1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

Lossless vs mp3 ABX results. (Among other ABX's)

Discussion in 'Sound Science' started by deadlylover, Sep 30, 2011.
1 2 3 4 5 6 7
  1. firev1
    Actually if you are interested enough deadlylover you can listen to Suara, she is only J-pop artist who releases in SACD or 24/96 for that matter but alas I don't think either of us has a way to rip SACD tracks. 
  2. kiteki

    Yes I have some native 96kHz, you can find native 96kHz (and other formats) available for free download at this link www.2l.no/hires

    Okay, I'll try downsampling to 48kHz next time to see if there's a difference... my hunch is foobar is confused by the massive amounts of 'silence' on the 24/192 files, above 20kHz.


    No, I was using m2tech hiface 24/192 over USB, to Sabre ESS ES9023 DAC.
    That's the reason why I started this testing, no point having 24/192 capability for nothing [​IMG]
    If we hypothetically say the difference is audible, my take on it is 16/44 versus 24/192 is like excellence versus excellence+1, or a great cake versus a great cake with icing on top.
    I'm really not sure if that tiny bit of icing on the cake is worth a five times larger sized box for the cake.
    However, if I can upsample the icing onto the cake while I'm eating it and keep it in a normal sized box, then that's cool with me.
  3. goodvibes
    Whether you prefer upsampling or not will have a lot to do with the upsampler and dac. If you have a really good oversampling dac with bit for bit reclocking and excellent filters, you should prefer files at their native rate as the higher frequency required for better filtering is handled by the oversampling. Tends to cost a bit to get the symetrical reclocking right. For me, I get sonic benefits with this type of dac with up to 24/96 native and not much after.
    Most less expensive dacs use an asrc to upsample/remove jitter so what happens before it can have variable results.
  4. Head Injury

    Sonic benefits that hold up under blind tests, right?
  5. goodvibes


    Easy with revealing enough kit as in not difficult. If you're ever in Chicago, you can test me.[​IMG] Reference is original 24/96 masters to down converted 16/48 in Wavelab pro with custom dither or analog masters directly transferred to both 24 and 16 bit respectively on a Nagra VI with custon linear supply as it's the first dig recording setup that we've found compares to the best analog.
  6. Head Injury

    Just download Foobar and the ABX plugin. I don't have to be there.
  7. deadlylover

    goodvibes, give it a whack with the ABX plugin test thing, and watch out for any funky volume difference stuff that might happen just like kiteki found.
    No need to feel bad/embarrassed for failing, because you're 'supposed' to fail anyway, and any positive result will be heaps interesting/helpful. I've tried high-res ABX and I just failed miserably left right and center. [​IMG]
  8. kiteki
    This afternoon I woke up from a slumber, in my dark quiet room, and started listening to some music.
    I was listening for a while, and then thought "this sounds too good, what's wrong?"
    My current 'listening studio' is simplistic, here's a photo of it:
    Music on laptop -> USB -> m2tech Hiface 24/192 -> Sabre ES9023 -> LME49610 -> RCA -> IEM/HP of choice.
    Music on laptop -> USB -> m2tech Hiface 24/192 -> Sabre ES9023 -> LME49610 -> RCA -> stereo receiver -> speakers.
    Sure enough, when something sounds "too good" or "unusual" I check the DSP settings in foobar:
    Right so SoX was turned on, makes sense.
    I turn it on and off on the fly, while listening to the track, with these arrows:
    I think the difference is quite faint, but I like the way the upsampling sounds, and I think it shouldn't be very difficult to ABX it,
    so I proceed to right click the track, and click on convert, here we have the conversion settings:
    and here:
    This time I selected upsampling x4 (176 kHz) instead of 192 kHz, however as usual, the replaygain scan is all stuffed up, which invalidates the testing:
    There's more issues I have with the testing, I have no way of knowing if the ABX comparator is successfully playing the files the same way as they are played to me in foobar, and after the testing the ABX comparator doesn't tell me which is which (A/B are constant, X/Y differ), so it's only purpose is localising a difference, when executed correctly, yet it doesn't show if the listener knew which difference was which.
    I mean hypothetically, if someone uses ASIO in the foobar program, how do we know the ABX comparator is using ASIO as well, with the same settings? It just takes a chunk of data and plays it, and checks if you can hear the difference, which is useful, but I think it could be more advanced.
    Here is what a downsample looks like, by the way, so as you can see, data is cut off and lost in the downsample from 192kHz to 44.1kHz, (user "gregorio" kept on insisting to me that no sound exists above 20kHz).
    IMHO gregorio was a good example of a "cynical scientist", where instead of practicing impartiality, they chose the side with the least snake-oil, and then defend it violently.
    I kept on telling him I agree with him that no one can hear above 20kHz, and he kept on saying the fact we can't hear above 20kHz is "simple proof", and implying that people that think they can hear above 20kHz are "idiots".
    Personally, I think there's a possibility that sounds above 20kHz may have some effect on the audible, sub-20kHz spectrum, via interaction, overtones, filtering, etc.
    Either way, as is demonstrated in history over and over, and by looking at "scentific studies" that are completely devoid of intution, discussing facts on paper sometimes leads nowhere, or to error, and as you can see in politics, humans are extremely partial to their personal circumstances.
    I think the best thing to do is to find out for yourself.
    I think it's nice that the developers of audio equipment keep on pushing the envelope, yes sometimes it might look like it's being pushed too far, yet along that path insightful discoveries often come along, so why not?
    Anyway, my experience so far has made me lean towards the conviction that upsampling is more important than bit perfect.
    The reasons are simple, I think the difference between 44.1 and 192kHz is there, yet faint, the difference in size of the files is colossal, no portable devices AFAIK support 24/192 playback, and you'd fill the entire device too quickly, at least now in 2011 when we are limited to 64GB DAP's which are not sufficient, yet perhaps in 2016 with 640GB DAP's, then native 24/192 would be ok, (purely talking in a sense of space here).
    In other words, I don't want native 24bit/192kHz files on my computer, when upsampling sounds just as good if not better, from what I can tell?
    The next reason, how much data is in native 24/192? Hardly any at all... so if upsampling sounds almost just as good, then what's the point in native?
    I'd rather watch youtube and DVD's all upsampled into 24/192, than watch youtube... in native bit perfect, right?!
    So yeah, what my testing and personal listening experience has led me to, is that I quite enjoy the faint effects of upsampling, via the SoX plugin in foobar, so I definitely wouldn't mind having a USB DAC/Amp that upsampled everything I fed it, instead of the one I currently have which only supports native.
    The next point I have though, is the quality of the upsampling, which is a very complicated, instrinsic process, right?, so if a DAC only advertises "upsampling" and that's it, how do I know what the quality of the upsampling, is like?
    Here's an example:
    Here I'm upsampling 22.05kHz (which sounds horrible) to 64kHz, as you can see the information in 22.05kHz is being heavily cut off, due to the flatness at the top, and yet the 64kHz upsampling still has that flatness at the top, shouldn't it predict what the information will look like above 11kHz and create some fake information up there?
    Finally, here is that Rihanna test again:
    In this test I have not ticked the box, since when I perform this test and I have ticked the box, I learnt how to listen for slight volume differences, and after a while it became really easy.
    Then I unticked the box and started failing a lot, but I'm getting better at it now... and I am not listening for volume differences anymore, I am listening for air between notes, effortlessness, fine-grain/coarseness (for example in vocals), and difference in emotion. 
  9. Head Injury

