Can lossy digital audio be better than lossless?
HTML-код
- Опубликовано: 26 авг 2023
- Data compressed audio trashes 90% of the original recording so surely it's worse than lossless audio. But what if lossy, data-compressed, audio could be better than lossless?
EQUIPMENT USED TO MAKE AUDIO MASTERCLASS VIDEOS
CAMERA - Sony Alpha A6600 amzn.to/3uj7Dtq
COMPUTER - Apple Mac Mini M2 Pro amzn.to/47tW6Wg
MICROPHONE - Sennheiser KH 416 amzn.to/3QWWvuO
ALTERNATIVE MICROPHONE - AKG C414 amzn.to/3SLqCad
BUDGET MICROPHONE CHOICE - Rode shotgun amzn.to/3G0ni36
MICROPHONE PREAMPLIFIER - Comica LinkFlex AD2 amzn.to/40SMHFT
TELEPROMPTER - Padcaster Parrot alternatives amzn.to/40A55TJ
TELEPROMPTER SOFTWARE - PromptSmart Pro promptsmart.com/
AUDIO INTERFACE - Behringer UMC202HD amzn.to/3ugj6db
HEADPHONES - Sony wired amzn.to/3sEjlht
KEYLIGHT- Amaran 100d amzn.to/40A5Gor
FILL LIGHTS - Switti RGB amzn.to/3ufhOPu
SOFTBOX Phot-R alternatives amzn.to/3QGoTjJ
LIGHTING STANDS Phot-R alternatives amzn.to/3QWXOKe
MONITOR - Samsung UJ59 amzn.to/3MEoBsk
MONITOR ARM - Invision alternatives amzn.to/40SLqi5
MIC ARM - Neewer amzn.to/49Ob0sD
TRIPOD - Velbon alternatives amzn.to/46gu2oa
GREEN SCREEN - Elgato alternatives amzn.to/3SG2F3L
DESK - FlexiSpot E7 Pro amzn.to/3QVIWvS
VIDEO EDITING - Apple Final Cut Pro www.apple.com/final-cut-pro/
IMAGE EDITING - Adobe Photoshop amzn.to/3FZ7Ww0
AUDIO EDITING - Avid Pro Tools amzn.to/3QA5Wz7
NOTEBOOK - Leuchturm 1917 hardback amzn.to/47zonKX
PEN - Sakura Pigma archival pen Sakura Pigma archival pen amzn.to/47zz3cP
OSCILLOSCOPE - Hameg HM 203 alternatives amzn.to/47tWN1O
BEAUTY - Philips 3000 nose hair trimmer amzn.to/40BdLJo
Disclaimer: Videos on this channel may include affiliate links and may feature software that has been received free of charge with an NFR (not for resale) licence, equipment that has been loaned or donated by a manufacturer or distributer, or equipment that has been bought for less than full retail price.
I do know if I want my CDs to sound like vinyl I place a bowl of Rice Krispies between the speakers and my listening position. 😃
Coco Pops work fine too.
You obviously haven't listened to a modern turntable and quality pressings. I thought I could never listen to vinyl again, but my late uncle left me his collection, plus the new records I have purchased have changed my mind. I'm not going to pretend it sounds better than digital, but it definitely has a quality that I find very pleasing.
You need two bowls if you're listening in stereo!!
@@ac81017Do you put the Rice Krispies on the left and the Coco Pops on the right, or is it the other way around?
If you guys have records that sound that bad, then you’re not taking care of them properly.
Mastering is far more important to sound quality than lossy/lossless comparisons. I've never met anyone who can tell the difference between lossless and lossy at 320kbps (MP3) or 256kbps (AAC) when tested under scientific conditions, regardless of the quality of "the system". Encoders are now so good now that even the "difficult samples" are encoded to a quality that hardly anyone can distinguish.
Arguments on this are really just a defence of the people who bulk bought the snake oil. My take is that people should spend more time enjoying the music, and less time worrying about it.
I read an article stating exactly this like 15 years ago! So I guess humans are bad at learning?
some of us want the best audio quality wether you like it or not, my passion isn't just music, it is sound in general.
yet you can sit with your DAW and flick the switch to hear what you lose when you encode MP3 and it's NOT insignificant and you CAN hear it, therefore the difference is by definition, audible. Fact. The tests you refer to claim to make a complicated comparison simple, it isn't simple. It's age dependent, hardware dependent, source material dependent, experience dependent. So given that I can easily, it's not a small amount, hear the audio that's lost from an MP3 we can conclude that lossy is worse. I regularly buy FLAC instead of MP3 and the air and imaging in the high end is different
"My take is that people should spend more time enjoying the music, and less time worrying about it."
Why tune into this channel then, shouldn't you be elsewhere 'enjoying the music' and not mixing it with the plebs?
I think people should be fined for the wilful abuse of platitudes online.
@@Simbosan
I agree with you, I too can tell the difference between typical .mp3 recordings and better formats such as .wav and .flac.
If you know what you are listening for it isn't difficult to hear the differences (think -"hollow-ness").
This is again the reason that for archival purposes I do not use "lossy" formats.
I may however choose to compress things into .mp3 when making up play lists and media for our vehicles.
This is because my goal there is to store a lot of music on a relatively small memory stick.
In typical vehicles (or ours at least), with all the other noises and distractions creeping in, what's the point of anything much better than .mp3?
Not so when listening to my "hi-fi" system at home though.
There I want the absolute best I can manage, within reason of course.
So while I do want the best audio quality I can achieve in any and all situations (including computer listening), I also try to be pragmatic in my approaches toward them.
