How can a lossy compressed minidisc be "cleaner" than a lossless CD or similarly lossy but higher bitrate and more efficiently compressed digital file?
It was mastered from a higher quality source? There could also be errors in the CD mastering, like setting the gain too high and chopping off the top of the song just to make the whole thing "louder". The digital files may have been taken from the CD since nobody can be bothered to go back and find the masters anymore.
> The digital files may have been taken from the CD since nobody can be bothered to go back and find the masters anymore.
All my music collection is FLAC files, bit-perfectly ripped from my CD collection. So, sure, they're digital files, compressed, but lossless. It is, for all practical purposes, just as good as a CD (as long as my external DAC is as good as the CD players' built-in DAC).
The original CD may have been badly engineered but that's another topic.
3 000 CDs fits in more or less 1 TB in lossless (but compressed) CD quality (just to give an idea of how tiny it is using modern storage standards).
Some people only use FLAC as archives but these files are so tiny in this day and age that I don't bother converting the FLAC to mp3s: I sent directly the bitperfectly ripped FLAC files to my DAC / stereo setup.
CDs in a player always have errors. The errors are covered up by a couple of stages of error correction. One is a transparent CRC, the other drops the effective bit depth down to 12 or so bits.
The Block Error Rate (BLER) on CDs can be surprisingly high. So a CD in a player is always slightly compromised.
When you rip a CD with a good ripper the drive does multiple passes over problem areas until the results are consistent. The final file is bit perfect, except on very damaged CDs.
You must mean CIRC. CRC in the context of errors is something completely different. CIRC offers pretty good error recovery, so except for bad scratches, CDs and FLAC (assuming it's also 16 bit, 44000 Hz) are identical.
MiniDisc doesn't brick-wall-compress the sound. Instead, it uses automatic gain to prevent clipping while recording from line-in. ATRAC encoding is also devoid of any compression.
MD players generally have VU meters, and you can observe your music, since they generally have recording functionality and used in studio and/or interviews, getting sound as transparent as possible is the aim.
Fun fact: Portable MD players are generally called "Portable MD Deck" by Sony to signal it's not a mere player but an audio equipment. Most MD decks contain indexing and primary tools for chopping a record to tracks and tagging them.
ATRAC is not devoid of compression. It's a lossy format which uses a crude form of spectral encoding - a bit like an MP3, but the model is less sophisticated.
There is a rare variant called ATRAC Advanced Lossless but it was never available on many devices.
Oh, I think I failed to tell what I'm trying to tell.
ATRAC is of course lossy and compresses the audio. The compression I mentioned was loudness maximizing/homogenizing compression which is used in mastering process, or in audio pipelines in general. So we can say that, ATRAC is devoid of any compressor stage which makes the resulting audio more loud by default.
That loudness maximizer/equalizer also called compressor so it's easy to misrepresent or write misinforming things when talking about both at the same time.
I used to record, in stereo, everyday life. i started with a 32MB smartmedia recorder, then to a cheap-ish compact flash recorder, to a minidisc "net" recorder - the audio i recorded with that netMD is still impressive to me, it feels like you're actually in the recording that i haven't ever replicated, even with the same microphone on other, more advanced devices like tascam dr-05 and the tascam quad channel, or even line in / mic on a computer, regardless of the sound card. I can go put a pair of headphones and a single AA battery into my MD device and be transported back to my college campus, a few of the industrial companies i worked at, the concerts/performances i was asked to record for friends, or even a small gathering/party/game night.
the only stuff i've heard that was recorded by amateurs that comes close (and possibly surpasses, depending on who's running the conversion / editing) is DAT recordings using the microphones shaped like a human head, for "perfect" stereo separation.
anecdotal, for sure, but the "compression" - ATRAC - was extremely good, especially at that time. I am aware that CDs have "no compression" on the data, so obviously DAT -> CD would sound better than MD -> CD, but this is my opinion.
First of al, ATRAC is a strange format. It's optimized for perception as far as I understood it. It doesn't color sound but can reproduce the dynamics and details very well. It's important to understand ATRAC has quality levels: SP, LP2, LP4. SP fits 80 minutes to a 80 minute MD. LP2 fits ~160 minutes, LP4 fits ~320. SP is extremely detailed and transparent. LP2 is really pretty good. LP4 is not good. Is not bad, but for listening on the move. It's not suitable for joy-listening on higher resolution systems. The deficiencies will be definitely heard.
This sound quality has another reason. The DSP/DAC combo on MD players itself. Earlier versions were called Type-R, improved versions are called Type-S. Latter one can reconstruct sound very impressively. It can make SP and LP2 sound really good. It also has a digital amplification section. So, It's only converted to analog at the latest stage. Result is a very dynamic and clean sound. Especially if you convert to ATRAC from CD or any lossless source.
My Hi-Fi (Sony CMT-HX3) also has a 35W version of that DSP/Amplifier combo. It really sounds fantastic* with CD, MP3 and AC3. Didn't try ATRAC, but it won't disappoint I presume.
Sony's higher end systems and DSPs are a work of art. They sound clean, crisp and detailed. They can sing through a pair of good headphones or through 2x85W behemoths. This is why they're timeless and regarded highly. With just Bass & Treble settings, it beats a lot of presumably higher end portable players.
A similar sound quality was present in Creative's Muvo^2. It was plain enjoyable to listen. Apple's iPods famously use Burr Brown DACs but, the sound was too flat and dull in the earlier versions, lacking punch and energy. They fixed it after iPhone 5.
