Inside High-res audio: PCM vs MQA vs CD: 2L Sampler Comparison
Вставка
- Опубліковано 14 чер 2024
- Objective comparison of the same content in high-resolution PCM, MQA and CD rate. Video produced in 2017 and uploaded now. Special thanks to 2L for providing sample content: www.2l.no/hires/index.html?
ASR Forum discussion thread: www.audiosciencereview.com/fo... - Наука та технологія
Keep them coming Amir, great work!
While on the subject of MQA, years ago when the annual audio expo was held in the metropolitan city I live in, someone from Meridian had a room demonstrating MQA shortly after its launch. The guy was doing very controlled auditioning demos where he would play some generally well known music tracks encoded with MQA and then went on to ask the audience if they agreed on how wonderful MQA sounded. Most people noded in agreement with some verbally expressing their positive impressions. But I, being the unfaithful and scientific minded bastard that I am, despite my name, had to be the only fly in the proverbial ointment. So I asked the showman to play an MQA track followed by its red book CD counterpart. He responded that he couldn't do it because he didn't have the original file on his server and that they didn't perform their demos in that fashion and asked me if I didn't agree that it sounded great. I replied, sure it seemed to sound good but unless we could compare it to the original CD track, we the audience had no point of reference to compare to, specially since the playback was performed under a setup, i.e. speakers, audio gear and a listening room that we were unfamiliar with. So I told him that his demo was not valid and didn't convince me of MQA's claims; you could tell from his facial expression that he didn't like that. It just goes to show you how unscientific and deceptive their whole approach to selling MQA technology is.
The point one should keep in mind with regards to MQA is why mess around with the integrity of the original PCM signal to save some bandwidth that is no longer an issue these days and have to pay some royalties to MQA Ltd. on top of that. Might as well use FLAC which is losless compression at around half the original size and and is totally royaty free.
All issues for MQA will undoubtedly be irrelevant when Spotify and Apple Music get into the high res game. I truly believe the biggest reasons people are picking Tidal over Qobuz is modern interface (including iOS app) and slightly better library. All the hassle of working around MQA through getting the right equipment (to unfold etc) is something the mainstream audience does not have time for nor are they close to understanding why MQA is necessary.
I have Tidal and have no issues whatsoever but as soon as Spotify releases hi-res I’m moving over to join my wife and kids and take advantage of the family plan (more affordable too). Then I will be able to download my playlists to cut down on internet usage. Something Tidal doesn’t offer.
The benefit of MQA is that it corrects the time smearing of the ADC at the recording studio and at the output. Therefore only recent recordings since 2019 that were recorded in an MQA equipped studio are going to sound good. There is no point listening to an old song that was MQA level 1 enabled and comparing it to anything. Only if it was recorded originally on a Digital Audio Workstation that has the MQA plugin is it going to be worth listening to. Also, most audiophile systems aren't powerful enough to take advantage of the MQA time correction - so most people can't hear it and hence dismiss it as snake oil. But on a Hi End PA system the difference in clarity of a full MQA enabled track compared to FLAC is very obvious, as can be heard here: ua-cam.com/video/Cl5ULnX4viU/v-deo.html
@@fx-studio mqa shill found
@@traceeaw Why is it that anyone who takes a stance that MQA is a good, valid format worth considering is accused of everything under the sun? You really need to open your minds a bit as almost every top chart song is available in MQA these days, and not for no reason....
@@fx-studio As long as there is no scientific validation for MQA's greatness and only words that comes out are marketing words. There is high reason why most of ASR audience understands that it is snake oil.
Any competent mastering engineer would not be surprised at these results. I have been banging on about this for years. Higher sample rates do not automatically mean better quality and can actually degrade the sound with added distortion and noise. For the listener playback of a well recorded CD at16/44.1 is very hard to beat. It is tempting to assume bigger numbers must be better but higher sample rates can introduce problems for little or no audible improvement. Also MQA is a failed solution looking for a problem.
MQA is a scam, using woo-woo marketing to fool credulous audiophiles just as surely as fancy cables and silly tweaks did in the past. Except that most silly tweaks don't run the risk of making the sound objectively worse whereas *MQA does objectively degrade the audible signal.* It's trying to put a commercial brand name upon something that's objectively worse than open lossless formats. And then hold claims of "high quality" audio behind spurious license fees.
@@sjwright2 Tidal MQA is fraudulent. Hopefully Spotify hifi will kill MQA when released later this year.
Take a look at Mark Waldrep’s site realhd-audio and the video from GoldenSound.
@@geoff37s38 shame Amir banned him from ASR lol on the backs of others and 180’ing on a swivel any time that suits
I always thought bit depth was more important then sample rate
@@RossKyle95 amir banned Mark? Why was that? They seem to be on the same page afaik?
loved this Amir! it confirms what I always suspected. CD quality is rather pristine for distribution. a producer today still needs to work hard to create content that exceeds the capability of the lowly CD. then we can FLAC that for easier sharing. so what can be better? MQA is an idea but then gets you into licensing and politics, meh. "Hi-Res" PCM just wastes lots of bits to encode a bunch of noise and artifacts way outside of hearing ability. is there a format that increases actual audible quality as you increase digital file size? I'm tempted to conclude that 24bit/48kHz PCM is as good as it gets for humans, then you FLAC it for transport.
Agree 100%, honestly CD quality is good as it gets for most people, 24/48 goes just a step beyond.... No need for DSD or MQA, .WAV and .FLAC are good enough.
abx blind tests between mqa and flac would most likely show how little (none, most likely) difference there is
@@thomasward00 Some (not all) DSD and DXD tracks sound better. The same goes with hires tracks: not all of them sound obviously better. Some sound worse: maybe they messed up trying to resample a low res track.
Great analysis. Particularly pointing out the MQA compromise when playing in a "backward compatible" system without a decoder. Thank you!
I wish I hadn’t found your channel!, . Since I started looking at your videos i am getting more and more interested in this science, so much so I am to have to start at the beginning ( very basic level ).
Please keep up the great work.
Thanks! This was another excellent presentation. Hi Res appears to be the emperor with no clothes.
I really hope that you and people liking your comment are not also the ones complaining about DAC reconstruction filters pre-ringing etc. Because, that is exactly where this whole high-res story comes into play and helps DAC reconstruct the time-domain more faithfully to the music.
This emperor is very much so dressed ;)
@@marcelwenting The emperor might as well be dressed, but if the color of his clothes is beyond the visible spectrum, it's going to be transparent to us, mortal humans, anyway.
Great timing... super interested in this.
Great Amir , very impressive analys
Really missed your videos. Thanks a lot.
@Amir, Thanks a lot for sharing this valuable knowledge, Will you be able to do a video on how streaming music content quality is like; e.g TIDAL HIFI or Spotify High Resolution .. since you explained the MQA ..I don't think we need to look at TIDAL masters .. but I am interested to understand how the CD-quality 16 bit / 44.1 KHz is transformed and streamed?
Am I right in summarising that the sweet spot for HiRes delivery according to this video, within current formats, appears to be (of course non-MQA) 88.2->96 (which will always come as 24 bit)? This appears to have preserved the best behaving content in the audible range, without delivering toooo much excess content in the ultrasonic. Normally a DAC cuts content above ~ 25kHz anyway doesn't it, with something like a linear phase fast roll-off?
24 bit/48 kHz should be the minimum. Above that, I prefer 88.2 kHz sampling as it is a nice multiple of 44.1kHz. So yes, you are generally correct.
I am a mastering engineer. I can say from my workflow perspective. Those peaks could be produced by the encoder (MQA) as expressed by The análisis. A good DAC will sustain the clocking to avoid sample peaks in combination with a true peak limiters and filters.
I tent to cut from 18@20kHz simply because the market I work for dose not reach above 18. Also the encoding from a digital distribution is not great so avoiding to much information in the top end and controlled low end brings a safer encoding processing. MQA is just a tone shaping tool that should be in the hands of professional and not on playback.
This was extremely illuminating
Thank you so much for these videos!
Thank you for your time.