As you said- "age dependent, hardware dependent, source material dependent, experience dependent" (and environment/situation dependent). 🙂
I have a highly resolving system with large full range electrostatic speakers. The recording quality is very obvious through these speakers. I have tried most streaming services but Spotify Premium at 320KB/s is indistinguishable from so called lossless. If you listen intently for a difference then you will hear it. This is how the brain works. The quality of the original recording is far more important than the distribution format. In a home system room acoustics and loudspeaker choice have a huge effect on the listening experience.
Apple Music is almost 95% lossless with their streaming. I have all of my music ripped at 320 kbps and it’s synced to my iCloud, and streaming my library from it has it pop up as lossless for the majority of what I own.
I do have a lot of albums that aren’t on Apple Music (2 were tape cassettes but I had my composer friend, that used to work for the same label the cassettes were from, rip the cassette in lossless so the quality is immaculate) and they sound very outstanding. I stream my music on my HomePod Gen 2’s and HomePod Minis, and I’m way more satisfied with my results of music.
Actually if you stream spotify from android there is bug that sometimes it plays loosy, same with deezer which is annoying and you can hear it. Either way I use apple music and I like it
Brilliant idea my man, I couldn't agree with you more! Just give us the highest resolution possible the most important 10%, and dither it (with noise shaping of course). I'm ALL in.
There’s another issue with lossy coding and that’s generational loss. Run audio through 3-4 generations of lossy codecs and the effects caused by that lost 90% (ok 89% to keep Betty happy) soon become apparent. This is an issue for broadcasters and studios as well as anyone who isn’t disciplined with their audio management. The main reason to use lossy audio is to save storage or bandwidth but these days these are cheap and plentiful so lossy codecs are becoming less important. I’m curious therefore why we need to represent audio in small files when linear lossless does it just fine. If space is at a premium then use FLAC or ALAC both of which are lossless and about 50% of the size of the equivalent linear file.
MP3 originally came about to allow the little players to accommodate more files. So back then it did matter.
Today, we can think about the million song repositories on various streaming services. Stored as WAV or even FLAC files they're going to occupy a goodly number of terabytes of drive space. Using AAC or MP3 formats probably reduces that by more than 50% .... Which in a massive server farm is and amazing gain in space, power consumption, heat and transfer efficiency.
Remember, these mediums are storing disk files and making them available on demand... they are not re-encoding them every time. You can transfer the same file out an infinite number of times with zero degradation.
I can definitely hear the difference between Spotify and Tidal and music purchased from HD Tracks. But the difference is not as big as many audiophiles would like to believe. Some people like darko have no use for high rez whatsoever. The quality of the recording is really the ballgame. Compression is the death star. So many modern recordings just have nothing to give no matter what.
Darko just released a video on this very topic.
The popularity of vinyl says it all. For me the main difference is vinyl is mastered much better most digital is bricked, and that's where a phenomenal medium is spoiled by moronic and corporate misuse.
Indeed ... it is truly ironic that the first storage medium capable of encompassing the full frequency and dynamics of a good performance is being used to distribute the least.
A crappy recording will still sound bad lossless. A great recordinbg will still sound great lossy. So there are more factors in play.
I think the Nyquist-Shannon sampling theorem says that lossless 44kHz PCM is, in part, already what David proposes. The theorem states an analog, 22kHz band-limited signal (generally encompassing human hearing) can be PERFECTLY reconstructed from a 44kHz sampling.
Exactly 👍
Just a couple hours ago, I listened to an old 96Kbps mp3 rip of a beloved album on some new Audio Technica earbuds... my FLAC and WAV files sound phenomenal on these, but this album sounded flat, with no separation between instruments, and the electric guitar and bass distorted when they were playes together. Those are somewhat extreme cases to contrast, but it's where my mind is. Thanks for the ever incisive videos!
"old 96kbit mp3". that's the problem old codecs sound very bad. it's not just the bitrate. the codecs have improved a lot, to decide what data to delete and what data to keep or even imrpove. You will not here a difference between FLAC and modern AAC.
Stuck with mostly listening to you tube on my phone due to unfortunate circumstances. So I built my latest system out of the aftermath to sound good that way and enjoy the music. Also thanks for the uploads out there and yes many sound pretty good. Even get to see some nice tables and cartridge. It is noticeable.
I once worked (not in a technical capacity) on the launch of Radio 3's 320K AAC online streams (we called it 'HD Audio' at the time and trialled it with the Proms in 2013). It was very exciting. We put on listening sessions in a studio in BH with an audiophile set-up. Journalists came in and were played various versions of Radio 3 output through a lovely pair of monitors and not one of them could tell the difference between the 320K AAC streams offered online and uncompressed (in either sense) audio straight from the desk in the next room. The rows in the comments were heated and always fascinating - and many people continued to insist that the existing FM signal must be better…
Even though I'm rapidly approaching 70, my hearing is way better than it should be. I have good sensitivity ( I notice noises my wife who is 12 years younger can't) and I can still hear up to about 13 khz. My fine discrimination is not what it once was though. I can still tell almost any 128 kbit MP3 from a CD, but I'd be hard pressed to tell a 256 or 320 kbit MP3 from a CD. Low bitrates affect some genres of music to a greater or lesser degree depending.
Bit depth affects things way more but glad your ears are doing well!
As long as nothing audible, which means noticeable, is going on I truly think life's too short to be fussing about theoretical improvements in sound quality you might get by demanding hi-res for everything. I grew up with cassette-walkmans. Good ones sounded great, and I have a true fondness for every discarded one of them that has never been matched by any digital format since.
Thanks for the video. You forgot to mention that some people WOULD care about what they are missing IF they knew they were missing it. Your statement didn't divide me into a camp, but it did make me lol. Fire those assistants...
Those assistants (let's call them "handlers") do have some redeeming attributes. They remind me of the ever present (and never present) Samantha of "I'm Sorry, I haven't A Clue" BBC Radio Four fame.
