.. Better be a clear difference to justify the much larger file size. Some difference might be possible. Better? Maybe, but my guess is not definitively in any way.
43 responses Add your response
Are you able to articulate why this should sound better, at least in theory? I guess that there are potentially at least several ways one could upsample/upconvert, (a) as you have, by using pre-player software that permanently converts the files, (b) by using software in or associated with the program that plays the files, to upconvert it as it is played, or (c) through the hardware on a DAC that upconverts the bit stream once it gets there. Without getting into the debate regarding whether up-converting is worth while in the first place (of which I can see both sides in theory, yet experienced no appreciable benefit in my system), is there a reason why one method should be better than another? Don't purport to have an answer, just curious.
Background: for the past month or so, I was was using Bit Perfect to upsample during playback by powers of 2 -- ie., 44khz and 88khz files played at 176khz, 98khz files at 192khz, etc. Then I switched it all back to native resolution in integer mode. Think I prefer native, but I spotted very little difference, albeit sometimes more on some tracks than others. Given this option to upsample on the fly with only the click of a button, not sure why you'd choose to permanently alter your files by running them through an algorithm to extrapolate and add in hypothetical detail that was not recorded in the first place -- at which point you no longer have a native file, and the only way to return would presumably be by squashing it through yet another algorithm to "prune" out the large majority of the data to take it back down to 44khz. (Just thinking about that makes me a little squirrely, as I suspect that there is quite simply no going back.) Put differently, if you take a lossy mp3 and upconvert it to 16/44, you're not turning it into a CD-quality file, just a larger one. No? I'd love to hear why that's a great idea -- and perhaps it is in practice -- but having trouble getting my head around it in the abstract. Thanks.
I guess I confused you as much as I was confused. I wasn't trying at all to articulate why it would be better, but simply find out if it could be. My experience mirrors yours. Usually I prefer Native resolution playing files, but there were a few that I prefered up converted...as you, in multiples of 2. I did have some 16/44 that was a tad smoother and seemed to be a tad faster when upconverting to 24/176, but normally, I found native to be ever so slightly harmonically richer... That was upconverting a file through the DAC. Last night, I actually changed a file resolution. Well, my observation... Converting Flac to WAV, on normal to very good recordings, I found the WAV file to improve over Flac, a touch more focus and a touch tighter bass, no loss of detail. On High quality recordings???not so much, had a hard time telling them apart, sometimes felt like I could, but most likely me just thinking my old ears still have it, most likely no difference. Thats were it stands as of now, but I've only listened about an hour and a half. Again, this started because I respect Steve N's opinion and have seen in a couple of threads say that he recommended converting FLAC... I just happened across the upconversion. I'm going to try 24/44.1 next.
Haven't really compared FLAC or ALAC to AIFF or WAV but it makes sense to me that uncompressed would be theoretically capable of higher fidelity - no need for the CPU or software to convert on the fly. Storage space is extremely cheap, especially compared to overall hifi costs - so my attitude is why bother with compression at all.
Up conversion, however, makes no sense to me whatsoever. IMO, anything which alters the original signal is distortion.
Thanks, Tim. Was hopeful that perhaps you'd stumbled on something cool, but sounds like we're pretty much on the same page after all: curious, but the jury's still out.
Steve, have you found that permanently up-sampling files with a purpose-built program (ie, Izotope) is beneficial? As they say in the old Starship Troopers campy romp: Would you like to learn more? Yes. Many thanks.
"Steve, have you found that permanently up-sampling files with a purpose-built program (ie, Izotope) is beneficial?"
Depends on your DAC. With most DACs, this type of upsampling from 44.1 to 96 can be very beneficial. More detail and improved dynamics. Smoother vocals.
The problem I have found with FLAC is that on many systems it compresses the sound a bit. I believe that it is not changing the data when run statically, but dynamically something is happening to corrupt the data. With cheap disk prices and AIFF format, there is really no reason to play FLAC files. Convert them to AIFF or .wav.
"With most DACs, this type of upsampling from 44.1 to 96 can be very beneficial."