Did the peak in the CDQ file show on other formats if you set the horizontal scale to 22.05khz? I have some recording equipment and keyboards with high frequency hissing, especially stuff from the 80s with lcd displays. Personally I think it’s better to use 44.1/48khz files if available, there is no point in playing content above hearing range as it wastes headroom although at these levels I doubt it would make a difference either way. On the recording side of things, I know for a fact that a lot of ADCs sound better at lower sampling rates even though they support 192khz.
Thanks for the videos, I really enjoy them.
Hey Amir. Please make null test all these formats against cd quality and MP3 320. Please show all tidal fans what for they fighting with.
Thank you for this explanation. It really helps. I would be curious to know which format/sample rate you would select if you were purchasing or listening to the music blind? If you were just listening and couldn't look at the graphical data of the music what would you set your player/streamer to? Thanks so much for all of the work you do for the audiophile community.
In general I like to have the original stereo mix prior to mastering. That way there is no loudness compression and I get the bits as they were produced. I can then convert it myself to whatever format I want with proper filtering and dither. Otherwise we are at the mercy of the mastering engineering in how that conversion has been made.
Great video with excellent explanations!!!
I'm not convinced that the information above 24-25 kHz is actually music. Most microphones used in studios cut off pretty sharply around that area. I think it could be a result of the analog to digital conversion "smearing" the signal in the frequency domain
Could be. You can cut off the ultrasonics and then pitch shift them to see if it is musical. I am too lazy to do it so don't ask me. :)
@@AudioScienceReview I'm also too lazy/busy but I might make a post about it on the forums when I get time =)
@@AudioScienceReview The user SquidCaps wrote an excellent comment about this topic on your first video on the topic - here is their comment: ua-cam.com/video/AhDvNe6Pw7c/v-deo.html&lc=Ugx8xPW-pg2vslpog094AaABAg
I agree completely with what they have to say. Even if the data in the 30 kHz and beyond range is actually music, it's still better to lowpass it away because if the transducers cannot recreate the signal reliably, then nothing good comes out of having it. Furthermore, even if we have transducers that canr ecreate all of that, it's not good for our ears to be exposed to near-ultrasonic sounds, especially for a long time, because it can lead to hearing loss, even though we can't hear this stuff.
@@ayane_m Is this an MQA specific thing you are talking about?
@@ayane_m wow there, hold ur horses. hearing loss from signals at less than -100dB recorded and then played back through a chain that is not designed to reproduce it, so it will be attenuated further? come on, that is just horse manure.
Furthermore, everyone here seems to have forgotten about DACs and time reconstruction filters. It is one of the biggest downsides of the CD sample rate and THE reason to go high-res. It is not at all about hearing the music above 20kHz, it is all about making your DAC sound better because it has more time-samples to work with.
Note to self: if comment exceeds 3 lines, type it in a text editor, then copy it over. *sigh* Stupid youtube cleared my original comment just as I was finishing it.
A repeatable comparison, and one easily done in Audacity (minus the real-time FFT).
Spectrogram settings in Audacity:
Max Freq: 400000 (it's default to whatever Nyquist is)
Gain: 0db
Range: 120db
Window Size: max narrowband
Window Type: Hann
I downloaded and compared the following file types for the following tracks:
DXD (aka PCM 352.8), 96kHz, CD (44.1), and MQA full resolution (note: a decoder is needed for MQA, otherwise it's just CD + encoding in the 15-22kHz range which appears as noise).
*MAGNIFICAT 4. Et misericordia*
Same as Amir used in this video.
Same results, but I'll add a couple more comments.
MQA unfolded twice is an improvement over folded MQA, but it still doesn't match up with CD, so..... yeah, still not lossless.
Also, I found it odd that the DXD file was downsampled to 96kHz instead of 88.2kHz.
Overall, this is a poorly recorded track, in my opinion. At the very least, the equipment used wasn't up to task.
*Chromatic Fantasia and Fugue in D minor*
DXD: One noise tone at 55kHz. Nothing above that.
88.2kHz (hey, they downsampled to a proper factor this time!): *zero noise* Yup. No noise tones anywhere.
CD: Same as 88.2kHz sample, but less freq. range. Nice and clean.
MQA before unfold: Same as CD except for encoding in the 15-22kHz range which appears as noise.
MQA unfolded twice -- same results as before.
In the end, I still like 88.2/96kHz sample rates (or maybe it's the 24bits) if they've been recorded and sampled properly. Where I notice the benefit most is from analog sources, but I see no benefits above 96kHz, personally.
Okay, I've officially spent more time on this comment that on the analysis. Y'all open up your audio editors and test this with some of the other tracks in the link, just for shits and giggles. I've got a bed to smother. Good ni...gh...errr... morning. *haah*
Guess they don’t expect someone to analyze there product. Thanks again!
Hi Amir, another beautiful video and very useful. Once you are done with digital format it would be nice to show us the difference (in this graphical way) between FLAC and analogue signal i.e. from the turntables. What is that people really love about analogue music - is it visible or it is just a matter of "touching my records" etc. Thank you as usual. ;-)
What people love, I'm pretty sure, is the noise and distortion that records produce.
@@glgermain what people love is nostalgia
For music produced during the loudness wars (and even modern pop music), a given vinyl album will likely have more dynamic range compared to the digital version. Look up examples on dr.loudness-war.info.
Yet another piece of info that shows that MQA can only degrade quality compared to CD. Even subjectively this is noticeable. MQA made the treble sound smeared.
Very good info - thanks for this !!!
MQA doesn't play 44.1 kHz at 44.1 kHz on output. It always doubles the sample rate x2 so it will be 88.2 kHz on output that throws away little "artefact" beyond the audible area. It would be honest if Amir could do measurements with the DECODED MQA file instead of undecoded raw one.
Good work Amir - interesting and informative content as usual. It would be interesting for someone to get a vid conference with 2L and Morten Lingberg as they are at the forefront of MQA production - get their take on this and the whole non-white glove produced music sitting in Tidals library and where they see it going for the future. I'm pretty sure he's pro-MQA. If he thinks he is getting a better reproduction who are we to argue unless people don't want the hear the studio original as intended? (as per the intended results from true MQA processing)
What is the point of that kind of conference? MQA is a scheme to make hi rez files take up less bandwidth. This is will known and undebatable. MQA can not make an original hi rez file sound better than an original hi rez file. This is also well known and not up for debate. No one in their right mind thinks that MQA somehow provides a "better" reproduction than an original hi rez file. It is not adding anything that was not there, rather it processes and subtracts information that is deemed not to be useful.
So, why would you want MQA? Because you have a slow internet maybe? Because you want to make streaming less expensive for the likes of Tidal? Because you like the branding?
@@petermartin9494 There are not many interviews with studio engineers/ producers and musicians outside of the MQA youtube channel. What I was interested in was whether the original producer was getting a more authentic reproduction of the music once the ADC and DAC colouration were corrected for in MQA (as quoted in MQA literature). If the producer says he getting a more faithful reproduction (as do many on the MQA channel - but they only the show the positives ) then you've gotta think that correct application of MQA works, but not necessarily on some of the mass encoded stuff added to tidal by labels. I'm not biased either way at way at this point, but I'd like to try more examples of MQA to get what it claims it can achieve.
@@tobermoryman The idea that MQA somehow corrects for ADC and DAC coloration is frankly ridiculous. As if it has some kind of magic that corrects any coloration in any ADC and DAC. If this were true all anyone would need is a $100 MQA DAC since MQA would correct any issues with it. MQA is a compression scheme to make streaming audio files more economical, end of story. There is no evidence supporting the claim that it corrects ADC and DAC colorations. Anyway, it is unlikely that a serious recording engineer would use an ADC with some kind of coloration.
whether you can hear the noise that mqa cd adds above 15khz or not it's still being pumped into your tweeters and will distort the musical information that they're suppose to be working on.
If I get your points correctly:a) if you pay a premium for a product, you should get a premium quality product whether you can test for it or not; b) those producing downloadable content have the wherewithal to properly polish the material.