High definition tracks on RUclips sound amazing. That just proves that it's the the actual recording itself which is the limiting factor.
Stand up and take a bow that man for telling it as it is rather than the load of bollocks we often hear from so called audiophiles 👏👏👏
Sounds like you don't even know the meaning of the word AUDIOPHILE Mr Gobshite. 🙂
Actually, the 44.1kHz/16bit standard is perfectly adequate for the final artifact (not for mixing). Some loss is already there in the form of dynamic range (bits) and frequency representation, but when carefully mastered we should never be able to hear that.
Some modern digital formats as used via Bluetooth have very audible degradation, my noise-canceling set fortunately also connects using a USB cable, and the difference is immense.
The biggest improvement i made was cable lifters. It was night and day ! The more expensive ones obviously sounded the best.
Do you need to feed them regularly, and what do you do when they're on holiday?
😂
Interesting idea to focus on what matters
Btw I always find the word "psychoacoustic" funny, because it actually pictured how "audiophile" works.
Lossy digital audio played on a very well set up system can sound better than lossless digital on a badly set up system. Imagine what the sound would be like in a room with lots of glass walls and marble floor. Bob Katz (mastering engineer) feels that higher resolution audio (better than CD quality) provides better sound stage aspects to regular CD quality. Since compression happens in the frequency domain, it keeps more of the tonal aspects but loses in the timing aspect. ie. More smearing. The more compression, the more smearing. Bad rooms have more sound reflection off walls, which also causes more smearing. This impacts the aliveness of the sound. ie. Does the instrument sound like it is in the room, or does it sound like a recording. The goal of very good systems is to disappear so that you can't hear the equipment, and your focus is only on the music floating in the air.
I think it's very possible to have better audio where unneeded elements are removed. Especially now, that we have tools to measure quality in ways that we didn't before. Aside from locally anecdotal evidence ("I don't like it" or \\"Well, I do!", etc.), we have the ability to send various versions of audio out in to the wild, and have the number of streams, downloads, replays, purchases, and probably other data I can't think of be sliced and diced and reassembled into a cogent picture of preference. I feel certain that your staff there would agree. 😀
It's an interesting thought experiment. I'm going to have to say lossless is always better than lossy. I would argue that if you are focusing only on ADC converting the 10% we care about, you're wasting some (most) of the resulting bit stream. The bit stream (eg 96/24) can carry all the detail with the same ease as only carrying the 10% that our acoustic systems can detect.
There's an old saying, "It is better to have it and not need it, than to need it and not have it."
Resolution can be a novel idea, but, jokes aside, this just makes me ponder my own hearing, and/or lack thereof. I ran a sine wave the other day and was sad to listen to it drop off earlier than it used to. I don't feel I'm missing out when I enjoy music, but I'm clearly not hearing as much of it as I used to.
Such is life (aging). I wear hearing aids that can only afford me up to 8kHz, but I can still enjoy music and can judge one speaker from another within my limited envelope. I do suffer some tinnitus as well, but that doesn’t prevent me from splurging on nice toys 😉
@@billd9667 Another music channel recently mentioned that he knew a guy who, after years of touring without hearing protection, couldn't hear anything north of 2k. But the dude still loves music; a reminder to be grateful for what you have... and enjoy your nice toys! :)
Funny you mentioned cameras and lenses, because there is an analogy there. A high resolution lens on a high resolution camera can cause aliasing issues; which is why most digital cameras have an anti-aliasing (low-pass) filter in front of the sensor. In essence, the image is blurred a tiny little bit _on_ _purpose_ to avoid aliasing problems. In absence of this filter you could get distracting Moiré patterns appearing in the images. In theory a lossy audio format could serve a similar (though not identical, i.e. not necessarily anti-aliasing) purpose if the source material has issues which would otherwise become an audible problem when processed into a lossless format.
@4:02 "Any conversion to digital is going to be less than perfect", Does that include the digital conversion/processing inside of Our brains, post semicircular canals and vestibulocochlear nerve that caries the now digitized signal? I do agree with you that We don't require ALL the information that exists in the," Analog", domain, outside the brain reality. An example, very small transducers in our laptops/phones cannot reproduce the frequencies of musical Fundamentals but do reproduce the Harmonics, so our brains can fill in the missing data of the Fundamental so that we can sense the Pitch, as the Harmonics yield Timbral information. Yet on systems that can reproduce the Fundamentals, that is where the Hi Fi in the experience comes into play making the experience quite thrilling.
If "Hi-Rez" audio was really that good, why is it that whenever I stream just about any music I've enjoyed for the last 50-odd years in anything over CD quality, it sounds (or maybe "feels"?) more like nails across a chalkboard? I play those same songs straight from either CD or Vinyl and they sound just fine, with perhaps CD getting the every so slight edge, perhaps more because of the absence of the clicks and pops, which I sometimes don't even hear on a well-pressed album. I do realize that it's probably because my house costs me more than my stereo system OR my cables.......or probably not. I don't have a crazy stupid investment in hearing things that probably aren't really there.
I have the impression that what's removed with the algorithm is some of the room acoustics, reverb, so that the instruments breathe and you get a sense of the room they were recorded in. That's something that I don't want to lose.
I don't think the algorithm is smart enough to decide to remove a single instrument or room quality ... it's based on levels, audibility, so if the instruments breathing would be audible it would be preserved.
I fully expect to tune into the digital chump one day and find his two weirdo assistants have done him in and produced their own video.
The dead giveaway I am listening to something lossy, is after 30mins when my ears feel sore and fatigued, which NEVER happens when I listen to lossless formats. The other giveaway is if the music is more acoustic and complex in production and recording. If the music is using a lot of compressed synthesised instruments and samples, it becomes much harder to tell the difference, although the sore ears still happens after 30mins.