YEs, it can, but it all depends on what specifically is done during the upsampling and that is done correctly and that the playback system is sufficiently resolving and dynamic.
The upsampling process is essentially an opportunity for the designer to digitally remaster the raw data as desired. There could be more to it that strictly just up-sampling alone. There are many common enhancement algorithms that might be applied.
As usual, the devil is in the details in each case.
Uncompressed upsampled files will be much larger. BEsides disk storage, network or computer bus bandwidth limits are more likely to come into play. IT is not without some potential risk/downside as well.
So if you do it, make sure what you hear is worth it. Bigger is not necessarily always better when it comes to digital files.
Unlike analog audio, which records a smooth and continuous waveform in real time, digital audio must capture audio samples in a series of discrete steps numerous times per second. The more often you can sample, the smoother and more faithful the recording will be to the original analog sound. (Remember, sounds in the real world that our ears hear are analog.) A standard music CD has a sampling rate of 44.1 kHz, whereas high-res audio can go up to 96k or 192k. (Most tests have concluded that 192k is overkill without much practical benefit.) Aside from the occasional concert movie, almost all feature films and television shows are mixed at a rate of 48 kHz. This is unlikely to change in the foreseeable future due to a variety of logistical issues in the film production pipeline. For example, 96k files require twice as many mixing resources as 48k, which means that only half the number of channels on the console would be available. Not to mention that the average movie soundtrack is comprised of numerous audio elements recorded at a variety of quality levels. 48k is a common standard that works and doesnÂt seem ready to change.
I have Musica Pristina Virtuoso music server and here is my experience.
1 When I rip CD into WAV and FLAC (two separate files) and compare both - WAVE is dramatically better then FLAC which sounds lean, thin, 2D and castrated
2 When I take FLAC file I have downloaded from HDTracks.com or hidefenitiontapetransfer and convert it to WAV format I never experience sound degradation described in TAS recently. Converted WAV is better then FLAC but not in 100%. So I do convert as there is no downside (besides space which is cheap now) and potential improvments. Sound characteristics as I described above
3 When I upsample file, WAV or FLAC, (I use Musica Pristina software) then I hear tremendous improvment in highs: silkier, extended, air, instrument separation. Bass does not change (a bit tighter) but is not so promiment in comparison. Its natural, as we just improved highs and ear is not used to it! Later on, bass sound "normal"
I believe that upsampling results depends grealy on the DAC you use because if it does its own upsampling then done twice it could be worsen, in theory I believe. My DAC is Esoteric K-01 USB DAC portion of this SACD Player.
Done right - the SQ is better then majority of analog gears I auditioned in my life
Dob - It helps to understand why upsampling makes things sound smoother and silkier. Its not due to the upsampling itself, it because a different digital filtering is being auto-selected in your DAC. This higher frequency roll-off filter has less impact on the SQ.
If you can manually select the digital filter, then the advantage of upsampling the file goes away. You can select a high-frequency filter for all files for instance. That is what I do. This is even better than upsampling because you have the best of both worlds: the data is not modified and the digital filter is not impacting SQ.
"Dob - It helps to understand why upsampling makes things sound smoother and silkier. "
I do not beleieve its always the case and depends larfely on the given DAC (with its filters and other parameters)
I believe later on you also said that if X then Y and else its Z etc....
I think audiophile must trust their ears first of all. If I like the taste of the given soup I don;t have to learn all exact ingridiens and methods of cooking involved in its preparation - leaving it to the brilliant chefs as you are.
Always ewmjoy reading your comments. Thanks a lot
Any ideas about transcoding on the fly from FLAC to WAV?
Can this sound as natural as AIFF or WAV?
Has transcoding evolved to close the distance?
On reading your posts above, I think you were saying it is better to let the DAC do the up sampling rather then upconverting via IZotope?
Is there a way for Izotope to convert on the fly like transcoding?
Have you heard a difference converting a 44.1 file to 88.2 or 176.4 vs 96 or 192? Or does it make any difference?
Does AIFF keep all the artwork sorted as well as FLAC? Does AIFF sound the same as WAV?
Thanks in advance!