Exactly!
how do they manage to have this included? failed formatting?
You should do what a cd player looks like when using regular SACD, maybe blu ray audio comparison
So basically what I am gathering is that the mastering is the key. Actually the analogue to digital conversion isn’t up to par in the production of the music. So most hi res audio is either useless or just adds noise and distortion.
Am I understanding this correctly?
Thank you.
I would say mixing rather than mastering, the later doesn't really constitute as much part of the work as the former
Once again MQA takes another slap to the face
I realise that Amir wants to maintain his objective neutral stance, but failing to emphasise the obvious stupidity of the MQA format is a disappointment, especailly given that it's objectively worse than CD quality in the audible band. There's nothing MQA does that's useful to anyone, other than turn on an Idiot Lamp on some DACs. The only _theoretical_ rationale could be file size savings, but even that's ridiculous in 2021. If you have the bandwidth to stream netflix video at _medium_ quality, you can stream the very finest of uncompressed HD audio. Why would anyone who cares about *EXTREME AUDIO QUALITY FOR BATS* even _want_ to save a few megs on an objectively worse format?
@@sjwright2 anti time smearing filters. If you can’t tell the beneficial difference then I hope you haven’t spent too much on audio because it’s clearly wasted on you 😅
@@Rhcpbedders lol
@@Rhcpbeddersyou can do that without messing with the file. They could just use the mqa filter with upsampled pcm and it should do the same thing if it makes a difference
With the amount of streams being generated, MQA does make an environmental difference, at least because of its file size - now the extra energy needed for whatever it takes to decode it, implement and promote it is impossible to calculate I guess… anyhow, it is a dead format. There was more to MQA than just the compression, which I do not fully understand - a way to reverse filters in dacs, a way to make sure it was lossless in the sense that what was fed into it also was what came out - some type of watermark or error correction system…
nailed it! thanks amir 👌
How about dac filters and aliasing etc... Is high res music identical to low res in audible band let's say below 15khz ?
VERY interesting to 'see' what's going on above 20K. Thank you. Maybe it would be fun to see a couple of FFT's of regular, red book, CDs?
It is not as much "fun" since most if not all of that spectrum is taken up by what could be music.
@@AudioScienceReview oh, OK, thanks for responding.
Hi Amir, is it a bug in your Adobe Audition that makes it show 32 bit no matter the files actual bit depth?
The bit depth shown should correlate to the file. As the sample rate
I notice this just now but didn't understand it when I watched the first video a few days ago where I commented on the clipped sound example.
Edit: Downloaded to see what's going on and it's a bit weird here actually. 24 bit files showing up as 32 bit in Adobe Audition, which they should not do. I also checked the files in Izotope RX which shows the files bit depth as they should.
It probably converts it for editing.
@@lightbit553 That shouldn't be a thing but I'm not using Adobe Audition more than checking these files to see if I had the same thing... so maybe you're right : )
Audacity does this to prevent losses while editing. It is more precise that way.
@@lightbit553 Do you mean Audacity or Audition? : ) Maybe you're on to something then.
Audacity. I don't know for Audition, but I assume it is similar.
I'm lost about the 48 in a 96 file why not the full 96 being used and what is the rest of it used for
Was this music atleast mastered well? Atleast then there could be some justification to the price. 16/44.1 is all you need but alot of the high resolution stuff is mastered better which is the real appeal.
What he failed to understand is that this is why you don't do unfiltered DSD->PCM conversions. Either you output to a DSD capable DAC or you use a filter between the conversion. The sacd plug in foobar for example not only enables you to do the conversion on playback, also includes different types of filters for higher frequencies.
What does DSD to PCM conversion or SACD have to do with this? Amir is analysing files which are in PCM format. And in the rare case the avarage end user would play DSD media, it would usually not get converted to PCM as most modern DAC's use bitstream technology internally.
@@4everB2 if you look at the files he is using then you will see that they are dsd files that have been converted to pcm so that they can be open in audition which does not support dsd files. The massive noise on the high bands are from the dsd format of which it is a feature. So in short, everything.
@@JohnTwo1 It doesn't matter to the end user what happened in the production process, the only thing that matters is that he receives a PCM file with garbidge in the high end. It's also not relevant that you can't open a DSD file in an audio editor, because this is an analysis of PCM files. It's also impossible to determine the root cause of the high frequency noise just by looking at a PCM file. And if DSD conversion is done right it doesn't necessarily lead to high frequency noise. Why would you pay extra for that as a customer?
Flac does not support dsd. It is not suitable for it.
does mqa make the quiet autible and if you sample at above 16khz it's not autible?
Nice video Amir, very interesting. Some comments:
1- the odd peaks of unnatural tones @19.4Khz in the 16/44 Flac, couldn't be a aliasing phenomena of the brickwall filter?.
2- At about 10:30 you mention the non-existing high dynamic range of music in these frequencies, I guess as a way to show how little information from ultrasonic bands MQA needs to pack in the folding process (beneath the noise threshold in the audible range of 20-20Khz). Then in 11:15 you show how this folding adds non-musical information from some 17-18 Khz up in the MQA file, and you label it as an "unquestionable degradation of the PCM file". I think what you are seeing there is not only the encapsulation process, but also the noise shaping of what's below the noise threshold of the file. MQA alleghedly reshape that noise, specially from mid-bands range, into the upper octaves. This is because human hearing threshold at that range is much less sensible (Fletcher-Munson), and so in that fashion they actually GAIN headroom in the bands that matter the most. It is not an anomaly of MQA but what the do by design, looking for an enhancement of the signal. Also, this shows how you can't expect in MQA a bit-per-bit match against the original file, as all the noise floor is fundamentally different, being replaced with real information, then dithered to still appear as noise as in a normal redbook file. This is not a cause for alarm, but in fact a mere description of the process MQA does.
3- In the explanation of that lower dynamic range, you limit the comment to the inaudible range. But the same happens in several upper octaves of the audible range: no real instrument, because of vibrational laws of mother nature, is able to render high amplitude of harmonics in the upper octaves: the exponential roll off of music you mention in the video. What I think MQA does here . In other infamous tests published in ASR, feeding the test signal not with music like here, but with high amplitude test tones in higher frequencies instead, it is shown that MQA soften the shape of those square waves test tones, and in general some roll off of higher frequencies above 16KHz. Again, that's by design: as the algorithm doesn't expect high amplitude musical content there, the slow roll off filter is in fact shaving the upper bits in those frequencies. That's not a defect: it is exactly what makes MQA sound better, imho.
4- I found very noticeable how much cleaner the signal of MQA in the spectra compared with the other formats. It would be nice to have a screenshot of exactly the same second of music of all these formats. MQA, I think, is showing much more clearly the harmonics of the chorus singing this Magnificat, in a way the others don't. I think this could be explained by the more clear signal of the MQA that got rid of the pre and post ringing of the impulses in the music because of the soft filters applied (and, rigorously speaking, every sample is an impulse response). As each one of those pulses has less ringing, the result is a clear signal. The audible impact is a more clear and focused sound, imho.
5- This test was done without MQA unfolding by hardware, even with a low sampling rate for MQA standards. I guess the same thing noted above regarding the cleaner signal achieved would be seen even better in a higher sample file, unfolded by hardware.
Again, thanks for the test and video. I don't know if purposely or not, it discredit much of the criticism of MQA that's have been so popular recently. Best wishes,
Pablo Bañados, Chile ("mieswall" in ASR)
Point 3 had a crop of content... "what I think it MQA does here ." this continued with: "... is to implement backwards way into the audible range some very slow roll-off antialiasing filters instead of the brickwall-like ones of RedBook. In that way they achieve the much better impulse response typical of MQA files: 3 to uS instead of the 500-5000 uS of RedBook, or the 150 uS of high-sampling PCM files".
Awesome. Just awesome.
i wouldn't consider mqa as "hi res audio"
Indeed, it is a lossy and DRM-encrypted below the norm of CD quality, and which degrades the original with garbage...
The problem in that Amir has been doing some damage control when the MQA hoax is debunked.