I suspect the real difference is the EQ and dynamic range compression. All those recording with electronic instruments probably really emphasise some narrow frequency band and are full of dynamic range compression. That kind of thing is bound to tire the ears rapidly.
@@martineyles I think it is the consequences of what is removed from the audio, to achieve the small file sizes. No matter what the style of production or mixing, lossy always makes my ears sore after 30mins. This can happen with certain types of lossless mix too, but not with musically sympathetic mixes. Sometimes due to EQ and range compression, and always when lossy.
That would be a niche market, so small, that no one would develop.
Mp3 sounds different from the CD format. Just listen to the cymbals. The degradation is quite obvious when listening to metal or rock. The negative effects can also be heard very clearly on tidal at low audio quality(320kbps/). A few years ago there was a popular blind test online to identify formats. Despite the low representativeness of the samples(the crappy mainstream genres themselves) I was able to get scores of 5/6 after a while of ear training and 6/6 the second time, which shows that the difference is audible. With the fact that I was testing with good quality headphones(akg k701), on distant speakers the differences are less noticeable.
I can detect a difference between mp3 and CD quality, but streaming above CD quality adds nothing to me.
Can you hear the throw-away? Depends on the type of music and a few simple acoustic parameters of your room and monitors. A 90% throw away does not impact a wide swath of rock/pop. Even audible differences become just another flavor of the production and who can say what's "right" unless you A-B compare. But take a properly recorded and mixed orchestral/choral work and the /complete/ loss of depth -- the front-to-back flattening in the lossy encoded sound -- can make you weep.
To hear that depth you need a room reasonably free of time-smearing early reflections hiding those subtle details that provide depth cues in the first place. (A lot of audiophiles do *not* have properly treated rooms. The overall clarity of their mega-buck system is diminished by room problems. In those settings listeners might indeed be "happy enough" with lossy encoding because they've never clearly heard the information that provides dimension.)
In studio mastering I use a tool that lets you hear what a lossy CODEC is throwing away (different bit rates of MP3 and AAC). We'd hope it'd just be noise that's tossed but especially at lower bit rates the amount of actual music getting removed is shocking. Even 320 KBpS falls short. And FLAC or SLAC? Nope, not lossless, contrary to marketing claims. Only a little better than 320 MP3s. This is all easy to hear on even a modestly better system but in a properly-designed room.
These days storage is cheap. Don't lossy encode. Don't be suckered by CODECs claiming lossless. Even better-produced rock/pop is worth the small incremental cost of storage and deserves the better sound quality. And see what can be done to improve your listening environment.
Saying that mp3 throws away 90% of data is like saying that flac (lossless compression) throws away 30% or 40% of data because it uses less storage memory.
The assistants
Betty & Debbie should wrestle ala Duran Duran’s “Girls on Film”. That solves lossy vs lossless
have you ever subtracted an MP3 from a wav to see what the difference is? quite the eye-opener. most of it was wisps of HF energy & tiny phase errors, from what I could see.
I've just done the exercise again here, with a piece of music I recorded & mixed at 44.1/16 (it's a demo, ok? 😀) & of which I had a corresponding MP3, as I'd emailed it to some folks.
the remainder, after subtracting the MP3 by phase inversion, is broadband noise that never peaks above -70dB, but it's difficult to tell if this is information *lost* from the wav or just moved a bit by phase errors in the transformation. reaper & max, since you ask, & a natty tascam 16x8 on my imac.
LDAC Bluetooth codec basically does what you're proposing. It can upsample 320kbp mp3 files to basically sound high-res. The only hangup is that LDAC can stream up to 990kbps, but the converted 32khz LDAC file needs thousands of bits more per second to send than what the container can hold. For comparison CD quality is about 1400kbps, which is also higher than LDAC's highest transfer rate. So hopefully what you're hearing is identical to what you're describing.
LDAC is, as you said, just yet another codec. It does not anything to the PCM stream decoded from any of your lossy audio files. It's just that the codec itself was designed to support 96KHz input, and often on devices that aren't very "configuring-friendly", the policy is to process/mix sound at the highest supported down the line (to please consumer or "audiophiles"), so the consequence is that 44.1/48KHz stream has to be resampled to 96KHz before it gets (re-)encoded by the LDAC encoder (i.e., it's essentially "SRC", which is something that at least audiophile consider to be really "damaging" to audio quality). Audio resamplers aren't even designed to interpolate so that the output stream will sound better. (Or, that's just not how digital audio works.) In other words, it doesn't even claim to offer anything e.g. MQA or "reconstruction for lossy" tech in the older days claims to do.
@@tomyan3392 well that's the thing, LDAC is not like other codecs, it uses a mixture of advanced decoding methods to make the music sound better. There was a conference where Sony's team was criticized to no end because they claimed LDAC could make a 256 file sound hi-res though LDAC. I believe this code does feature some level of interpolation that we do not see in other codecs like APTXhd. So you're right in a sense that nothing is added, yet we still get more due to higher sampling rates.
In the late 1990s and early 2000s I spent a fair bit of effort transcoding my vinyl collection to MP3/192 files (the best we had at the time). After some fiddling with levels and a couple of different softwares, I had a setup that would let me do needle drop transfers that were sonically indistinguishable from the original vinyl. (clicks, pops, and all)
And before you jump on me, yes, I did blind-AB test them with the help of a friend. In almost every case the disk file and original vinyl were close enough that I had to look to see which was actually playing at the moment. The result was about the same when my friend took his turn.
Today, 20 years later, this collection exists in the form of a curated collection of MP3 files all "declicked" and normalized to the now standard of -16 lufs for streaming audio and there's no concern about "what was removed' as I still enjoy it daily.
Is this better than vinyl?
Well, if you consider that I reduced an entire wall of albums that would eventually deteriorate and sound awful to about 30gb on a hard drive that will retain it's sonic character forever... yeah, it's better.
