thanks for the replies. I am sorry but I was inaccurate in my choice of words in my first question. The iFi web page does not say it improves sound quality, but that by changing to the DSD format it removes some of the harshness from the mp3 format sound, thus making it more 'analog' sounding and thus, for me at least, more pleasant to listen to. I appreciate that the lost bits can never be recovered. I suppose that the conversion is something like applying a filter to the sound (?).
Anyway - I cannot try it out for my self because I cannot find a way to convert from mp3 to DSD, and this is where I would like some help if possible please.
Mitch
It won't sound any different. DSD and PCM will produce identical audio. The DAC chip in your sound card will convert PCM to similar presentation as DSD is without wasting disk space or consuming extra system resources.
The iFi web page does not say it improves sound quality, but that by changing to the DSD format it removes some of the harshness from the mp3 format sound, thus making it more 'analog' sounding
Usual audiophoolery.
To convert PCM/mp3 to DSD you need foobar2000 with foo_input_sacd and foo_dsd_processor - https://sourceforge.net/projects/sacddecoder/files/foo_input_sacd/
Using this method it is impossible to save result to disk. Conversion can be done on-the-fly during playback only.
Thanks again guys,
Just so you don't think I am making this up, I thought I would include the link to the iFi Technical Note (hope I am allowed to do this);
https://ifi-audio.com/wp-content/uploads/2018/07/JRMC-MP3-to-DSD512.pdf
The method does use the Foobar plug-in, but as far as I can see, it only runs in JRMC (which I do not have) and although I have the iFi Micro USB, unfortunately I do not have one of their DACS. There are a couple of professional software suites that may do the conversion, but they are very expensive. I had hoped that by just using the Foobar plug in I could do it totally within foobar using my own DSD capable DAC.
From your advice however it seems like I might be wasting my time, but even so, just to appease my own curiosity, I would like to have a go if anyone could tell me how (without buying JMRC, an iFi Dac or expensive software).
On the plus side, the work on the time machine is going well, just a few minor issues to resolve. However, as interesting as I find digital audio, my first trip will be back to 1960 to buy as many new Decca SXL recordings as I can!
Thanks all,
Mitch
https://ifi-audio.com/wp-content/uploads/2018/07/JRMC-MP3-to-DSD512.pdf (https://ifi-audio.com/wp-content/uploads/2018/07/JRMC-MP3-to-DSD512.pdf)
Seems like a load of BULLSH-T.
If your MP3 sound harsh, they will of course show the same harshness after creating DSD out of them. And why DSD? Linear WAV surely has at least the same capabilities when using higher bit depth and sample rate (I surely do NOT recommend doing so)
The only thing that I can imagine might possibly help - apart from the obvious bullshit-enhanced placebo - is if the MP3 file exceeds digital full scale and clips upon playback, AND somewhere in the conversion/playback chain there is something that will do the right thing with DSD but not with MP3.
A peak-aware player avoids this. In foobar2000, use ReplayGain tags and under Preferences -> Playback at the second-to-top pull-down, choose an option that prevents clipping according to peak. Which is a good thing to do in any case, so I take the opportunity to recommend it even when the issue is total bogus and likely borders on fraud ...
Edit: Just for the sake of the bullshit here: Does not fb2k work in 32-bit floating-point internally? So if upsampling is good for your placebo, then you should get it automatically! :-D
This is like copying a VHS tape to Blu-Ray or DVD... You don't get Blu-Ray/DVD quality. Except anybody can play a standard format shiny disc but you're stuck with this oddball DSD format.
MP3 is lossy, but a good quality MP3 can often sound identical to the uncompressed original, even a high-resolution or DSD original, in a blind ABX test (https://hydrogenaud.io/index.php/topic,16295.0.html).
There is one thing you can do to "partially restore" an MP3 - There is something called a harmonic exciter effect (https://en.wikipedia.org/wiki/Exciter_(effect)) that generates high frequency harmonics. It can be used to replace the high frequencies that are thrown-away by MP3 compression. The big problem is, if you are hearing compression artifacts it's usually not the loss of high-frequencies you're hearing, and the other problem is that it's impossible to know what the original high frequencies were so you can't restore the original sound. But, still you can "improve" or "enhance" the sound, just as you can enhance the sound with EQ any other effect. And, you can save-as WAV or FLAC, or make a CD, etc. There's no need for DSD.
There is a free exciter plug-in (https://wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#Harmonic_Enhancer) for Audacity (also free) if you want to play-around with it.
Right, this is beginning to get beyond my level of competence, but nevertheless very interesting - it is great that one can learn so much from this Forum - thanks very much.
Just about to go and download Audacity and the plug in - a new strand of interest! However I still cannot help but wonder why the blind testing the iFi guys did produced positive results. I cannot see how the article in the Tech Note would be a sales pitch as recommending a conversion like this would not help sell their DACs - they must have heard something - but WHAT?
As a 'city boy' not sure that I would know what bull s..t sounds like - is it better than mp3 ???