@@brunngraggan679 My "damage control" is limited to damage people do our understanding of audio as you just did by calling MQA DRM. There is no DRM in MQA. You can copy any file, give it to your friend, do whatever you want. Your "right" have not been managed. Digital Rights Management is called that for a reason. Get educated, don't spread gossip that is not technically correct and then you won't have me on your back.
@@AudioScienceReview MQA does not restrict making copies. However, it uses cryptographic measures to restrict certain other uses, for example the playback on devices that are not authorised by MQA or the digital copy of the "unfolded" audio. That is the textbook definition of DRM. I honestly cannot understand why you keep insisting it has no DRM, especially given your technical knowledge and credentials.
@@katjanonsen1690 It's a format that requires a player to support it. The fact that support is a license instead of a hardware dependency is kind of besides the point. It's like calling DVD DRM because you can't play it back on a VHS player.
@@joshfoss7407 DVD has DRM because of CSS, again using cryptographic measures to rectrict certain uses. DVDs can't be played back on players (software or hardware) that do not have the CSS decryption key. That is in fact similar to MQA, the exception being that DVD uses symmetric cryotography while MQA uses asymmetric cryptography.
This channel is a breath of fresh air in a demon haunted world.
Are mqa cds better than sacd or cd.is it worth buying an mqa cd player like technics gl700 mark 2 sacd player.
How udible can be this difference in mqa version?
Amir your work is excellent, congrats for bringing the audio community real arguments. I am a 51 years old audiophile. Best case scenario I can hear 15 kHz, only some kids can hear up to 20kHz. So what is the use of spending a lot of money in digital downloads with resolutions audible only for bats and dolphins. It's absolutely useless, just a waste of gigabytes and money. Most components have the added disadvantage of generating tones in the audible range because of intermodulation, that is the price to pay for an extended bandwidth, so probably with high resolution downloads you hear sounds that are not in the music an undesirable result of electronic circuit design. I think CDs (44.1kHz 16bit) are pretty much what everybody needs, just take care of getting a decent CD player. I like SACD, but must recordings in DSD where recorded with the usual bandwidth, you pay premium for a better decoding technique, resolution wise it's also useless. Most studios record at 48kHZ 24 bits, that is far more than what is actually need for 100% of the humans... the large numbers are just a marketing scheme and a way of ripping money out of the ignorant.
In general there is no reason to get high-res music if you can assure that the conversion from high-rate format that was used in the studio to 16/44.1 is proper. Problem is that we have no idea how that conversion was done. It could have truncated bits instead of dither for example which can be audible. It can also use the wrong filtering causing potential audibility issues. And finally, it is possible to hear the noise floor of 16 bit audio. So all else being equal, I like to have the high-res file. But if folks want to charge me more for it, then I like to have it be pristine and in every way better than 16/44.1.
Very enlightening Amir, Thanks. Just one question if I may, this high quality audio containers work like a ruler that can have infinite length between it's values or as a bucket that will fill up from the bottom up? Is it correct to say that a higher sample rate will have more information between the sound curve nodes?
My pleasure. Higher sample rate can do that if the original content had higher bandwidth. As I show in my videos, there is some information above 22 kHz that CD can hold so there is some objective merit in that. Subjectively though you can't hear that ultrasonic aspect so it doesn't matter that the faster transitions in music were represented more faithfully.
@@AudioScienceReview thanks for your response much appreciated although I ment to say bitdepth not samplerate
Can you do a comparison with DSD? I’m curious to see how it compares
I wonder if the claims of MQA having superior time alignment and phase correction have any merit?
Superior to what?
It has never been studied to demonstrate to be correct. In general, time accuracy of digital audio has to do with its bit depth and bandwidth, not the sample rate, filtering, etc.
It'll visually make more sense of you added a marker at 22khz?
Great, informative vid as always
It would but requires editing the video which I am loathe to do. :)
@@AudioScienceReview fair do's. I just assumed u can plot markers within the Analyser, like OTDRs. Never mind
I don’t think this is a fair assessment of MQA without full hardware decoding.
When I was young, I could hear those "ultrasonic" frequencies, something I discovered while working with an ultrasonic motion detection system. Back then there wasn't much in music recordings to hear, gramophone records struggled to get the top octave, as Did Compact Cassette. And of course the CD was bandwidth limited by the infamous "brick wall" filter. There were the 19 kHz pilot tones for FM stereo, and MTS stereo TV broadcasts had a similar pilot at 16 kHz. Some FM tuners had a switch that allowed the pilot trap filter to be switched in or out, but no TV sets had similar switches. Of course one could hear the CRT "sing" when repetitive picture signals caused the yoke to vibrate.
Back when amplifier power was costly, and ferrofluid cooled tweeters was still to come, a lot of us young music lovers burned out tweeters as we tickled the dragon's tail of clipping. I figured those days were long gone. Imagine the irony of an audiophool suffering the same fate because of buying recordings that used ultrasonic noise as filler. It does seem that the Ri-Rez pushers are getting smart after being exposed as selling CD-DA rips padded with zeroes as being original works. When the extra samples and extra bits contained _no_ information, they were easy to spot. Then again, so all those birdies.
Thank you Amir! But how about electronic music which was made @ 96 kHz? Please check something like BT tracks.
Can do but what is "BT tracks?"
BT is an artist
Visited this page multiple times with multiple hardware setups and really just couldn't hear a difference with my ears.
Anything I thought I could detect was all placebo and didn't stand up to a/b testing.
I'm a SACD, DSD, DSF freak, but early on back in 1998/1999, when the format was released I knew it was "menure like"😝 When I read about the noise shaping implementation, intuitively I thought that most of it would be the bulk of the touted ultrahigh frequencies content of SACD. I know it doesn't deliver, but the whole thing surrounding the format interests me and it's like a hobby to me. I guess I'm still hoping that the mythological potential of the format someday is expressed. It's a "hoot" tracking down old SACD players that do the native DSD to analogue conversion (if there's such a thing), obscure SACD discs releases, finding out which BS "high-end" brands still produce players today, the never ending reissues from "audiophile" labels and important albums (how many different issued Kind of blue SACDs can be released?!), struggling and unknown artists that use the format to record their albums because they know there's an audiophile market, etc.
In my day to day listening, despite having the whole SACD set up (multichannel please! 😜) I use a pc with a topping DAC and I'm good, but it was the "journey" into SACD that made me aware of important artists and wonderful now classic albums.
you can always get a PS3 and then get an HDMI to 5.1 rca adapter (which also has an hdmi out to the tv so you can see what you're doing)
Will the 19kHz tone on the CD file is due to aliasing issue of the hi-res file? Then, the down sampling software is fraud.
I don't know. As consumers we don't have the necessary information to work backward to figure out what is going on. People creating music need to investigate them.
To my ears, MQA sounds dull compared to CD flac. It sounds like more than just the high frequencies are being messed with.
I had this same experience but the main reason I quit my subscription was because some of my favorite artists releases are now all MQA and the regular lossless quality versions of the albums long gone. I'm talking about classic rock albums that have nothing to do with this gimmick format being ruined bc they're only streamable in MQA now
CD 16bit resolution -- will noise below 96db be heard ?
A very quiet room is 25dBA. An extremely loud system reproduction level is 105dBA. The range between these two extremes is thus ~80dB of dynamic range. In order for 96dB of dynamic range to be _genuinely_ useful we'd need to be seated inside an anechoic chamber and listening at volume levels where any signal peaks would cause rapid hearing loss.
@@sjwright2 Pardon can't hear you.
In theory yes. Please see this article I wrote on this very topic: www.audiosciencereview.com/forum/index.php?threads/dynamic-range-how-quiet-is-quiet.14/
This goes to show that it can pay off to have a 2nd order filter at 45kHz to dampen out much of the noise higher up.
Interesting analysis. For sure it would have been better if those spurious noise spikes at 38+KHz had been eliminated, but it needs to be emphasized that they are at a low level, about -110dBFS. Therefore the risk of intermodulation products falling into the audible range and stress to tweeters and amps is almost non-existent.
