With my 70yo ears and my ESL63 panels, I still can perceive that Hires 24/96 or 24/88 files sound better than 16/44.
My hypothesis is that, at least on recent classical music, the downsampling to redbook is badly done, be it on bandpass AND dynamics.
Is it deliberate or not, that is the question…
But, as you write, more important are the sound take and the mixing/mastering.
With my 70yo ears and my ESL63 panels, I still can perceive that Hires 24/96 or 24/88 files sound better than 16/44.
Here’s the article I wrote on the subject:
Non-Italian readers may perhaps use Google translate.
I would like to thank all of you, your comments have been very helpful to me. In particular I would like to thank @PhilDaintree for introducing me to Max Waldrep and his work, which was fundamental to my article.
Thanks for your interest, let me know your feedback!
Just to confirm my understanding of Mark Waldrep did:
- He took a wide range of his 24/96 masters
- Downsampled each to 16/44.1. I believe the top-end filter was at 20Khz
- Upsampled each again to 24/96 and level adjusted to make sure they had the same volume levels as the original hi-res tracks
- Then the blind test was performed between two 24/96 tracks, one original, one the down/upsampled version. You would not know which is which.
The spectrum analysis of the two files shows that the upsampled tracks max out at the CD frequency ranges. When differentially compared, you can see that the original hires version has the additional ultrasonic range > 20Khz
The reason for performing the comparison between two hires buckets is that if you switch between a 16/44.1 and 24/96, there will be natural volume differences because of the sample sizes and that will taint you comparison to favour one over the other. It also has the effect of displaying this on a lot of equipment and that would largely invalidate the blind aspect of the testing.
His results and process were presented to the AES in a paper. So this entire thing had the right academic vetting.
I agree with you.
What I suggested is that if he down(up)sampled HIS files wit a rigorous method, it may not be the case in the commercial/marketing reality : Hires (or so-called) files cost more…
“Dedicirt wird dieses Werk Niemandem außer jenen, die Gefallen daran finden” (Franz Schubert, 1797-1828).
HiRes is pure marketing. I don´t mean to offend anybody but this is “wanking with numbers”. The higher the better . I would even go so far that nobody is able to spot the difference between 320kBit ogg (i.e. Spotify premium) and flac of any bitrate with 99,9% of the presented material in a double blind test. Here´s why:
The problem with human hearing is that we don´t hear with our ears but with our brain. And that is in most cases knowing what is playing, seeing the glossy expensive gear vs. some well constructed and assembled DIY speakers which measure objectively better but unfavourably lack a walnut finish and golden spikes.
People who clearly express their dislike for lossy music formats simply don´t understand psychoacoustics, which is mostly the principle of auditory masking (https://de.wikipedia.org/wiki/Maskierungseffekt, in German, but there is an English version, if you prefer) which is inherent to every human being.
Of course there are dramatic presentations on the net that there IS a difference (obviously, where should the data go?) showing “diff”-Files which are quite impressive at a first glance. “Oh my gosh, I´m missing all that stuff?”. Yes you are, but your brain would have sorted it out anyway.
This is not an exact analogy but more apparent than auditory masking: Imagine coming from outside into a dark room on a bright sunny day . You almost don´t see anything until you eyes get used to that. That is (sort of) why you are not able to hear silent - in comparison - content while the loud content dominates. Of course you can clearly hear the “diff” content alone, but only when the rest is not present.
So the theory behind that is: When you KNOW, what you are listening to, you have a clear preference for “more horsepower” (which is HiRes in our case), because it must be better. You paid more money for it in the most cases anyway. But that is subjective. If you are lucky with that, I don´t mind. Most products of the so called “audiophile” industry would not sell, if our brain wouldn´t play that trick on us.
OT: The burning in / breaking in process is the same. The only thing that breaks / burns in, is our brain. Simply speaking: we are getting used to the sound after some time. There is no part in the hole audio chain which changes its behaviour over time (but over temperature ).
44kHz/16bit Is more than sufficient for playback. 96kHz/24bit is needed while recording because it gives you more headroom for mixing. But yes HiRes files often sound better, that is because the recording and mixing was done with a great deal of care and knowlegde, and since: why not selling it as “HiRes”? It´s there anyway, it meets the demands of the customer and finally you can make more money.