Dob - higher sample-rate files can be better for high-frequency transients, but upsampling 44.1 has its own added distortions I have found. If a DAC has a high-order digital filter, then leaving a 44.1 file at 44.1 is usually the best course. This result depends on your DAC of course. I have one of the few DACs that sounds great with 44.1. At shows, the attendees believe that I'm playing hi-res when its only 44.1 rips.
Any ideas about transcoding on the fly from FLAC to WAV?
Can this sound as natural as AIFF or WAV?
Has transcoding evolved to close the distance?"
It's the real-time uncompression of FLAC that causes bad audio quality. This would not change that.
"On reading your posts above, I think you were saying it is better to let the DAC do the up sampling rather then upconverting via IZotope?"
No, it is still better to use Izotope because the algorithms are better than any hardware resamplers, plus you get options to adjust.
"Is there a way for Izotope to convert on the fly like transcoding?"
"Have you heard a difference converting a 44.1 file to 88.2 or 176.4 vs 96 or 192? Or does it make any difference?"
Depends on the DAC. It is more a function of how good does the digital filter at 176 sound versus the 192 filter I think.
"Does AIFF keep all the artwork sorted as well as FLAC?"
"Does AIFF sound the same as WAV?"
I wish it did, unfortunately it sounds very diffuse and unfocused to me. I dont understand why because the only real difference is that the L-R data comes R-L instead....
Thanks for your thoughts. I've been reading that transcoding Flac to WAV with i7 multi core processors maybe the key to solving the problems like timing and latency add to the decoding process as the envelope surrounding the file is opened. There may eventually be (or maybe there are now) be some software written that assigns one core to the transcoding process. I know that in dbPoweramp has implemented such a scheme in their newest versions but I don't think it is used for transcoding.
In my latest tests, I can hear the difference between Flac and WAV files and know it is there. But if I just listen to the Flac and am not actively comparing it one doesn't really notice. Encoding files with the no compression or 0 compression is much closer to WAV.
I've also found it makes a difference in the processor in the PC doing the transcoding. For instance my WHS machine has an older AMD Athalon processor and while it is very quick to send out the files if I use it to run PS Audio's eLyric from the server with transcoding turned on, it clearly sounds worse than sending the files to my Windows laptop with an i5 processor running eLyric which performs the transcoding.
Using Fidelizer on both the WHS and laptop make a big difference as well.
SGR - I have found that the differences in FLAC ALAC and AIFF compared to .wav are I believe limited to software using the audio stacks, like USB and Firewire. Networked audio may not have these problems.
The sound quality differences are primarily with imaging, focus and soundstage width. Each of these has a slightly different effect. If you dont have a highly resolving system that is tweaked to achieve pinpoint imaging and wide soundstage, you may not hear any differences.
Maybe this was covered, but anybody know a good, reliable program that can do bulk conversion of a library of 1000's of .wav files to FLAC?
I would do this mainly for the metadata tagging advantages of FLAC versus .wav. My library is mostly .wav, just a few FLAC, but I would like to investigate moving more towards FLAC in the future.
jRiver or dbPoweramp can do the trick. I believe if you go to the FLAC website there is a free converter as well. There may be others available free. Do a google search.
My system does resolve the differences and they are clearly discernible. I have never used USB only Ethernet and I'm learning more tricks all the time about how to improve the sound quality. Using CAT 7 Ethernet cable, providing linear power supplies for routers and gigabyte switches, using Fidelizer, using PS Audio regenerators and filters for all front end components and any piece of computer related hardware, are a few of the almost insane lengths I've gone to. I suspect that many of these tweaks would help systems with USB interfaces as well.
It would be interesting to compare USB to Ethernet streaming to find if there were inherent advantages in sound quality from one system to the other.
In my system, comparing the same WAV to Flac files, I don't hear differences in imaging, but there is just a little more snap and live sound with WAV and maybe a little more blackness. However if one is not comparing them purposely, they are so close that one is really not aware that, "Oh, too bad that was a Flac file not a WAV."
I'm investigating types of computers, their components, and operating systems to find out how I can improve this aspect as well.