Also it needs to be said that the 2L recordings are among the highest fidelity, best sounding recordings out there. My only criticism of their sound is a slight tendency for brightness.
ok so what about the differences seen in the audible range? are these hi res files that have a bunch of noise and random tones any good in the audible range?
Unfortunately our tools are inadequate to useless for that type of analysis. Good stuff is mixed with the bad stuff so no way to separate the two without statistical analysis to separate noise from music. If I had no other work to do, I would build such a tool but I am busy making these videos and doing reviews. :)
One notice about 2L is that these “test bench” files you download are all in FLAC format. How this affects the original files I don't know. Just worth mentioning : )
MQA uses FLAC. It isn't an entirely separate format. It's flac but with the audio content and metadata modified
@@GoldenSound Thanks. It makes sense with MQA then but not when comparing the other formats since they all get restricted to the conversion to and playback of FLAC.
Does it matter at the end. Not sure but for pure technical comparison it should matter?
(Also worth noting that I see other issues in the frequencies than Amir. For instance the spike at 19kHz in the 16/44.1 file is not there anymore.)
flac is both a container (format to put audio bits in) and lossless compression. Neither changes the nature of the bits as analyzed or played.
@@AudioScienceReview Thanks, the little I know about FLAC. (I'm on Mac so I never use that format.)
@@cornerliston mac supports flac. Everything supports flac nowadays. Also, you can just as well use pcm in wav container (by decodinh mqa flac), it does not matter, all data is hidden inside pcm words.
Fella im sure it really matters not if the music is still enjoyable. Maybe to the everyday listener the measurements, digital errors etc matters not and also cannot be heard. What are your thoughts?.
I am sure if the waiter spit in my soup as he brought it to me it would still taste the same. All else being equal, I prefer that he not do that. :)
@@AudioScienceReview ha ha spit in soup is entirely different but I get ya. How would you know someone has spat in your soup and what if the food still tasted good, regardless?. Right back at ya Amir ha ha.
@@AudioScienceReview Look at audio products as cakes made up of various ingredients. There could an ingredient in there that is not perfect for the intended purpose but the overall cake still tastes good IE the musical experience.. Or it could make the cake, IE the musical experience not sound good lacking in certain areas hindering the overall performance, taste, etc. I like cakes ha ha.
Comment to support the channel. Thanks for another great video!
But we need to have the ultrasonic security systems recorded in our music!
In this video about noise in audio, you actually have some noise in your own audio! :D Maybe need a better microphone? Other than that, great video!
These are very old videos I am uploading now. My new setup doesn't have these issues.
Hello!
Let us hear this signal above the acoustic band. Cut the band below 16kHz with the FFT filter, boost it and play at a rate changed by half or more ("adjust sample rate" function).
Greetings from Poland!
HI there. The filers are there for you all to do that. :)
As an engineer, and not a scientist, I would be grateful if you could give me an idea where you're going with all this. As a bit of a scientist, I understand that you are collecting data for some yet to be resolved theory. Fine. Except I'm struggling to figure out where you're going.
It is intuitive that, if you record audio data more precisely than you can sense it, the additional data is just extraneous. For example, if one good musical instrument's harmonics bandwidth rolls off naturally at 17,000 Hz, what is the point of recording data (for playback to humans) to 200,000 Hz? Are there time-domain or phase accuracy aspects which improve the human perception? Does the additional speed and precision in parsing the music waveform provide audibly greater playback accuracy to a trained listener?
Our compatriots at Philips and Sony presented us with CD Redbook, lo these 40+ years ago. They too were balancing audio quality vs. media cost vs. SOTA digital technology. There were far more hurdles than most mere audiophiles comprehend. Arguably, they did a good job.
I suppose, after 40 years, we should have moved on in all aspects of music recording and distribution. Is this why you're looking at hi-res audio files?
I'm grateful for the efforts, nonetheless. But I feel you're taking me well into the woods with no trail to follow. Where are we going? And, how do we know we're there? -Cheers
Pretty simple. For a while I was purchasing a lot of high-res audio downloads. I had automatically assumed that such content would be recorded without any of the faults you see. My hope is that by showing the problems, content produces will go back and clean up the production. Indeed, this has already happened in the case of Sound Liaison sampler video I did. So no different than reviewing hardware: hoping it brings improvement.
Continuing to look at hi-res downloads to see if they live up to their reputation seems like a fine use of time to me. Hopefully it will also keep record labels in line.
Why to people worry so much about sample rate? Isn't bit depth more important?
both are really, but see high sample rate is needed more for recording and editing. so selling a high res file keeps a studio from having to use dithering when truncating down to 16/44 for cd, which is adding noise into the signal so should technically be closer to what they hear in the studio. (unlike mqa). personally i've heard some marvelous sounding cd's and really all we need as they're mastered properly and HDCD's could have been the answer to the loudness war if the studios insisted on releasing a hot mix. too bad microsoft bought and murdered the format. Grateful Dead about the only ones still releasing them.
@Audio Science Review I have a DAC that has a working blackwall filter, would enabling this be a good stopgap?
You don't even need a brick wall filter. Just a gentle sloping filter will do the job just as well if you have the choice, many of these are already 10-30dB down at 30kHz.
/Edit: ignore this thought error. See asr reply below.
It would not help. The sample rate determines the cut off frequency and since these files are high sample rate, the bandwidth is quite wide. The only way to get what you want is to resampling the music to lower rate, then the sharp filter would cut off the ultrasonics.
@@AudioScienceReview Bummer! thanks.
A tone between 19 to 20 kHz is probably a fire alarm pilot tone. If the system has a public address feature standard EN 54-16 requires testing the wires and repeaters and so on at least once every 100 seconds. Most systems do this continually for simplicity.
Also for simplicity they purchase amplifiers, repeaters etc. that are designed for audio. A truly ultrasonic test tone would work, but then some engineer needs to justify to another engineer why they're expecting an op-amp rated for 20 kHz to pass a 21 kHz signal. "Because of course it will" is a bit too much seat-of-the-pants common sense for the life safety business.
Probably the right call overall but it does mean a lot of real venues and recording studios are contaminated.
DSD in multichannel music or Blu-ray Audio in AURO3D or ATMOS!
When you listen to well recorded multichannel audio music you never go back to stereo!
How do you know it's well recorded? Unless you've done an analysis you'll never know.
I think there are too many variables/techniques in analog or digital audio playing/recording/listening path. Microphones, cables, analog to digital conversion, summing circuits, eq, compression, limiting, effects, mastering, digital to digital algorithms. Also we don't have the same hardware/ears used when the audio tracks are finalized. I personally prefer Qobuz master tracks versus the Tidal ones because there isn't any conversion in the middle. Also I always listen to the music with active speakers, the same used in the professional studios.
There are many factors indeed and hence my request that proper checks be put in the production to catch issues. Just using one's ears is not enough. Objective analysis like i am doing takes little time. Now if they did not publish high-res content, that would be one thing. But when they do and routinely charge more for it, then they need to pay more attention.
@@AudioScienceReview My two cents: I play the violin in a Radio Symphony Orchestra in Germany. Although we are a ‚real‘ orchestra in that we tour and play concerts asf we also produce a lot of music. When we record music however everything concentrates on the interpretation and tiny inaccuracies regarding ensemble work and balance and such. The conductor has the final word, the sound engineer points out stuff we didn’t hear and normally this gets done in a matter of three sessions of 4 hours each for an entire Symphony. 100 people are sitting on a huge stage, 4-6 engineers running around, sometimes cameras for additional video productions. I think because there are so many factors AND people involved in a production like this it‘s next to impossible to do that. This is not a controlled and permanent studio setup situation - sometimes after a session they have to remove everything from the stage because another group might be performing later that day on the same stage….
I just think it can‘t be done - under circumstances like this, which is normally the case at least in classical music.
@@goons7ersakass866 The recording session is just the first step in the production. Once the recording is done, you can use as much time you want to produce the master from the recordings. The analysis Amir mentions should be done as a part of the mastering process.