Modern lossy encodings are not the devil, because they respect the science (physics, biology, psychology). The higher the bitrate, the more far the “lost content” is away from the audible threshold. That was not the case in the early days of mp3, which was sometimes poorly encoded with audible artefact, so please don´t refer to that age.
My background: I´m an acoustic engineer in automotive with a passion for music and good recordings. And a clear dislike of anything “voodoo”-ish some clever marketing is trying to sell to us for lots of money. I build speakers for over 20 years. Most of the time I listen to spotify and I´m happy with that. The reason I sometimes listen to CD is: The streaming provider of my choice tends to have only remastered versions of old material. Unfortunately this is mostly done with reduced dynamics as the modern customership, mostly listening to music on smartphones, seems to like that more.
I also listen to vinyl but, since this is objectively inferior (dynamics, SNR), because of the special sound and nostalgia. Like some people (not me) prefer tube amps. There is nothing wrong with that as long as you know, what their physical shortcomings are and why that is so pleasing to some ears (brains! ).
Cheers from Bavaria
EDIT: typos removed and refined some expressions since English isn´t my native language
EDIT2: After refreshing my knowlegde, auditory masking is not only done in the brain but also an effect of the mechanics in the ear and the neural signals, sorry for omitting that in the first place.
Thanks for your comment @arndtinger , the psychoacustic aspect is in fact very interesting (and crucial).
I also absolutely do not demonize mp3s: modern algorithms are capable of doing a great job (especially with quality rips such as 320kbps). The rest is done by our brain.
Tube amps are worst than transistor amps from technical point of view, but they are still popular because the second harmonic distortion they introduce can be pleasant for the human hears
That’s exactly what I meant. In addition to that, because of their high output impedance they tend to follow the speaker impedance with frequency response.
I personally prefer ClassD amplifiers, the “tube-like” distortion can be added with DSP.
Next speaker project, which I have just begun, will include Hypex fusion plate amps. Will be my first experience with Class-D, looking forward to. At the moment I have Class AB im my living room (Dayens) and in my bedroom (Vista Audio) which are both from Serbia and offer excellent value for the money.
But BTT: (or better to psychoacoustics) I myself being a victim of the “higher → better” thinking am currently ripping CD´s as long as this: Volumio running CD drive at noisy high speed - #7 by arndtinger isn´t solved sufficiently.
Despite of me claiming, that 320kBit spotify is already sufficient for a “high end” experience, this is not only including material, which is not (yet?) available on spotify (e.g. Nationalgalerie, which i was very fond of in that time) but also some CD´s which are in my opinion very well recorded or have the above mentioned remaster problem. It simply gives me a better feeling to know that I have the seemingly better quality at hand, because only the living room gear sports a blue-ray player (Cambridge CXU). Bedroom and the future project (will) only have RPi´s with Volumio.
You see: Even being conscious of the fact that there can be no audible difference (science says so), subconsciousness plays tricks on us. So don´t rely on subjective opinions, double blind tests are the only way to find the truth.
EDIT:PS: And of course Measurements! Floyd Toole (mentioned in Emanuele´s Article) has done great work to establish Spinorama as a gold standard for Speakers, and SINAD is well established for electronic gear. Visit Audio Scene Review, a valuable gold mine for reliable Data.
(Though a SINAD of > 100dB is well beyond the audible threshold, it gives you a good clue, how well the gear is / was constructed)
There is a second clue here: 16bit gives you dynamics of 96dB (mentioned above). You ever heard at 96dB for a longer time? Then I bet you don’t have neighbors . → Sufficient dynamics for listening @ home.
your last remark, you made a strange conclusion. Dynamic range is not playing music at 96dB, it’s the value between the lowest and highest volume. So if your amplifier is set to 0dB, 16 bits can reveal sounds between -96dB and 0 dB. Not many people can hear sounds < -90dB
Sorry for being less precise here: Yes you are right. 16 bit → 96dB is a range. So if the loudest parts of your content plays at 96dB volume (which is in my opinion very loud for listening at home) the low volume parts (e.g. the silence between tracks) is 0dB volume which is the human hearing threshold. So there will be “absolute silence” or a “perfect black background” as mentioned above. Hope this clarified things.