Thanks for your input.
Thanks for that info.
Which program is best from a usability perspective in order to just point to a source music library and convert to a target without any additional manual interaction? I"d like to be able to fire off the program just once and have it convert all my thousands of files in one reliable shot.
I get excellent results with wireless lan connection to music server. THis approach provides excellent electronic isolation between noisy computer gear and sensitive audio gear. Also excellent for physical isolation between the two as well.
I think I have heard some similar differences to what you describe between FLAC and .wav, enough to prevent me from going all FLAC so far, but have not done enough critical comparison to say for sure.
In my system, comparing the same WAV to Flac files, I don't hear differences in imaging, but there is just a little more snap and live sound with WAV and maybe a little more blackness. However if one is not comparing them purposely, they are so close that one is really not aware that, "Oh, too bad that was a Flac file not a WAV."If a WOW system's like Sgr's barely discerns between FLAC an WAV then, I, *a mere mortal* should be just fine with FLAC.
It's true, WAV is vastly superior to FLAC. Vastly. FLAC sounds like AM radio in comparison. Never mind the fact that they are 100% identical when played back, the exact same bit stream, identical to the last 0 and 1.
The 1's in the WAV file have more depth and clarity than the 1's in a FLAC file. And the 0's have a darker background, like a veil has been lifted. In fact, the 0's in the WAV file have more depth, they're crunchier, more robust, and holographic.
I think the 1's and 0's in a WAV file also contain more fiber, as my bowel movements are noticeably stronger after listening to WAV instead of FLAC.
BTW, I've also found enormous gains in changing the wall plates on my electric outlets to rhodium plated platinum with carbon fibre inlays, wrapping my interconnects in virgin albino crocodile wool, and de-magnetizing my wooden speaker stands.
I do hear a difference. WAV does sound better than Flac in a direct comparison. Especially if the FLAC file is heavily compressed. Flac files at zero compression sound much closer to WAV.
It has been suggested that it is decoding the envelope around the Flac file might cause software jitter that is sent to your DAC. Many have found that the newest i7 processors have much less trouble (if any at all) decoding Flac files This is all conjecture at this point as few know for sure.
Apparently there is more difference between the two depending on if you use Ethernet via DNLA vs USB. Most have reported less differences if the Ethernet via DNLA which is what I use. In most if these systems one can enable transcoding the Flac to WAV on the fly. I use this also.
So until you have tried streaming the files via both methods maybe you should be a little more open when deciding what I can hear.
I do hear the differences between Flac and WAV and they are much larger via USB than DNLA Ethernet. I have spent lots of money and improving my Ethernet system to minimize these differences.
In my system the difference though detectable have become smaller and smaller until I don't really care. Yes WAV still sounds a bit better as described above. I'm hoping that like others, upgrading my transcoding PC will finally make the differences between Flac and WAV undetectable.
At this point, WAV still sounds to just a little bit better. So close that I'd rather take a little hit in SQ rather than having lost tags and retagging again and again.
Hopefully you have learned some things about Flac and WAV playback systems. I'm always ready to learn a new trick. In audio, I have learned to never say never. Little things can make a big difference to the sound and all things have the potential to affect the sound of your system.
Would you believe treating the cd before it is ripped and using footers under your ripping drive during ripping could effect the SQ? Or that the quality and type of Ethernet cable can be heard? Many would say not, but my ears and system have been proven to be right by those who have kept an open mind and listened for themselves.
So long and farewell for now and someday I'll learn to keep my deaf ears to myself.
Sorry if I have offended you. You do things right. First you report what you hear and then you try to explain it - and its fine and how it suppose to be, first observatio(s) and then speculation/discussions and this is how we learn.