You think "there are too many variables/techniques in analog or digital audio playing/recording/listening path" to be able to do what?
There were always lots of variables in recorded music, and yet lots of great recordings were made.
Good recording. Good mixing. 44.16.eg.cd is all you need.people talking about mqa.mqa is just sales talk.i have not got a so called hi end system. Cd audio research. Amp.kimber cable. And staxs. Head phones. I've got 30 year old cds that sound amazing.
Without the ultrasonic THD products of the mic preamps used in the recording session the sound just can't be any good.
Which preamps?
@@Wizardofgosz The mic preamps used for the recording session.
@@billmilosz ultrasonic THD preamps. Tell us more.
@@Wizardofgosz For a recording of music, you need microphones to pick up the sound. If you are using microphones, you will have to have microphone preamps. All amplification circuits created harmonic distortion. ALL. Better circuits create less harmonic distortion, but ALL gain circuits create distortion of various kinds, including harmonic distortion. The mic preamps used in the original recording session will produce 2nd, 3rd, 4th, 5th, 6th, 7th harmonics-and on and on. The mic preamps used in the recording of the music put on this so-called "hi resolution" will produce harmonic products way up in the spectrum- ANY mic preamp would.The music will not likely have any energy up that high as musical instruments tend not to produce ultrasound, and recording studio mics couldn't pick it up if they did. So, in these high resolution recordings there won't be any actual content produced by the instruments up in the high ultrasound regime- the only thing that will be there will be artifact like pilot tones from equipment leaking in via stray capacitive coupling, noise, and harmonics resulting from the harmonic distortion of the gain circuits used to make the recording. Since in general the higher the gain in a given stage, the higher the distortion, and the highest gain in the recording chain will be the mic preamp, it stands to reason that if the digital recording is able to capture these nth harmonics way up near the AM broadcast band, and there are no actual sounds recorded during the session up that high- then the only thing that will be on the recording in the highest region of frequency will be noise, artifact and distortion. Since subjective-fetishist audiophiles will ALWAYS say that these hi-rez recordings sound so much better than recordings made to redbook standards, ideo sequitur, the reason that these recording sound "so much better" is because they have this ultrasonic noise, artifact and harmonic distortion present which recordings constrained to 22 kHz wouldn't have.
@@billmilosz this is what I wish you said in your first comment.
Could you imagine such an instrument that only creates tones in the inaudible spectrum? It would finally be something I could play without offending any humans.
ever heard of a dog whistle?
@@marcelwenting Lol..Now I wonder if any musicians have ever included dog whistles in their music.
Thank you. The snake oil SLAYER guiding the noobs to real music. Very informative, thank you Amir.
So I will stick CD and CD quality. If the master is not good enough (and how should one know....?) where's the sense in HIgh-Res, especially given the fact, that most of us are too old to hear above a certain frequency.....
The main aspect of high sample rates to my understanding is a better time resolution. there is no benefit in high frequency signals beyond our hearing range, but -68db Noise @160KHz should not be allowed.
The MQA snake oil degrades the audible content (just slightly) only to store inaudible HF Noise that no one can possibly hear, but there is no way to restore the better time resolution of that high sample source material. So You end up getting less quality with a slight boost in the upper freqencies and more Noise, but I can't see any benefits for that trade off.
Therefore I have quit Tidal (=MQA) and enjoy Qobuz for now....I have to say the Tidal App was more intuitive though and the MQA/HighRes Flac discussion is somewhat academic, since I could not hear much of a difference.
Time resolution is a function of bandwidth and bit depth. It has little to do with sample rate.
@@AudioScienceReview Well, maybe I got something fundamentally wrong. I try to explain what I think is going on (very simplified) and be happy to learn where I'm wrong:
To digitize a Signal you basically take a snapshot of the Level/Amplitude of that Signal at discrete time steps- the samplerate- and assign a discrete Bit value to it.
Now, with a sample rate of 44.1KHz you get a snapshot every 0.0227ms.
But You can't possibly know whats happening between these samples. Until You increase the sample rate to, lets' say 192KHz. Then the distance in time between each sample is only 0.0052ms.
That sounds to me like a much higher 'time resolution', more steps per time interval.
You say 'Time resolution is a function of Bandwidth and Bit depth' , but isn't the Bandwidth directly coupled to the sampling rate? Like for a 44,1 KHz sample rate you end up with a bandwidth of 22KHz and 192KHz gives you a bandwidth of 96KHz and so on.
I Dont know what the Bit Depth has to do with Time here. To my understanding, the bit depth of a sample defines the dynamic range, the possible discrete values btw. 0 and maximum, but how is this related to time?
Would love to learn a bit more here...thanks, and keep testing!
@ReaktorLeak Hi there!
As far as I know, when digitizing an audio Signal you need two basically independent things:
1. the Sample Rate. This defines the time steps between each measurement/sample and defines the max frequency that can be reconstructed (on a bandlimited signal according to nyqusit shannon) A higher Sample gives you more Samples per Sec, thus the time resolution is increased.
2. The Bitdepth. This is defines how many Integer Values of the signals Voltage you can have at each and every Sample Point. 1 Bit would be 0 and 1, Signal or silence - not very usefull for musik ;o) 16 Bit gives you already 65.536 different values, 24Bit a whopping 16.777.216 Levels.
These two Numbers (SampleRate and Bitdepth) are like X and Y on a graph, where Samplerate is time accuracy and Bitdepth is Signal accuracy.
Or to use an analogy from Digital imagery: the SampleRate would be the number of Pixels per Inch and the Bitdepth defines how many different colors each pixel can have.
quantization errors are indeed a 'problem' on the Bitdepth side of things (at least in theory), not on the time (or Phase).
So still, I don't see a connection between bitdepth and time resolution, but would even go as far and say Sample Rate IS time Resolution in this context.
Of course I could be totally off track here....
anyways, enjoy your music, regardless of all that digital magic ;o)
@ReaktorLeak wow, thanks for the effort! I see what you mean.
But I always imagined ( and still do to some degree ;o) the Signal to be static and the samples going forward in time, whereas you are moving the sinewave along the Time axis, while keeping the samples fixed in time...dont know if that made any sense...I try again:
In my 'model' of audio digitizing things go like this: A Signal (limited in Bandwidth) enters the AD Converter. This ADC takes a measurement of the Voltage at a given time interval - the SamplingRate. At any (Sample)point in time there is only one specific Voltage. nothing is moving throgh time. Its a very fast array of individual static measurements. As long as the clock of the ADC is accurate enough, there is - from the position of a single sample - nothing moving in time. Each single sample is somewhat 'timeless' so to say.
I Imagine this a bit like a movie: its a set of still images shot at 24Frames per sec for a traditional movie (but without the motion blur ;o) On a modern high speed Camera you get more Frames per sec = a higher Sample Rate = a better time resolution.
And all that is totally independent on the Bit depth at which the samples are stored.
I have been seeing Sample Rate and Bitdepth kind of as the QUANTITY and QUALITY of a digitized signal.
Thats at least my impression of this magic, and of course there are a lot of details left out, but i thought I got the basics right....or not?
Thanks again for diskussing this, cheers!
@ReaktorLeak I totally agree, it does not matter where (or better when) on the Waveform the samples are sitting, as the reconstructed wave remains the same. In fact, I think the temporal position will be a bit different every time you digitize (or resample) a Signal, depending on the internal clock of the converter.
Nevertheless, still I think that (at least in theory) a higher samplerate equals to a higher time resolution (the Quantity of samples) while the Bitdepth defines the precision of each of these Samples ( The Quality ).
If any of this is audible....thats another issue. Honestly, I dont think there is anything missing from the good old 44.1/16bit CD. And I dont see any point in going over 88.2 or 96KHz/24b. At least on the consumer side thats all more or less voodoo. Moving the listening posision a few inches or removing the front cover of the speakers may result in more significant changes to the audible experiance than the high res version over an cd version...But hey, more is more, so people will keep asking for more bits and samples....just because ;o)
cheers
Amir, why are you trying to bash mqa now when you were gate keeping it a couple of weeks ago?
