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

Upsampling a la Chord Blu Scaler and Sony DSEE HX - Bandwidth Extrapolation?

Discussion in 'Sound Science' started by ruthieandjohn, Jun 4, 2018.
2 3 4
  1. ruthieandjohn
    I have read many glowing accounts of the Chord Blu MK 2 Scaler and the Sony DSEE HX, both of which seem to me to claim to "restore" information missing from standard CD recordings at 16 bits, 44.1 k samples/sec. The mental picture, as drawn by Sony, looks like this:

    DSEE HX.jpg

    The enhancement (left) extends the bandwidth of the signal far beyond the yellow segment that truncates at mid frequency (right, the yellow portion).


    Does this really work? This is taking information not in the signal, but making some guess, I guess, given the spectrum up to 20 KHz, how it might extend to the higher reaches addressed by high-resolution, higher-sampling rate, audio.

    If it DOES work, are there products that do this off line? The Chord Scaler costs upward of $10K to do this in real time on a standard CD, keeping up with the signal from the disc. It would seem that one could also off-line process the CD signal to end up with this upscaled, bandwidth-extrapolated, "better" signal.

    By the way, the Sony DSEE-HX is described as only applying to recordings at less-than-CD rates, e.g. MP3 128kb/sec perhaps, as a way of restoring some of the "missing" information, but when I listen on my Sony NW-WM1A with and without this feature turned on, I only hear difference, not improvement.
  2. bigshot
    When I first read what you posted, I assumed that they were upscaling CD sound to high data rate sound. Not much purpose to do that. But this is for low bitrate MP3s? I don't see much purpose to that either. Even if you simulate the lopped off upper frequencies, it still won't help with the real problem with low bitrate lossy- artifacting. This seems like a way of polishing turds. I don't know why you wouldn't just encode at a transparent data rate for your lossy codec and not need this at all.
    Last edited: Jun 4, 2018
  3. castleofargh Contributor
    how did you comes up with the idea that this DSP from Sony had anything to do with Chord's crazy high number of taps? ok they upsample/oversample, but so do most DACs around the world. everything else is different.

    sony's DSEE is about taking moderately garbage signal and making up some more garbage that kind of looks like it could have been part of the original signal(which of course is a vague guess because you don't pop extra data out of a hat). that is done based on some algorithm that doesn't actually increase fidelity. sometimes the result does sound a little different, and I guess that's when a listener will have to decide "do I want almost half my battery life gone(true story) because of some DSP that's supposed to improve the sound in a way that doesn't really make sense?" and the answer is probably no for most people. those who like the result can use it, because even if it does nothing for fidelity, as any DSP, someone might like it subjectively. and I guess that's the cool thing about choice.

    now Chord products don't try to make up crap out of nothing, they just try to get the most fidelity out of the encoded signal. and they seem to do that just fine. in a somewhat overkill use of taps, but aside from that, it doesn't pretend to turn poop into gold, only signal into something very close to what it was(like any DAC usually does pretty well without costing 10+K$).
  4. ruthieandjohn
    I got the idea perhaps by misunderstanding what upsampling DACs do, and hence equating them with what Sony DSEE HX does... bandwidth extrapolation.

    Any instruction or tutorial on the purpose of upsampling in a DAC, e.g. what Chord Blu MK2 Scaler does, would be most helpful. For a signal that comes off of a CD at 16 bit, 44.1 samples/sec per channel, I don't understand how resampling it faster has any purpose. I definitely have found the Chord Hugo and Hugo 2 to be superb, I suspect because of the large number of taps of the filter (and the Blu has even more), but I had thought that just provided more rapid brick wall cutoff and the opportunity for some spectral shaping, rather than in any way changing the analog signal that is rendered from the CD bits.