Simply, Mihaitaa with his "mere mortal" approach tempted me too much. As you can see many so called audiophies are trying to apply their ignorance of reality and "calculate" music instead of listinging to it. Just above Lupin the 3d described to us that zeros are zeros indeed and shared with us description of his thought process, or as he put it "bowel movement". He will never guess, that "1" is actually electrical pulse, say 1 v amplitude for illustation and could be a) mispaced in time due to the jitter and b) greatly distorted (say reduced to 0.499v) due to noisy power supply, non linear DC:DC convertors etc and be mischaracterized as a "zero"
To comment further, I am trying not to diminish huge difference betweet two formats but improve AQ and make it as close to analogue front end as possible. There are number of factors involved and I am sure I don;t know even half of them. Many extremely useful suggestions are from Musica Pristina people ("Spectron").
At any rate, my apologies again and all the best ! This hobby (addiction for me, actually) is fantastic for life!
"He will never guess, that "1" is actually electrical pulse, say 1 v amplitude for illustation and could be a) mispaced in time due to the jitter and b) greatly distorted (say reduced to 0.499v) due to noisy power supply, non linear DC:DC convertors etc and be mischaracterized as a "zero""
Well, this is certainly possible, although not very probable. Most of the effects in digital audio are not due to bit-errors, but rather from jitter.
This is why it is critically important to address the jitter of the source deviceand master clock as a higher priority than the format, computer, software or even the DAC quality. All of these things are second-order effects compared to jitter.
When FLAC and wav sound different, its probably more due to the different software in play for each and how that is written, designed, and performs more so than the format. Decompressing FLAC files will probably require more CPU processing, but should not be a problem if done right. Of course, things are not always done right, and many factors can come into play when playing digital music files, so differences in performance between the two in any particular case would not surprise me and reasons why may not be apparent.
FWIW, I think it's odd that different software players (that can be tested to show bitperfect output) can sound different playing back the same WAV or FLAC file. Heck, even different versions of the software can sound different playing back the same file.
It is thus not inconceivable that WAV and FLAC despite having bitperfect data can sound different. I remember that the Pure Music designer mentioned the need to minimize sudden/minute spikes in CPU load to improve performance and that was one of the goals of their software update. So it is not just a % of the CPU load that is averaged over time that we need to see but those sudden spikes. There was a talk at RMAF about 1-2 years back by an ESS engineer which talked about the need to look beyond steady states but also how the system reaches steady states (ie does it oscillate through large swings in values before reaching steady states). He found that large swings seemed to have a negative impact on the sound quality. I think there's a lot to the computer playback chain that we are only just beginning to understand.
For example, 96k files require twice as many mixing resources as 48k, which means that only half the number of channels on the console would be available.
Mr.Jazz1959,I do not know a lot about recording ,could you please explain to me how recording at 96k will use twice as many of available channels on the recording console as recording at 48k.Thank you
IT boils down to the software must be able able to run correctly and fast enough to maintain a in-memory cache of data that is available at the exact time needed for playback. Playback happens in real time, so data must be continuously streamed, made available and applied at precisely the correct time in order for things to sound best. Real time applications like computer audio up the ante in regards to what is needed for optimum performance.
Memory is shared and virtual on most general purpose computers. MAny programs may compete for available memory. WHen there is not sufficient physical (fast) memory available, virtual disk based (slow) memory is applied as a supplement to allow things to run though not as fast.
If data is not availble at the exact moment needed for playback, software has three choices:
1) pause or wait for memory to become available again. THis may result in an interruption or delay in playback until data is once again available
2) reduce the bit rate of the data stream. This would result in not all bits being used and would affect sound quality accordingly, although teh music may continue to play. A program/system designed for audiophiles would not chose this approach, but it might be applied otherwise for more casual listeners without concern.
3) some combination of 1) and 2)
With software/computer programs anything is possible and may well occur unless care is taken in design to avoid it. Specialized audio streaming devices like Squeezebox are essentially specialized and dedicated (not general purpose) computers designed to optimize performance. They take a lot of the mystery and variables that can affect the sound quality out of the equation.
SO in my opinion, the system used to stream and play FLAC or WAV files is a much bigger factor in regards to sound quality than the format itself, both of which are lossless and essentially equivalent in terms of information content. Its what content gets delivered and how well that matters most.