Simple: I am neither bashing nor defending MQA. Those are simpleton characterizations people put on my views so they don't have to understand the full analysis. Don't pigeonhole me that way and you won't see any inconsistency. As with everything else I do, I arrive at topics objectively and without emotion. If something is wrong I say it. If it is not, I do the same.
From the little I have read, MQA is guesswork.
Amir, noise shaping is a dithering technique to trade-off aliasing artifacts for noise. The idea is that gaussian (or shaped) noise, being uncorrelated to the signal, is less noticeable than steady-state ("fixed pattern") signals. In high-end imaging (my line of work) it's been used for decades to alleviate Mach banding, but of late with HDR video it's fallen out of favour. We prefer to properly allocate amplitude and frequency sample (and sample precision) to match the perceptual characteristics of the human visual system. Now, when it comes to decomposing a signal into a Fourier series like you do here, bear in mind you are looking at an aspect of the signal that isn't perfectly correlated to perception : the ear doesn't do an FFT. What the ear hears, is not the ultrasonic decomposition of the waveform, but the slew rate of the waves, because it is a mechanical device. Now, the ear's ability to perceive fast wave rise times is notoriously limited, and probably to the tune of 1/f. This means that the massive amount of data your FFT is showing above 20 kHz is inaudible as pure harmonics and partials, but it packed back into the source audio signal, it contributes to its acoustical shape. Now, I agree with you that this contribution is small, and exponentially-so. This is the reason that apodizing filters are not brick-wall, and basically should roll off this "junk" gently. Now, that is silly in these audio formats is the bit budget waste. While you can't get rid of the sampling rate (you do need to preserve, say 96 kHz), the Just-Noticeable-Difference in amplitude within the ultrasonic range quite possible could be encoded with 4 bits.
Simple noise shaping is not exactly the topic of these videos. What is being emphasized is severe noise shaping required by 1-bit quantizers used in DSD format. With baseline SNR of just 6 dB, massive amount of quantization noise is generated without shaping it. Problem we have now is that people are shipping high-rate DSD which pushes the playback filter up to hundreds of kilohertz so all of this gets "played" by the system. And takes up file space. This is the point of this series of videos, i.e. how the bits are utilized. Of course noise shaping in general is a powerful technique and one that must be used for proper conversion of high bit depth uses in production to lower bit depth used in delivery (e.g. 24 to 16 bits).
Dither in video is a big problem. I remember when we were working with major studios on HD production for HD-DVD and Blu-ray that we would routinely find content that they had just truncated to 8 bits causing banding. One major studio's own animated logo at the start of every release had this problem!
As a fascinating thing, if you dither 10 bit video to get 8 bits, often you are better off using 10 bit original because 8 bit dithered content is harder to encode so winds up not being much of a saving versus just handling 10 bit.
Maybe I misunderstand you, but the ear drum isn't connected directly to the auditory nerve. Mechanical information is translated to electrical potential by the hair cells of the cochlea, which are individually tuned to react to specific frequencies in a manner indeed equivalent to FFT. Human hair cells don't react to frequencies above 20 kHz and that's that I'm afraid, just like the cone cells of your retina are incapable of reacting to the 940 nm infrared light of your TV remote.
I suppose Amir's point is that nobody in their right mind would pay extra for video content featuring light outside the visible spectrum, that information is clearly garbage, but somehow high resolution audio has a curiously large following.
@@doctorzingo You are right and I do not dispute that but I am not talking of steady state tones, I am talking of impulse response. When a transducer is excited it convolves the stimulus with “window function”, as a result it’s not an all-or-nothing situation, but one where the response falls off at a certain rate. This falloff contributes a small amount of information to the dominant signal. In human vision the shape of the spectral response of retinal cones and rods is a complex function of stimulus magnitude and contextual stimulus (called surround in color appearance models). In the spatial domain, which is the equivalent of frequency in audio, there is a difference between absolute resolving power and perceived resolution, which is a function of stereoscopic and motion perception. All I am saying is that we ought to be careful using rule-of-thumb metrics for what really is a complex dynamic phenomenon. Now I will not dispute the fact that you can get pristine audio from 16/44.1 : the model works and well-engineered equipment does a great job at analogue waveform reconstruction. I do believe there is some faint gain to be had from slightly higher sampling rates, although I’m more skeptical of higher bit depths because the ear doesn’t have a uniform JND across the audible range. I believe like we do in imaging, that one needs high frequency acquisition, high bit rate signal processing in post-production (because you quickly stack up rounding errors), but for consumer delivery, Red Code is perfectly adequate when processed by competent gear (Im using a Topping D90). In have found however that in complex classical music with dense choral textures or in piano concerti, there are fleeting (brief) moments where I obtained cleaner playback at 88 or 96 kHz.
@@phpn99 Correct me if I'm wrong, but if I understand you correctly, what you are saying is that there is an additional sound property that can be described as "quickness" or "transient response". But that is precisely what the highest frequency hair cells do - they are tuned to react to the fastest changes in cochleal fluid pressure. There is no mechanism to register even faster changes.
A 20 kHz low-pass filtered square wave sounds unfiltered to the human ear, but doesn't have an immediate rise on the oscilloscope. Or to put it another way, a perfectly rising and falling square wave requires an infinite amount of harmonics, i.e. an infinitely high frequency band.
@@doctorzingo I agree : No such impulse response exists in natural phenomena ; slew rate is limited because there is no such thing as infinite frequency spectrum (it's an issue with laser optical data transmission, for instance). The ear peaks in its ability to perceive amplitude differences well below 20 kHz, into the 5 to 10k band if I recall, above which perception falls like a brick. So the static hearing system (ear + brain) is a band-limited transducer that convolves the input with a window function. But I think there are dynamic phenomena at play in hearing, the same way that there are with the visual system : Adjacent stimuli that together (in micro-succession with audio), skew the response. I am literally splitting hairs here.. (or ciliae ;-) ) so do not take my point as argumentative to prove the same sort of point subjectivists serve Amir every day. I think higher frequency audio capture has some merits, that are definitely in the territory of diminishing returns. There is no economic or technical justification to charge more for "high-res" audio, because these data rates are pure commodity today. Charlatan tech like MQA is nothing but a ploy to milk gullible consumers; just like cables. I believe DACs and amplifiers are pretty much nailed with todays tech, and the very best can be purchased sub-1000 dollars. The remaining frontier in my opinion, definitely is loudspeakers, and this is an area where I hope to see much progress over the next decade. The ability of tweeters to accurately reproduce the sort of high-res audio content discussed here is likely the biggest argument against all these self-professed "esoteric" formats. I am curious to see what Amir will do in this regard, since he's started publishing speaker (and HP) reviews. I'd like to see inter-driver delay evaluated (and how it distorts the waveform across the crossover band), as well as the ability of a driver to trace a well-defined impulse. The reproduction error would best plotted as distortion vs frequency vs amplitude, also in an IM distortion plot. In visual systems we use a unit called the Delta-E to specify the smallest perceivable amplitude error (it's actually just a discrepancy, but it's commonly used to gauge the visibility of calibration errors). I wonder what JND unit can be used in audio to serve the same purpose. The sone ? Thanks for the discussion; always feels good to geek things out. Cheers.
Hirez Audio tries to fix problems, that are not really there.
As I understand all this debate, MQA is more about deblurring and cleaning up the temporal info with a surgical approach, and then pouring out all the dirty water. This radical approach has had some side effects. Meanwhile, upsampling can also achieve deblurring by adding, as it were, more and more water to dilute the polluted teomporal info and leading to ever larger volumes.
I say this as a DSP software engineer- You have absolutely no idea what you’re talking about.
Jealous of the 2L studio... Genelec monitors EVERYWHERE!
This is awesome Amir, however it is really nitpicking. I tried to reproduce this and I could not because 1) not all spectrum analyzers can be set to linear scale 2) not all spectrum analyzers can be set to -150db and beyond. So mastering engineers are not seeing this because they usually use the log scale and the volume is so low - practically inaudible - so noone is there to listen for that.