Mitch
A peak-aware player avoids this. In foobar2000, use ReplayGain tags and under Preferences -> Playback at the second-to-top pull-down, choose an option that prevents clipping according to peak. Which is a good thing to do in any case,
It is good practice though the effect of encoded data going over full scale being audible over a listener’s entire library isn’t all that likely.
However I still cannot help but wonder why the blind testing the iFi guys did produced positive results.
Blind doesn’t mean the test was properly controlled.
removes some of the harshness from the mp3 format sound, thus making it more 'analog' sounding
Yes, the all too common ignorant placebophile bullshit. Nothing new here.
Why only “some?”
The iFi web page does not say it improves sound quality, but that by changing to the DSD format it removes some of the harshness from the mp3 format sound, thus making it more 'analog' sounding and thus, for me at least, more pleasant to listen to.
The iFi page is incorrect, whether deliberately so or not. The red herring there (or bait?) is the "analog" sound claim. You cannot hear bits and bytes, you can only hear analog sound, so ALL sound is analog. I suspect that the "harshness" alluded to is an indirect reference to the long-held and erroneous idea of "stairsteps" in digital, making it harsh. The idea that digital is in any way harsh (or stair-stepped) just because it's digital is completely false and only shows a lack of understanding of A/D and D/A conversion.
By all means play with the idea for your own curiosity but beware of convincing yourself that there's an improvement. Only a properly controlled test will establish that (e.g. ABX).
Why only “some?”
Huh. I didn't see that nuance at ifi-audio.com, and lo and behold, that document says
"takes the digital harshness out of the music".
Everyone knows that it isn't the significant learning from the thread, so unless the OP feels like clarifying whether the "some" was injected due to not fully buying the claim or whatever ...
The OP needn’t clarify anything.
For those who still get even a modicum of entertainment the use of “some” was pretty obvious: digital is inherently harsh so mp3 can only be made less harsh.
(https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcTFD_M911OLAkZqMKSw6IIabVf596PTmY5ciomDkEHf93Y1mdd6)
Pavlovian
Someone else us now asking a similar question:-
https://www.computeraudiophile.com/forums/topic/55026-mp3-to-dsd-conversion/
For those who still get even a modicum of entertainment the use of “some” was pretty obvious: digital is inherently harsh so mp3 can only be made less harsh.
Digital is definitely not 'harsh'
[MODERATION]
I'm terribly sorry, I meant to reply to your post, not edit it. You spent a great deal of tmie writing your response. I was only be facetious, however.
I don’t believe the forum software has the ability to revert changes. If you happen to have a saved copy in your drafts I can restore it to that point.
-greynol
Thank you for taking the time to reply. I understand what you say to some extent, although I would not know how to go about setting up the listening test you mention in your final para.
Primarily I am a 'vinyl -man', and so I am well aware of the difference between good and bad recordings on vinyl (not why they are good or bad you understand - just that some sound a whole lot better than others), this is why I have been collecting early Decca SXL recordings for many years. I have had a CD/SACD player for a long time, but when it comes to choice of listening medium, it has always been vinyl. Now however I am trying to understand more about digital audio.
I have always preferred SACD to CD, because I felt, that in general, the quality of the mastering and recording were more consistently good, and the sound was a little more like vinyl I suppose. I have a fairly decent collection of DSD128 and DSD64 files (the 64s are mainly ripped from my now redundant SACD collection) and if not listening to vinyl it is these DSD files I tend to play - I have not listened to a CD for quite a while.
I also have a large collection of mp3 and ripped CDs which I do not listen to at all - and so when I saw the iFi Technote saying that by converting the mp3 to DSD it removed some of the harshness I thought it might be worth a try.
All the comments above seem to suggest that this may not be so - but that is academic, because I still do not know how to convert them and give it a try!
Thanks again
Mitch
Rollin told you how to convert everything you play in foobar2000 to DSD on-the-fly. But if you don't have DSD capable DAC you'll need to convert DSD back to PCM. Note that the audio will still not change though, it's just resampling and good resampling has always been transparent.
Funny how some attribute a sound to mp3. There were listening tests people prefered the mp3 over the original not because it was harsh suddenly. What mp3 surely does is softening transients so it is likely the other way around.
One should try to build up a simple comparison on his own before doing stupid upsampling because someone said so.
Rollin told you how to convert everything you play in foobar2000 to DSD on-the-fly. But if you don't have DSD capable DAC you'll need to convert DSD back to PCM. Note that the audio will still not change though, it's just resampling and good resampling has always been transparent.
I do have a DSD capable DAC which plays my DSD files perfectly. I had hoped to be able to convert the mp3 files 'off-line' , so to speak, into DSD files in ,maybe the mistaken, assumption that as the processor will not be working as hard as it would converting 'on -the-fly' there would be more chance of good sound. However I am grateful to Rollin for this - I have the plug-in and will try to work out how it can do this conversion - I know I should not expect any improvement, but at least, hopefully, I can try it for myself.
thanks Mitch
,maybe the mistaken, assumption that as the processor will not be working as hard as it would converting 'on -the-fly' there would be more chance of good sound
There's no maybe about it. Yet another audiophile myth. Same as people who seem to think that playing files from different locations on a HDD gives a different sound.