Unfortunately, digital playback mechanisms are not transparent to the user. There is nothing other than the hearing the resulting sound apparent to determine if if all this is occurring well or not. That's one advantage of vinyl. There is more there to see, feel and touch in addition to hear. It gives you something more physically tangible to sink your teeth into and perhaps adjust or tweak for better performance, if that is your thing. With digital audio, your listening fate is more largely determined by the equipment designers. Luckily, there are many good ones out there. There is an advantage to placing your fate in the hands of a trusted expert as well. Most will likely prefer that approach.
Mapman - with many people using aysnc USB the real time nature of the PC is minimal. All it really has to do is keep the buffer full and not get in the way of the aysnc USB requests. Filling the buffer is just not that hard. It is hard to understand how a computer with little else running other than the music player and with the CPU running at only a few percent of usage can have a significant effect on the timing of the aysnc USB over several minutes of music unless there is a serious flaw in the design of the player. There may be computer effects that influence wav and flac playback, but it seems that aysnc USB takes the real time aspect of the PC out of the equation. This is especially true for DACs that re-clock the data. Before async USB, the real time nature of the PC could definitely be an issue. But that seems minimal with async USB.
"All it really has to do is keep the buffer full and not get in the way of the aysnc USB requests."
That's pretty much always the case, isn't it?
Asynch USB might be faster/more efficient compared to other protocols maybe, but the data has to be cached upstream and readily available in any case in order to perform I would think. I'd have to read up on the aasynch/USB spec further, but not sure that alone is assured of solving the problem. Bottom line is the data has to be cached at a faster rate than it is needed to play, and readily available when needed in real time for conversion to analog.
Mapman - async USB puts the timing and request under the control of the external device, not under the control of the PC. The USB device is optimized for that and does not have other functions to distract it. So, all the PC and player have to do is keep the buffer full, which it should be able to do without much problem. Other than that, the timing is controlled by the external device.
Here is an explanation of how async USB works, from the guy who introduced it.
Hi Dob, all I wanted to say is that Sgr's system is far superior to mine (mere mortal) and yet he stated that the differences between FLAC and WAV he heard were relatively subtle.
I simply inferred that a lesser system, like mine would probably not reveal them at all.
If this called for your comment, so be it, I'll turn a "deaf ear" to it.
The first system I noticed the difference between AIFF and ALAC (compressed lossless) was with an iPhone 3GS docked to my car's Alpine Head Unit. Not even a very expensive set up. I used the Alpine's integrated amps and a set of Rainbow component speakers.
I had done that experiment after how I noticed a CD played on an Ayon CD5S sounded a lot more dynamic and punchy vs the same CD played on my iPhone docked to a Wadia iTransport and my friend asked to change it to WAV or AIFF. I thought he was pulling my leg but did it just as a test.
After the experience with the car system, I brought the phone back to the showroom where I had tested the ALAC vs CD earlier and this time, the gap was much smaller.
I would like to illuminate my "subtle" reference between Flac and WAV files. I have heard some systems where the differences were not subtle between the files. WAV was much better no contest. In fact when I first started with PC Audio, using USB and then moving to the Bridge and PWD II the differences were not subtle. It is only now with CAT 7 Ethernet wiring the system, a faster computer to transcode, better power supplies for routers and switches, and several other small changes have the differences between the two formats become less discernible. But I've spent a lot of time and money to do this. I did not set out to do this, it was not a goal, it just happened as a result of my system getting better.
I would love to close the gap, and believe upgrading to an i7 processor, a solid state drive and a few other changes might make the point mute.
There is definitely an advantage in setting your Flac compression to the least amount possible.
Sgr - there have been quite a few anocdotal reports that say that networked audio is less likely to exhibit these format differences in SQ. I believe is is a funciton of the audio stack, which is bypassed when using most networked audio streaming. It is mostly audible when using PCI cards, USB or Firewire.
The greatest and most consistent difference between FLAC and WAV when you rip your CD in both formats and then compare.
When you convert FLAC (ripped or downloaded) then "new" WAV file sound not worse then FLAC, practically 100% in my experience. In addition, many new "WAV" files sounds better (ranging from slight to huge difference) then original FLAC files. I do this conversion now automatically as I know SQ will not worsen but may (may) improve