Then, of course we have to consider the money made by selling off the vinyl collection.... Way ahead on that one!
The only problem I have with this is, that 90% was redundant data. The encoder catalogs the data to save it in a smaller size.
I like Betty's explanation. 😀
It added a great deal of clarity to the subject.
They are wrong about him not being funny though, IMHO.
Personally I tend to prefer "lossless" formats for archival purposes simply because I **might** want (or be able) to retrieve or otherwise process the lost information at a later date.
If that information is not present to start with, later processing or enhancements become impossible.
I also really appreciate the observation that the ability of our brains/ears/etcetera to process what we are actually hearing varies from person to person.
This might be a fair part of the reason some people can "hear" minute things in the audio stream that other people cannot appreciate.
If most people cannot process 90% of it anyway, for them having 100% of it becomes completely irrelevant.
Back to the $49.95 versus $4999.95 ears concept when evaluating equipment perhaps?
Or, perhaps they actually CAN hear the very same things but simply don't care.
It's just a thought.
I also agree with the comments of others that the mastering process (good, bad or ugly) is THE most important part of the process.
So- If that was done badly and the information simply isn't "there" to begin with, you cannot re-create it at the listening end no matter how much one may spend on interconnects, magic cartridges, 20kg turntable platters or $20k/channel amplifiers (not to mention speakers and used-car-price speaker cables...).
🤣
It is my understanding that most of that 90% you can't hear is background hiss, along with some very quiet upper harmonics, all in the range of 16-20k. So even though it's 90 % of the sound, it's not even 0.1% of the *music*. Digital doesn't know the difference between low-level random noise and a ripping flugelhorn solo. It's all the same 1s and 0s. But most of the "information" on any recording is actually that low-volume hiss which you cannot hear or which you otherwise ignore.
Most of that extra resolution is high frequencies that are either beyond the range of someone's hearing, very low amplitude, musically unimportant and/or unpleasant.
It's Paul McCartney!
When it comes to digital audio, total system latency is far more important than any form of compression or lack there of.
Too much latency has your transients literally bumping into each other creating the audible effect of a muddy sound and a less defined image.
This is actually the main reason why lossless is better.
Not because it is higher resolving or has more of the original data, but because it is lower latency because of the lack of processing overhead.
The less clock cycle translations your music file has to make before the DAC the better.
This is the reason why I have my entire library in WAV form.
WAV form is completely uncompressed audio, which means it is a direct bit-stream requiring zero input from processors.
I have tried even lossless compression like FLAC, but I can still hear the difference the processor interference makes.
I feel like I'm into audio, but not an audiophile because I can't tell a consistent difference between 320kbps MP3 or 256kbps AAC. Even then, the differences are minor enough that I don't really mind. I'd rather spend money on good gear to listen to what I have in the best way possible. It's one of the biggest reasons why I'm not out buying 1k iems, headphones, or multi thousands of dollar speaker sets
I'm curious about how transients are affected when converted to digital. I would assume there is some loss from [file] compression. But professional recordings are already compressed [dynamic range].. To determine if this is true, you'd need to take a "pure" track that hasn't been run through any filters, and convert it to a compressed file format. And then compare wave forms. I'm not an audiophile....but this is one possible difference you did not seem to mention? (Defining what could be different, and seeing if it is indeed different...)
Unfortunately, the wave form itself will not show what you are looking for (assuming a correctly implemented CODEC). You may in fact think that the decompressed file is better because e.g. you don't see clipping that (you think) is present in the original waveform - but that 'clipping' may not actually be there after the waveform is reconstructed in the DA chip!
@@diatonicdelirium1743 I'm not wanting or expecting to find anything here. Just asking the question. When I draw a curved line in CAD, and then simplify it, details like sharp points are lost. I didn't know if that would also be the case with audio recordings
@@fredygump5578 OK, the 'waveform' that is represented by the 44.1kHz/16bit format is actually a set of samples situated on the continuous line that would be the actual waveform. Using Nyquist's theorem we may assume that only real frequencies up to 22050Hz are in the data, and this allows us (the software or DA chip) to reconstruct the continuous line that represent all those frequencies.
If I use other mathematical means to represent those frequencies (not the line!), I may be able to get very close to those original frequencies as well, even to the point that they really sound the same to our (imperfect) ears. Some frequencies drown out against others, and thus the corresponding waveform may look very different!
All of this is a set of sophisticated psycho acoustic models to determine what things drown out, and some relatively basic encoding schemes that allow a nice stack of course-->refining layers to reconstruct the audio. Peel off refining layers to get a smaller file, at the cost of losing that refinement.
For starters, it matters a lot what you listen to, whether you can hear the difference between say CD or mp3 128kbs. Due to the current loudness war with flat compressed masters, the difference is almost impossible to hear. But with masters that are not flat compressed, I can hear it with intense reverb and other sounds with a lot of noise in it. Noise contains a lot of frequencies and the audia date reduction algorithm has a hard time determining what to keep and what to discard. Such sounds often sound clearly different in mp3 128kps. If you don't know what the original sounds like, you may not be able to hear it. If the sound changes continuously, the audia date reduction algorithm can go wrong so that it is very clearly audible as a kind of phasing / flanger effect.
I don't think the idea of sampling at a very high resolution is new. That is used for the super audio CD; one bit with a very high sampling frequency and a lossless data compression is used. I immediately want to believe that super audio CDs sound much better than "regular" CDs (provided the recording, mix and master are really good) but it seems that the world is not waiting for it and that the super audio CD has a silent dead died.
But in the end, the quality of the recording, mix and master, in combination with the headphones or loudspeakers and the acoustics of the listening environment, has many times more impact than the difference between Super Audio CD, CD or MP3 128 kbs.