    Last edited: Jun 4, 2018
  5. castleofargh Contributor
    for questions specific to the chord product, you'll surely find more information and help in the dedicated topic https://www.head-fi.org/threads/chord-electronics-blu-mk-2-the-official-thread.831343/ or by asking a direct question to the designer who's active on Headfi.

    about why almost all DACs increase the sampling rate of the signal has been discussed so many times, google is your friend on this. be it for technical choices, or even for hypotheses about why certain choices supposedly sound better to some(for the most part those are irrelevant, but passionate people will be opinionated). the main idea is simply that it makes things easier for everybody. be it band limiting or noise shaping like you mentioned.
    all DACs are converters, so they will change what they got in some ways, but almost none are trying to actually change the signal. only a handful of weirdos will try to design a DAC without fidelity as main target. when they use different methods, it's usually to follow a given philosophy about what they think is the most important variable to preserve. let's say some designer is obsessed with timing and phase, he can center his design around that at the cost of other variables of course, because we only just have the resolution of a CD to play with. in the end, while something like a CD is indeed strictly limited in the data it contains, the signal is always amplitude as a function of time, and we know that relation in sounds is defined as sums of sine waves. so a lot can be done to move the data around without really losing much of it. DSD being an extreme example of that, a 1bit signal that still ends up providing around 24bit of dynamic thanks to aggressive noise shaping and filtering of all the garbage that's been pushed in the ultrasounds. if such an extreme example of coding the digital signal can still end up equivalent to highres PCM convertion once in the analog domain, of course we can resample and do a all lot of things while still ending very close to the 16/44 signal of the CD being ideally converted to analog. they're the many ways to skin a cat.
  6. bigshot
    Upsampling is pretty much standard now, right?
  7. gregorio
    Colour me cynical, but the Sony unit appears to be a perfect one stop tool to take countless CDs masters and convert them so they can be sold as Hi-res masters and it would be essentially impossible to tell that this "Hi-res" version is not an original Hi-res master. We can in theory already do this and have been able to for years but it's quite a laborious task and I don't know how often, if at all, it's actually been done.

  8. bigshot
    Gregorio, is this similar to the mastering technique where they take the bass and pitch it down an octave to make it sound fuller? I have some Frank Sinatra CDs that have acoustic basses that sound just like electronic ones, because they've done this.
  9. castleofargh Contributor
    sony's DSEE takes a 44 or 48khz file and based on the pattern of music it identifies a genre and makes up crap in the ultrasounds that would have been there in a similar music, genre/tempo etc. I don't know exactly how they do it but sony DAPs have a "sense me" function doing exactly that and identifying musics by speed, genres etc based on the signal instead of tags, so it wouldn't be that surprising if they used that information for they patching signal with ultrasounds that were never here option. at one point they seemed to infer that it was also doing that for the signal in the audible range, but they dropped the graph and marketing in the later devices, so IDK. seems completely stupid, but perhaps they were adding random noise to decorrelate stuff in the lower bits and push the signal to 24bit? that part I really don't know. maybe that never existed at all.

    it's like MQA's wet dream, all the BS about timing, transient, filter way up even for the signal itself, etc, without having to bother actually having the data in the container. ^_^
    what's funny is that my DAP has that, and also has the opposite where it turns a highres file into 16/44(or 48?) so that we can use DSPs that would otherwise probably be too CPU heavy for the little guy(I'm guessing). or maybe they just never bothered making a version for highres since it wasn't a thing for their DAPs a few years before? in anycase, it's there on my DAP ^_^. how cool is that, wasting all the space on my µSD just to do real time downsampling. genius option for sure.
  10. gregorio
    One way you could do it would be to do effectively what you state but with the treble, pitch it up an octave, lower it by say 12dB or so and mix it with the original. That would fool most people looking at a spectogram but in places, under close examination, you might be able to tell that the harmonic series would be incorrect. A more fool proof method would be to use an Aural Exciter on the treble (which just adds harmonics) and maybe add a little white noise. The aural exciter part is probably not a million miles from what the Sony unit is actually doing.

    I don't think that would work, there's already random noise (dither) on the CD, so there's nothing left to decorrelate which hasn't already been decorrelated, all you be doing is adding random noise to the lower bits and that may of course be what they're doing. If it were me and I wanted to fake a CD to look like a 24bit file, I'd make a copy of the track, lower it by 96dB or so, then mix it back in with the original with some very low level pink noise as well and then just truncate back to 24bit. That would look pretty convincing in a spectogram and sound pretty convincing if you find a quiet bit and whack the level up!