I didn´t understand, what you mean by your last sentence:
missed then minus sign < -90db
I have the Zwicker (and successors) book, but I confess I have not finished to read it, yet…
Ok with good MP3s, OK with good 16/44 not inferior to “HiRes”, but I still suspect a marketing “bias” : as the HiRes files bring more money, the companies could be tempted to degrade a lot when downsampling to Red Book, so the buyer seems to hear a difference and ends by buying the Hires.
My experience in this domain is limited to a few albums : even with my 70yo ears I believed that the same album in HiRes sounded better : it put me “la puce à l’oreille”.
(RPi4, AlloBoss2, Boxem 4215, ESL63)
Haven’t thought about that possibility, but I suspect some big ruckus, if somebody finds out. Imagine somebody of us freaks downsample a HiRes file and run a diff between that and a original 16/44 from the same label. And find out there are diffenences which are not technically explainable. Even assuming the “degradation” wasn’t done by design, just by negligence. Post this result in some of the right boards and the reputation could be easily ruined. I think (and hope) no label will work at that risk.
I think the important word in your post is “believe”. There is a german saying litterally translated “Belief moves mountains.” And I’m sure something similar exists in many other languages, too. That’s the way placebos work (and they do). As I said some posts before: There would not be much left on the audiophile voodoo market, if this wasn’t the case.
EDIT: I have some CD’s (and Vinyl) which put me “right in the spot”. I even found some songs on Spotify which do so. The recording matters so much. IMHO the rest is marketing.
my 2 cts
- using a pc and feeding a dac via usb (through AQ forest, jitterbug or ifi filters): hires sounds predominantly notably better than cd redbook, which again sounds a bit better than mp3 files (320kB/s, slowest conversion).
- with a proper streamer (pro-ject streambox s2 ultra, running volumio, fed through a lin PSU, good fuse, power cable, AQ coffee usb, and a jitterbug) feeding the same dac, all sounds a lot better, and the 3 formats seem closer together, so that high resolution at first seem rather superfluous, but still music‘s fabric seems quite a bit more real and fluid, and the whole thing seems to come a bit more from anywhere except out of the speakers…
2 comments: this perception of mine isn’t tested on one suited track in redbook and hires versions (as it should, ie taking a non-upsampled original high res file, out if which a second, downsampled version of redbook standard is created for fair comparison), rather a general impression using a wider range of tracks (including pseudo hires upsampled from redbook).
Secondly, and this is a music‘s lover conclusions: if i buy (or record) music I try to get hires, if i listen, i take what is available (on ssd, Qobuz or Spotify (if it gets fixed again), and usually the (re)mastering is more important than the resolution of the file. I think I need to add, musical emotion is mostly influenced by the quality and balance of the hifi system, but rather least by the source file resolution.
PS: Nyquist Shannon theorem as far as i understand relies on a long sample, a rather infinitely long one; —> mathematical theorems should be interpreted within their limits.
About class D, have a look (er, an ear) at the Purifi modules : you won’t be disappointed.
Thanx for that tip, Sernyl. In fact I did look intensively and there is lots of information on Purifi vs. Hypex vs. Pascal vs. IcePower. I will try to keep this short in order to get not that far OT:
It turns out that - though better specs - the Purifis tend to sound a little bit more cold than Hypex. I don’t care much, if THD is far beyond or very far beyond audibility. And for convenience: with DSP there are only (so far) readymade solutions with Hypex NCORE, IcePower and Pascal (Four Audio). To raise my confusion even more: Now there’s NCOREx and those modules look surprisingly similar to Purifi’s Eigentakt. (Bruno Putzeys back @ Hypex again?)
Anyway the Hypex are already bought and I don’t expect a disappointment.
(and for peace of mind: the woofers will be Purifi )
Mine is made by Boxem, if I am allowed to write that…
I could not compare with others, coz I live in the open coutryside (when I am not at sea for months I had read a lot about it in french groups.
I enjoyed it at once (I still do), and incidentally was surprised by its channel separation, even while listening to the FM band.
I bypassed the preamp, so the amp is now fed by the Allo Boss2 (with volume control) / Rpi, of course.
“To nobody, except those who will take pleasure here” (Franz Schubert, 1797-1828)
To put things in perspective, the majority of high-resolution audio content is in the 24-bit/96kHz range and occasionally reaches 192kHz. The track transmits information from the original recording to the listener more effectively the greater its bit depth and sample frequency.