The main problem with SACD is not the format, but how it is used. Sadly a lot of discs are just lazy CD masters put on SACD to make a quick buck. However the discs that are created from the original pre-format conversion master source material, they tend to sound amazing with a super smooth detailed sound that is almost analogue in nature.
Psychoacoustic effects depend on the perceiver, and how they learned to perceive. Not all people will be happy with the same 90% discarded.
Okay, I subscribed. 👍
Welcome Grand Duke. Unfortunately there is no red carpet. DM
@@AudioMasterclass I am inclined to accept the fact and an apology this time, but only if the red carpet is reserved for the mighty audiophiles. 😁
One of the main reasons MP3 became popular was that the storage chips were quite expensive in the old days and young People just did not have the money to buy larger storage (hard drive /computer).MP 3 Players were reasonably priced particularly when compared with today's DAP players. Today storage chips are cheap and available in different storage sizes. As I grew up with LP and later with cassettes &Reel to Reel and other formats I never was interested in MP3 or any other compressed format. Well-recorded Music with a wide dynamic range did sound better to me uncompressed at least for me. Unfortunately, today's music has (mostly) less Dynamic range it's louder and makes you faster listing fatigue. Streaming music is easier with MP3 etc. than uncompressed music. But I fully understand particularly the young crowd using compressed files is a faster download and can store lots of songs on a small hard drive or MP3 player without spending too much Money. Every Music format has its own market otherwise it would not exist
In My DAC, sound system it seems that 192kps mp3 sounds better that 256kps. Am I going insane here? Anyone else noticed this? Are certain DACS 'voiced' for certain lossy files?
Goes to show there's always another can of worms waiting to be opened. DM
I can program my brain that it converts the music. You should try one day this method.
Perceptual masking is not a binary outcome, and hence "perceptually lossless" compression is never quite what is claimed, but it is certainly a sufficient approximation for most people most of the time, and therefore commercially ubiquitous. I suggest that tradeoff between restoring the missing "perceptually insignificant" data and restoring the additional resolution for those parts not masked is far from clear. It might not be easy for most people to hear either of these improvements, but we know that purist audiophiles will feel good about one and not about the other (see heated discussion on MQA elsewhere for a flavor of that debate).
I often wonder about the full reconstruction of any audio . Even a recording with top material doesn't record everything. We should have a codec which throws away information to save data but the whole process should be done so we could more or less later guess how to reconstruct the original. It would require an AI system which recognizes all the instruments and adds the missing overtones etc. I have never seen a filter, say in Audacity, which even attempts to reconstruct very low bitrate mp3. I wonder if AI could do it.
What do reported "artefacts" of lossy codecs sound like? Can I train my ears to notice them?
I don't think you should do that. You'll never unhear them. DM
@@AudioMasterclass Agreed, they are not pleasant!
Personally I don't find lossless streaming to make much of a point, because even if lossy audio could technically sound worse, are most of us even that focused that we would/could "care" when we listen to music? But at the same time, I'd only purchase lossless digital releases, because it "feels" bad to buy something that is not / less "archiving-worthy". (Unfortunately, UMG almost ruined it all with its pointless but toxic "watermarking".)
I don't think we can do better than cd, or at least not much. However, I'm pretty sure a 128kbps mp3 beats a 22khz sampling rate at 8 bits (16 bit computers of the early 90s normally could only deal with 8 bit samples with half of CDs sampling rate in their sound chips). There - lossy beats lossless.
I wanna hear air in my lossless recording. Yes, you heard that right. Air (clearly it's not an instrument).
Also, I am not so sure if "trashes" is the right word to describe the decrease/difference in bitrate offered by lossy audio...
IMO, resolution, as it pertains to electronics at least, can be quantified by the total intermodulation distortion measurement, which is rarely shared these days as it once was in the days of yore (1970s). One reproduced sound basically interferes with the other when played. It’s crazy high in loudspeakers, but can be minimized in the source and amplification. I think that this is where the recently popular SINAD measurement fails us
I will issue this challenge. Compile a 4 hour play list of your favourite music. Create 2 versions of that play list. One lossless, the other, not. Regardless of the resolving capability of your hifi. Regardless as to whether you think you can here a difference or mot. I bet you are far less likely to suffer listener fatigue with the lossless version. Try it!
💯
Sure if you mix the two up so it’s random which is which. But ok four hours?! Sitting there in the sweet spot? Sounds crazy long how about 30 min
I have a question. Possibly a silly question. This 10% (approx.) that we can hear: Do we all hear the SAME 10%?
Speaking personally, I find lossy music generally acceptable. For a very long time, I did NOT find it acceptable, but that's because it was at 128kpbs (at least, I ASSUME that's why it sounded terrible).
CD quality music (16bit/44. 1 KHz) is also acceptable for me - MORE than acceptable, in fact. Having said that, I did find it quite tiring to listen to it when I first got a CD player (a modestly priced Sharp) back in 1986. It sounded great by my standard of the day, but it did seem to cause some listening fatigue. It wasn't relaxing for extended listening in the way vinyl was for me. But I later found CD to be not quite so tiring (still a bit tiring) when I replaced that player (with a Sony) nine years later. I do, of course, appreciate that there could be at least one extraneous variable at work here (the DAC, maybe, or all sorts of other contributing factors).