    Last edited: Jun 5, 2018
  11. castleofargh Contributor
    as I said I'm totally guessing about that part. the original purpose of DSEE was for it to be applied to low bitrate lossy stuff to make them "better". while vaguely testing stuff I just noticed that it still affected the signal on 16/44 flac on my DAP, but I really didn't care about that DSP so I didn't look further into it.
  12. ruthieandjohn
    tansand likes this.
  13. Redcarmoose
    I don’t know why everyone only has stated the DSEE HX post processing is only on the Sony DAPs? But it may be just that the DAPs are more popular than the full-size Sony TA-ZH1ES.

    Back in November I started with the 1Z DAP finding the upscale process to be somewhat of a gimmick? It looked good on paper but going and changing the settings never made any improvements to my lower quality files.

    Later I picked up the Sony TA-ZH1ES and Z1R headphones. After folks telling me I was waisting my amp by not using the feature I started to play around with the settings.

    Then I started to hear a major improvement in how I was hearing my files. This is funny as I’m really into 16bit 44.1kHz FLACs and regular Redbook CDs played off a transport. I do at times hear slight improvements with DSD files, but in my humble opinion it is mainly the responsibility of the master and if that master was done right in 16/44...............all is well.

    The full size amp does the process either to a different level or the detail pull-out is of such a better quality that I hear an upscale on everything from 320kbps to 16/44 and even seem to hear the process improve with HD tracks though the difference is somewhat less. For some the thought of touching high resolution and jacking them up is something like mixing nice Scotch with Coke........a horrible degenerate regression!

    I had read that the full-size amp from Sony is doing the process slightly different and it makes sense as there is simply more room for the electronics, and the chips and circuits........thus.......not miniaturized for the DAP. Though adding to the surreal confusion..........I now hear DSEE HX as an everyday improvement with low quality files daily on the Sony DAPs?

    Who knows.........but I try and keep an open mind to maybe there being something to this digital tomfoolery? It’s not a giant change......but it is there none the less..............enough to make someone like myself a believer in the technology.

    I’m a strict believer leaving it on 24/7. If you read about users many leave it on except when playing HD or DSD tracks. And simply.......that’s what it is designed for.

    Maybe it’s placebo or maybe I didn’t know what to listen for in the beginning. But it’s a nice feature if your stuck with 320kbps and want a little lushness for free!
    Last edited: Jun 6, 2018
    skwoodwiva, ruthieandjohn and Blommen like this.
  14. castleofargh Contributor
    found this from a post of mine in 2014(wow I spent 4 years with the same DAP, that's at least love²).
    top2 16/44 flac recorded in with the crappiest computer line-in I ever owned. bottom 2, the same with DSEE active on my DAP.
    looking at that today I feel like Sony should have used all the BS marketing from MQA for DSEE and be golden. all the stuff about extra time resolution, less ringing and softer filter, without even having to to actually stock any ultrasonic content. so of course it would be even more ideal for streaming ^_^.

    one thing that might count or not, Sony has a habit of changing stuff while keeping the same DSP name, so maybe todays DSEE is different, maybe on different gears it's different, IDK, but it wouldn't be the first time.

    as I said almost all DACs resample. if only because most modern DACs are delta sigma DACs and to work with pulse modulated signal, you need to say "up" or "down" VERY often. only a few rebels want to argue that the pros don't vastly overcome the cons.
  15. Redcarmoose
    The fact that we are seeing Field Programable Gate Arrays allowing real-time processing of signal much like a DSP. This technology made big steps in the 1990s but now doing DSEE HX processing in a way only dreamed about 15 years ago.

    Still it is DSP style effects on another level. It emulates the richness often accompanied by higher bit-rate files. A subtle almost echo style of effect but a smart effect which gives the first impression of fullness and harmonic complexity. It’s adding a softness like what you can get with some oversampling DACs.... but it’s also a little more. The strange question is that the process has actually less to add with DSD and HD files. Almost like it studies the low bit-rate files and adds what’s needed but gets confused to add much to DSD and HD. The effect is still there but it’s less dramatic. It’s like it sees what’s already there and steps back a little.

    Correction.......no post processing is done with DSD files as they automatically bypass the effect.
    Last edited: Jun 7, 2018
2 3 4

Share This Page