There are free tools that show all of this. And certainly all pro apps can do all of this so I see no excuse for them not being addressed.
@@AudioScienceReview I tried with FOSS tools like Audacity and various free VST plugins (e.g. Voxengo SPAN, Blue Cat's FreqAnalyst) but did not manage. You are right though pro apps should be able to do this.
Just more evidence that 24/44.1 or 24/48 should be the maximum format for all music
No, they are crap because producers often are not educated on this topic, some will use low cost gear.
The advantages of high-resolution are greater bandwidth, timing resolution (left and right ear can detect 15 micro-second differences, which means at least 66.7kHz sampling is required), reduced filtering demands (the slope of the filter can be gentler thus avoiding phase shift problems seen with steep filters needed for low resolution), greater dynamic range and lower noise (144dB). But in order to reap these benefits you need a very high level system including the source, streamer, DAC, preamp, amp, and speakers, otherwise you will experience increased distortion that is higher than with low resolution recordings, intermodulation distortion in particular. IMD is a common problem at ultrasonic frequencies: if your amplifier and tweeters are not designed to handle these ultrasonic frequencies properly there will be significant IMD distortion that can be heard in the lower frequencies as well. For example, if the high-res recording contains frequencies at 28kHz and 30kHz a subpar amp and/or speakers will reproduce not only 28kHz and 30kHz tones but also the sum and difference of these tones. The difference will occur at 2000Hz right in the middle of the frequency range where the human ear is most sensitive. Low quality preamps and amps filter out ultrasonic frequencies in order to avoid oscillation and other problems, but this filtering is far from perfect.
Tape Bias would be higher than 40K on any professional deck.
Yeh, I am just guessing here without knowing exactly what they were using in these recordings.
Musical performances and music instruments produce much more higher frequencies than what we can ear. What you suggest to do is butchering these tracks. On the first graph, the part on far right can clearly be seen moving up and down in correlation to the music
@ReaktorLeak And neuroscientists know better
@ReaktorLeak So, you can't be sure that frequencies beyond 20000hz don't matter, since there is no scientific consensus.
Besides, those sounds are present in live music. Real instruments produce such harmonics. Even if it is not directly audible, there is no reason not to record the whole signal. Nothing is clipped in the analog domain. Digital should always try to replicate analog performances in the best way possible.
@ReaktorLeak And how do you propose to prove such a thing?
@ReaktorLeak I have only listened to each sample once, but I would say that #5 was the more detailed, followed by #1, #4 and #2 and #3 don't sound as good to my ears.
But they could as well be completely identical because the difference I hear is very subtle
@ReaktorLeak Interesting indeed. I have to confess that I converted the samples to DSD prior to listening. In PCM, it's nearly impossible to spot any difference at all
Amir, you should write a book or an e-book, atleast
The title should be "How I copied someone's findings after banning and ostracizing the guy for being right all along" although I don't think that's very catchy
@@Cystic_Fibrosis
There seems to be a backstory which I am not aware of.
I wrote a book once; it killed me. :) So hard for me to write things in text and get all the grammar, etc. right. Whereas in a video like this, I can just talk.
@@AudioScienceReview
But you already write the same material for the website. Besides, there is no material out there catering to the audio customer. It's all for the audio professional.
Also, the audio community is full of snake oil, with all sorts of shamanic wisdom. After few years, now I realise how simple the basic principles are!
What about anti time smearing benefits of MQA? How do you measure that by looking at these graphs? You’re taking a very narrow view by just looking at the waveform and frequencies
This analysis has to do with modifications MQA does to the baseband audio. What it does post decoding requires a different type of analysis that I may do one day.
@@AudioScienceReview thanks I would be interested in your analysis. I think you should make it clearer in your videos what exactly you are analysing and what benefits there may be which would require a different type of analysis
@@AudioScienceReview so just thinking on this if I understand correctly, your analysis here is basically what MQA encoded files look like without any decoding correct? Since I don’t imagine Adobe can even decode MQA. I think 90% of people posting negatively on MQA now in response to your video think you were looking at the decoded waveform.
Man you really need to be clearer about this stuff hey
So first you ban your forum member who did measurements on MQA and then you release your own... Coincidence?
can i have details ? im genuinely interested
@@dizzze GoldenSound (aka GoldenOne) did a full breakdown on MQA issues, he then got attacked by Amir on audio science forum (where Amir explanations looked shady, to say the least) and then when he defended/explained his methodology etc and running the thread he made he was banned. That's a BIG tl;dr
@@skedra7079 holy moly i didn't know that thanks, i saw the video and even if i had some hiccups with some of his tests i found the overall methodology really solid. Amir if you read this could you please explain what you had against goldensound methodology exactly ?
@@skedra7079 thx for the answer btw, have a great day
@@dizzze the thread was 100 pages long so it's a bit tricky to summarise. But if I were to give a short summary:
- I made a video where I'd gotten tracks published on tidal, and demonstrated various issues with mqa. You can see this on my channel. I also made a written post version on Asr.
- In that thread Amir made some statements which many saw as a defense/support of mqa, and contradicted his previous approach to analysis and reviews. This confused and angered a few people including some long time asr members/supporters.
- A while later, one of the asr moderators asked to have a phone call with me. They said that they wanted to lock the thread and for me to create a new one. I asked why, and they said that it was because they needed people to stop quoting/criticising amir's mqa-friendly posts. And that in the new thread no one would be allowed to criticise what Amir had previously said about mqa.
(guessing they wanted me to make the new thread so it would look reasonable/OK. Rather than them doing it themselves and having it look suspicious)
- I told them I was unwilling to help participate in censorship and that if you do not want to be criticised, you should not make controversial statements.
- A while later in another thread I was discussing my approach to reviews in what I thought was quite a civilised manner. You can read that thread here: www.audiosciencereview.com/forum/index.php?threads/do-objective-youtube-reviewers-exist.23019/
At about page 4, I saw that I had been banned for "disagreeing with management decisions" and apparently some of my posts have been removed too
- Amir then releases this video
This is also a fantastic example of the capitalist growth dogma.
Even if humanity achieved "audio transparency" 4 decades ago you just HAVE to "innovate" and bring something new to sell it.
All those precious ressources wasted for AC conditioners, audiophile network switches, the 500th iteration of the same amplifier design, redundant file formats etc.
Ironic that all the time he discusses high quality audio, his own voice is distorted.
ummmmmmmmmmm, please use a script going forward.
Doing yet another 180 and having the audacity to try and latch on to the hard work someone else had put in after banning that person from asr for disagreeing with you?
I wonder how low you can go...
There is no 180. This video was produced in 2017, not today. Here is the original: ua-cam.com/video/scAlD61GQMo/v-deo.html&ab_channel=AmirMajidimehr
It is the few of you who are confused about what my position is on this technology. You make up your own stereotypes devoid of facts and then complain when shown to be wrong? Spend a bit of time researching, asking questions, understanding the topic and you won't be in this pickle of making such an incorrect comment.
it's not music it doesn't count:)
I don't trust an audio analysis from someone who can't record his voice without it massively clipping... and apparently not noticing.
MQA can't work. It's method it calls audio origami violates the Shannon Nyquist criteria by 400 percent. You can't hear above 20 khz anyway. It's a solution that doesn't work for a problem that doesn't exist.
RBCD standards exceed every engineering criteria required of it. It stands out as the only achievement that has impressed me in this industry in the last 65 years. Its belt and suspenders redundancies in detecting and correcting errors is nothing short of remarkable. It offers over 64,000 loudness levels over a 96 db range and can reliably switch between any two of them within 1/44,100 of a second. Even if you beat it the difference will not be audible as the result of improved performance in the digital domain. Ask Dr. Mark Waldrep of Aix recordings on his site Real HD audio. He used a fair test on over 600 subjects and to his chagrin he concluded I was right and he had been wrong. He made true HD recordings. Down converting them to 16 / 44.1 made no audible difference. Methods that use different DACs for comparison are not fair tests.