As for hi-res music, I did find that to be better than CD quality IF I concentrated very hard. A few years ago, I compared versions of Lan Shui's recording of 'La Mer' by Debussy (downloaded from eClassical). I had downloaded both the 16/44.1 version and the 24/88.2 version. The latter definitely sounded better, but only if concentrating VERY hard (listening on a FiiO DAP with Audio-Technica headphones). More recently, however, I compared the 16/44.1 and 24/96 versions of the December 2023 album 'Caneuon Tyn yr Hendy' by Meinir Gwilym (a Celtic folk/pop album downloaded from Qobuz), and I really couldn't hear ANY difference. Extraneous variables? Maybe my hearing has deteriorated (well, it will have done to some extent by now). One variable I DO know about is that the recording of 'La Mer' is extremely dynamic (maybe too dynamic for most listening conditions), compared to 'Caneuon Tyn yr Hendy' which is only reasonably dynamic (very good by today's usual standards, but there is a little bit of clipping). So, two possible explanations there, but what do I know? I only know what I've experienced, but I can't say how or why (beyond some provisional guessing).
Are you saying you don't hear the difference between a WAV file and an Mp3? I don't think it requires particularly good listening skills to notice that... I think there's plenty of people out there with far better ears than mine and I think the difference between the two is blatant. I bet I would even do well at a blind test between 16/44 and 24/44 masters in WAV format. The differences there however are far far more subtle than the difference between mp3 and WAV. Mp3 just doesn't sound good. The normal audio on services like Spotify is objectively bad quality. If you pay for premium the difference when you turn on HD audio is incredibly obvious. I suppose when you can't even hear 15K!
Why play cd sound quality at 16 but yet 320
Is there really any need for lossless audio at this point with 5g cellular, gigabit coax/fiber internet connections, and $0.01/GB storage?
So you say don't use 20khz con't hear it
I download yt tracks and its great. 5.1 or 2.0 with my system. .
As an AUDIOPHILE a good stereo system is 50% synergy and 50% room acoustics/treatment . When it comes to streaming I've tried Tidal MQA, Deezer and Qubuz and found that Qubuz sounded best. I've tried all sorts of fancy dac's from £200 - £8500 with 32kHz to 192kHz, 16/24 bit, 2 and 4 x DSD 256/512 what ever all that means? Best sounding dac was an old Audio note tube dac with 18Bit Ananloue Devices AD1865N 44.1 / 48 / 96KHz input capability what ever all that means?? I don't read specs i use my ears and enjoy the music.
By not being at least a little bit curious about what all those numbers mean you are shortchanging yourself on a whole world of new insights into your hobby.
lossy audio.... you mean vinyl right?
Fascinating take on things, i think Mr Spock would find your point 'Logical' but in real life I have no idea so I will come back later when there are more comments to help me out but I'm sure somebody's nose is out of joint now. I got the joke but didn't give the polite response so here it is HA HA HA HA....cheers.
I think the 10% figure is misleading because there is both lossy and lossless compression involved. So while the file size might be 10%, the actual loss is not 90%. From there, I think what you're describing is a bit like MQA, which both adds more information AND is lossy. So this would be something like MQA Lite (TM). A file say 20% the size of a standard CD track, but with "better" resolution. Patent this quick and you can retire!
Yeah, Definitely not 10%!
If he was actually listening to 10% of the original data, he obviously wouldn't be as enthusiastic - it's like listening to a recording quantized to 2 bits instead of 16 bits. It's definitely less enjoyable and less inspiring.
The real loss can be estimated approximately if we compare the sizes of .LA (Very efficient lossless compression format) and .AAC, which will be in the range of 30-40%. Which is also quite a lot. But clearly not 90% loss.
ThE 10% is a bit misleading, better to compare CD audio at 1411kbps to AAC at 256k which is about 18%. And one could also compare say FLAC at 800kbps roughly with 256k. Still a dramatic dreduction no doubt. As for lossy sounding better than lossless, absolutely it is possible. 320k MP3 from a CD source is going to sound a lot better to PCM at 320k achieved through a reduction in sampling rate or bits per sample. So yeah. But lossy codecs take PCM as an input, what you propose is not really possible I think.
It might be, if lossy could retain 100% that all humans experience in music and store it in 1 byte.
I don't think "we can't hear 90%" is true. That data size can be reduced by 90% does not mean that this is all "thrown away".
If you run a sine wave thru lossy codec, does it also "throw away" 90% of it?
Easy to test ... use something like Audacity to make a file with a sine wave tone in it... save it as a WAV file then export it as an MP3 ... compare file sizes.
@@Douglas_Blake_579And for what reason do we compare mp3 to raw wav, and not to compressed lossless audio?
Are you aware that any lossy compression is also a mathematical compression (like an archiver), not that you're only throwing away extra data?
@@EG_John
I was simply giving you an easy way to compare the file sizes.
In the case of a pure sine wave you just might be surprised.
@@Douglas_Blake_579 You gave not an easy way of comparison, but a misleading way of comparison.
Compare FLAC file size to lossy file size. It would also be interesting to at the size of a flac that was made of a lossy compressed file (make a lossy file, uncompress it, then put that result into flac - it will probably be a different size to a flac of the original source).
A 320 kbps is 10% of the actual data size of a lossless file?! That is true in compressed form, but not when played! There are YT clips that shows what actually is lost (in sound) between Spotify and the original mastering file. There are just the attack of some dynamics and the upper part of the highs that you can hear as a difference... That is not 90 % lost (it can´t be as a 320 sounds actually great)... You mentioned the mp3 codec... The codec itself will compress the file to 10% of the data size (correct), but when played it will un-compress the data to a level that most people cannot hear in a very smart way and that is more likely up to 90-95% (?)... You need a high resolution rig or trained ears/brain to distinguish a 90-95% vs a 100% file... It is much more important how much ambitions the "engineers/producers" went through when they recorded and mastered the artist than 320 vs losless... My take on this discussion.
I do wonder sometimes if encoded at high enough bitrate from 24-bit audio, could lossy audio technically sound better (if that's even a thing) than "lossless" CD audio that is downconverted from the same master with dither applied. (In other words, dither vs psychoacoustic, who wins?)
I would very much like to know that
Apple phones don't have a micro SD card so it has to stream wirelessly. That's why AAC is still being used.