But we all have to learn somewhere :)
Want to mention the recent SoX at sourceforge does PCM->DSD conversion.
Oh no!! I had not heard of the HDD one - not another audiophile myth I might succumb to!!
The trouble is, that like a lot of people with a life-time of 'fiddling' with turntables, I cannot help tweeking. Unfortunately, the technology of my Garrard 401 is a little simpler than the mysteries of digital.
I think it is time for me to give up and say thank you to all those who have tried to educate me - I am grateful for your time and I have learned a lot. I know that on forums like this as well as asking questions, one should try and contribute to the problems of others - I feel a little guilty about this, but it will be some time before I know enough to be confident to offer any meaningful inputs - sorry.
While I was typing this a post from Wombat came in telling me of of some software at Source Forge. So about that giving up........
Mitch
While I was typing this a post from Wombat came in telling me of of some software at Source Forge. So about that giving up........
Relax. You don't it. Just let your media player play your music.
As we have said for many years, CD is able to achieve an audibly faithful replica of what is on vinyl and do so consistently. The same cannot be said about going in the opposite direction.
The Meyer and Moran study shows that the first portion of this exists outside of academia.
For those who still get even a modicum of entertainment the use of “some” was pretty obvious: digital is inherently harsh so mp3 can only be made less harsh.
Digital is definitely not 'harsh'
[MODERATION]
I'm terribly sorry, I meant to reply to your post, not edit it. You spent a great deal of tmie writing your response. I was only be facetious, however.
I don’t believe the forum software has the ability to revert changes. If you happen to have a saved copy in your drafts I can restore it to that point.
-greynol
I apologize for my strident tone anyway. I sometimes take things too seriously. I REALLY CARE -- and sometimes I am a little too vehement. That 'harshness' or 'starstep' idea (myself as an EE/DSP & software person) almost elicits a knee-jerk response in me -- and jerk might sometimes be operative :-).
There are *REAL* cases in nonlinear signal processing with digital processing where the effects of the 'stairstepping' (actually aliasing) are important. It is VERY EASY to unfortunately produce aliasing if one is designing digital compressors/limiters -- so there are some real issues there. When messing with normal linear signal processing -- then the rules tend to be simpler, and the starstepping/suprise aliasing/harshness issues aren't operative.
Think about this -- you have a fast attack compressor/limiter, and effectively when doing gain control -- the nice/clean signal is multiplied by the instantaneous gain control. That quick change in gain due to the fast attack will definitely produce evil sidebands. Those evil sidebands can easily wrap around Nyquist rate -- cause aliasing -- and create nasty/harsh/ugly sound. It is a VERY challening thing to do a really good fast attack compressor/limiter without IMD (it is an especially nasty form of IMD that is hard to measure, because constant signals don't really cause it.) A well design RMS compressor/expander does have a nice advantage -- the attack/decay tends to be slower than a fast attack compressor -- but these are just details.
I know that this board is really mostly interested in 'measured' results only -- no ad-hoc 'sounds good' are acceptable unless it is ABX. Frankly, I don't know how to quantize the real distortion resulting from a fast compressor/limiter, and doing a comparison by listening (ABX, if you want) is the only way that I know to decide between the 'kind of good' and 'really good' results.
The silly simple compressor that I have posted elsewhere has none of the heroic efforts needed to produce absolutely perfect (within limits of physics) that I'd need to put my name on a product. I give the toys away as demos and MAYBE useful for learning or casual use. Some day, after my noise reduction emulators are finished, might write a REALLY GOOD/truly low IMD compressor/expander. The stuff is a LOT more complicated to do correctly, than just to do it!!!
Again, sorry if I did appear to be strident. I sure wish I could explain things better -- the language needed to explain stuff is so tricky, even when I am not emotional, I cannot always communicate accurately...
John
It seems obvious to me that you’re not a jerk. You provide a wealth of helpful information. That’s a good way to respond. I’ve grown cynical and can’t help but just poke fun.
Want to mention the recent SoX at sourceforge does PCM->DSD conversion.
I can't see DSF or DSD mentioned in the current lost of formats supported - am I looking at the right page?
http://sox.sourceforge.net/Docs/Features
Thanks
Mitch
Sorry, i only saw commits at sourceforge and Mr. Rullgard contributing. His fork is here at github https://github.com/mansr
He is the person at CA who has answered and claryfied the DSD capable version is not at sourceforge and linked you the compiled version, mansr.
thanks - all understood
Mitch
@BrilliantBob Your brilliant side discussion has earned its own topic!
https://hydrogenaud.io/index.php/topic,117030.0.html