    That's not what upsampling does. You can never add data back into a file once it's lost. All upsampling does is resample the data it's given to a higher sampling rate, and fills in missing data with empty bits.
    If upsampling sounds different (and you can ABX it) then your system is doing something different with the sampling rates. It either handles the higher sampling rates better or worse. Probably worse in the case of 192kHz.
    If information above 20kHz affects the sound waves of audible frequencies from a headphone, that's not a good thing. Any interactions between the frequencies that are supposed to be there are already present in the audible band of the recording. Any additional interactions at playback are distortions and colorations, aren't present on the recording and wouldn't be heard in real life. They should interact once, not twice. So wouldn't that mean 44.1kHz is objectively more accurate?
  10. deadlylover

    I checked the waveforms a while ago when using the SoX resampler, and it does make changes to the waveform, rather than just filling in the parts with empty bits.
    It seemed to smooth things out, if that makes any sense. [​IMG]
  11. Head Injury

    Digital audio isn't my forte (none of this is [​IMG]) but that doesn't sound like a good thing. Resampling up or down? Were you using the recommended settings from here? Got any pictures? [​IMG]
  12. kiteki

    Please explain what's happening when I upsample 22.05kHz to 44.1kHz, which results in better sound...

    There is a possibility that my system makes 192kHz sound different, yes. It could be the SoX software itself, or it could be part of my system doing something different when it receives a higher khz signal, such as the DAC chip, or it could be that the system is flawless and 192kHz just sounds different. Anyway, I listed all the components involved.

    If sounds above 20kHz created artificial noise in the audible spectrum of a headphone, that would be bad, yes, and then sending a 20Hz-20kHz signal to said headphone would result in more accurate sound, yes. I don't think that's the case here, though.
    Reality is 0Hz-1millionkHz, our task is to record reality, by selecting to only want that 20Hz-20kHz section of reality, we have to use complicated filters to cancel out the higher frequencies.
    Apart from less requirement in filtering, it also seems like recording at higher frequencies results in cleaner impulse response signals, and less noise.
    So, on the studio side of things, it seems like recordings are done at 24bit/192khz, Afaik... and then they downsample it to 16bit/44.1.  They are however, sitting on a vault of information of 24bit/192kHz, that they could start selling to the public, the question is, is there any point in us listening to that 24bit/192kHz material? That's what I'd like to know.
    At the moment, I think I'm happiest with the the studio doing high-res recordings in DSD or DXD with expensive microphones and software, and then downsampling it to 16/44, and then I upsampling it back up to 24/192, that sounds like the most viable option to me, from my current listening experience and looking at the file sizes.
  13. Willakan
    If you are hellbent on upsampling, do so to 24/96. As documented by Lavry and Benchmark, due to the reduced oversampling ratios the result from playback of a 24/96 file is measurably better than the 24/192 equivalent. In other words, from the same source a 24/96 file will always be superior to its 24/192 counterpart.
  14. kiteki
    So you recommend recording in 24/192, downsampling to 16/44, and then the listener upsampling to 24/96... is that correct?
    Asus recently released a DAC/Amp that upsamples to 384kHz, just FYI http://www.asus.com/Multimedia/Audio_Cards/Xonar_Essence_One/
    I started reading the Lavry article, aready at paragraph 5 it says "In fact all the objections to audio sampling at 44.1kHz, (including pre ringing of an FIR filter) are long gone by increasing the sample rate to 60kHz".
    See page 2 here, comparison of 48 and 96 - http://www.cirrus.com/en/pubs/whitePaper/DS668WP1.pdf
    It seems that cirrus logic's take on it is that the audible difference between 44.1 and higher sampling rates is inter-channel phase distortion and pre-echo.
    They also note that it apparently used to say on Sony's website that the audible improvements of SACD over PCM are not due to the extended frequency range.
    Too bad gregorio isn't around anymore to join in with pictures of his studio.
1 2 3 4 5 6 7

Share This Page