The also don't have a headphone jack, forcing you to use Bluetooth audio (another lossy compression round but with a differently configured (optimised for latency) codec. Even AAC will be decompressed with different bitrate and other settings).
as i have said before i do not just have a passion for music, i have a passion for sound in general. i just want to experience the best audio quality wether in music, films, video games
Saying that mp3 throws away 90% of data is like saying that flac (losless compresion) throws away 30% or 40% of data because it uses less storage memory
FLAC compression is sorta like when a MS Word document is compressed with ZIP or RAR. The latter makes the file smaller without changing the words and logos etc.
I don’t think it matters. We have plenty of bandwidth and storage these days so why bother?
your an MQA believer.
1) I got 80% wrong when comparing mp3 320k VS lossless, clearly the lossless is higher fidelity, but subjectivity it depends what you are looking for,
analogy a JPG image with slightly higher color saturation to compensate for the lower color depth bit definitions,
2) Spotify OGG VS Amazon lossless,
I couldn't tell any difference casually or AB with a gap in between, but GAPLESS AB testing same volume MATCHING volume levels, shows an ever slightly better dynamics for lossless as reveled by modified Klipschorns a 110db sensitivity at the midrange(modified), which makes some sense BC lossy removes masked information that you wouldn't miss, but the information being taken out, contributes to the overall bulk of energy, thus dynamics suffer slightly.
I thought Spotify uses OGG for lossy streaming, not AAC… maybe I’m wrong.
@@SteveWille Thanks for the correction, looks like now it’s OGG
It was the other way around in the blind tests I did a few years ago. The only ones I got wrong were the over processed / compressed modern Pop styles, where it was impossible to tell which was lossy or lossless. However when the music featured real instruments and not samples, I could tell the difference every single time.
Why have eq then
Is this in defense of MQA?
No.
Hmmm... From what I understand. Lossy audio does not have a bit-depth.
Yeah it does. It is just masked by the file compacting method (similar to zip files) and then restored on playback. For example, the algorithm will take advantage of unused bits in each sample to use fewer bytes of storage, but on play back the unused bits are restored and reset to zeros as the PCM stream is reconstructed.
Fem NPCs are the best NPCs, lol.
At 2:20, you stated that both ears, and brain are involved in the human hearing system. Obviously, you are mis-informed because audiophiles don't use their brains when listening to a highly resolving system.😉❤
You're so funny. DM
Sorry David this doesn't make sense. Every recording engineer can already render their music at whatever quality they like whether its MP3 or WAV or High res. So what are you suggesting?
I am suggesting what I suggested in the video. DM
What we can’t hear effects what we do hear. Harmonics are effected from mixing with everything there.
I think your reasoning is way off. You can listen to 10% of what’s supposed to be there and not know the difference? And people are listening to you about audio?
I can hear differences in cables, DACs, amplifiers etc. Do you think they don’t matter either?
I’ve worked in audio my whole life and have worked in and/or for a lot of the biggest studios in LA/Hollywood. I can tell the difference.
Surely there is no such thing as truly lossless recording. No recording process can capture the audio in its entirety, whether that be sounds below the noise floor of the mic & pre-amp, or higher level than the mic can handle. Or frequencies that are lower or higher than the mic can capture. Also all recording adds some distortion.
Thus as all recording is lossy to some extent, the question is which process is the least objectionable.
I prefer the results of DSD encoding, it tends to sound more natural and open. But whether that is down to creative production choices or the actual encoding process I am unable to say.
That said I also like the Stagetec Nexus ‘mic’ inputs, these are true 32bit, they do not have mic amps as such, but convert direct to PCM what ever signal is fed to them from mic level up to +24dBu, ‘mic’ gain being applied in the digital domain after encoding, with the top 28 bits being fed into the routing system.
I am writing this comment before i watch your (bound to great) video. I have always though that removing some information from the audio might make it easier for amps and speakers to reproduce the audio with lower levels of distortion, both inter modulation and harmonic.
I don't think it's like You are literally throwing away 90% of sound information, there is also some methematics involved, think it more like a difference between JPEG and BMP image files...
The fact is that the 89% that is stripped away
is actually important to the well trained “good” ears that are seeking to maximize their sonic experience on the better playback equipment.
A fundamentally sad reality is that with many untrained ears and the “average” budget playback equipment, that 89% can almost always go unnoticed. 😉
Lossy is quickly becoming irrelevant, bandwitdth and storage grow so fast one could easily stream or store and play lossless, why bother compressing it?
Streaming services might say so that they can offer different tiers of service at different prices. Not to my liking but business is business. DM
If most people don’t care about missing out on 90% of the music, surely they don’t care about you putting lipstick on the remaining 10%. Maybe you should devote your studies to meatless meat.
The fact is you aren't actually missing anything. Can you honestly tell me that you can hear a single triangle tinkling away in the crashing crescendo of an orchestral performance? Or that you can hear the musicians footsteps on stage during a rock performance?
Can't hear it... won't miss it... unless someone tells you they were there.
@@Douglas_Blake_579 So everyone is happy and no further research is necessary. Have a good one.
@@PrismApplied
Not exactly ... but I do think it's far less of an issue than most audiophiles make it out to be.
@@Douglas_Blake_579 The problem with lossy is not that just that it eliminates supposedly inaudible items. It’s that it changes the sonic characteristics of what remains. However, if you don’t find the sound to be objectionable, party on. Why invest in equipment or technology that has zero impact on your experience? I hear that Spotify has a terrific music discovery algorithm. Enjoy, life is short.
@@Douglas_Blake_579Yes to the triangle, no to the footsteps. The triangle has sufficiently different frequency peaks at sufficiently high level to be heard and also not to be thrown out by a lossy codec. The foot shuffling won't be heard, and will also be thrown out to some extent.