TL;DR: Sony's DSP/DAC/Amp stuff makes any well designed audio codec shine. ATRAC can store dynamics very well. This is why being a Hi-Fi company pays in the portable space.
*: I have a more conventional, but vastly higher powered system. This small guy is not a match for it, but for a small hi-fi it really hits above its league. If you scale this design up (to 80+ watts) with proper speakers, I'm sure that it'll make a lot of more expensive systems sweat.
The DAC isn't part of the format; although I respect the desire for a good DAC. I believe all lossy compression codecs from the past 20 or 30 years are perception based; ATRAC may have been one of the first ones, but it sure doesn't beat a CD which has no loss at all, unless ATRAC/minidisc regularly used higher bit depths than are supported by CDs. The sample rate is already "perfect" (in that humans could never hear a difference)
You're right however, the DAC can make or break the resulting sound. If the DAC wouldn't be important, we wouldn't have Wolfson, TI Burr Brown or other extremely expensive units. Sony's DAC in this pipeline is like a polish. It doesn't blemish the sound after all that meticulous reconstruction.
On the paper and spectrogram, an SP ATRAC track is not a match for WAV or FLAC, however it has all the ingredients to saturate a high-resolution system and create an enjoyable listening experience. Also it had enough resolution to handle higher-end amateur mastering (in SP mode) and general recording needs (in SP and LP2 mode).
While 44.1KHz is perfect mathematically, 48KHz and 96KHz can create a smoother sound. That needs higher end equipment to hear clearly, but the smoothness is definitely there.
I would really encourage blind testing of one's assumptions about audio quality. Foobar2000 has this super nice ABX plugin that can handle automatic conversation to different codecs and or PCM sampling rates and depth from lossless sources. https://www.foobar2000.org/components/view/foo_abx
But move on to anything more modern than mp3, such as Apple's AAC encoder (there's a ton of bad AAC encoders out there) or Vorbis with VBR settings averaging at 256-320 kbps. You'll find it very hard to spot the lossy encodes in a blind test. Opus is astonishingly good at even lower bitrates.
> The guy who created the Vorbis codec explains why this really isn't the case and why high sampling rates with ultrasonics actually are bad for playback quality.
I'll take a look to it, thanks for the link. That sounds interesting.
> I would really encourage blind testing of one's assumptions about audio quality.
I'd gladly take an ABX test and stand corrected. digital feed's ABX test is pretty good but needs some quiet time to do it correctly. I'll try to find some time to take it. I tried but, it's not good when you don't have time.
> I can personally recognize full bitrate mp3.
It's possible to recognize some equipment (software or hardware) by ear because of its distinctive sound dynamics. I've listened a limited set of stuff (amplifiers, headphones, CD players, etc.) and can recognize each of them by ear. I think starting early and playing some instruments in groups help with ear training.
> But move on to anything more modern than mp3, such as Apple's AAC encoder...
The latest generation of encoders sound extremely transparent. You're right about it. It's a good thing actually. I'm not grumpy about not being able to brag about identifying the codec of the file I'm listening to. I just want to hear enjoyable, detailed, high resolution music. If my ears and brain are happy about what they're hearing, I'm happy.
Actually, here's an online ABX blind test set with several modern codecs at high bitrates, as well as the bad old mp3 garbage that ruined the reputation of lossy compression.
I really challenge you to reliably detect the lossy sample in the tests for Apple iTunes AAC ~256 kbps and Spotify high quality. http://abx.digitalfeed.net/list.html
The reason we have very expensive consumer DACs is that there are a lot of retired people who listen to music all day and have nothing better to spend money on.
48khz vs 44.1khz has the same sound quality in the human-audible range (maybe not if you're a dog); the actual reason DVD is 48khz instead of 44.1 is that it syncs up with 24fps frame rates better.
> The reason we have very expensive consumer DACs is that there are a lot of retired people who listen to music all day and have nothing better to spend money on.
After a certain point, yes. I'd personally wouldn't distinguish a %0.001 THD Burr Brown DAC and %0.0001 THD Burr Brown DAC in normal listening volumes with a good amplifier and set of good speakers. I inadvertently own both variants (one is in my CD player, the other is in my sound card).
However, being a former orchestra player, I assume I can hear a really bad DAC and a difference between 44.1KHz file and 48KHz file under some circumstances (like high detailed cymbals or other sounds in treble range). I'll happily stand corrected though.
None of this stuff is real unless you can pass an ABX test. I doubt it's real just reading it.
ATRAC is not optimized for anything that other codecs aren't (and technologically it's quite bad, especially compared to Opus but even compared to AAC). And there aren't magically better DACs any more than there are better power cables; they're all good enough as long as they're electrically compliant with your headphones.
> None of this stuff is real unless you can pass an ABX test. I doubt it's real just reading it.
That's of course fair. I can't ship you my MD player (that thing is priceless for me), but I can share a triplet of files for ABX testing. If you wish, I can do the following:
- Rip a track from an album before loudness wars began to FLAC.
- Rip the same track with Sony's own MD studio to ATRAC SP.
- I can re-encode said ATRAC SP track to FLAC.
- Ship you three files (A, B, X) where A is FLAC, B is ATRAC SP and X is either of one.
If that's OK for you, I can do it this weekend.
I'm not claiming anything or hoping for any outcome. My aim is to test, discuss and broaden my mental horizon. I've written these things by listening same stuff in many formats through the same audio pipelines (audio cards, Hi-Fi systems and speakers/earphones) for many years and, I instinctively know how every one of them behaves with any combination. Also, I've played in orchestras (big bands and symphonies), so my ear is somewhat trained to find nuances.
Making the some tests and discussing these things will only make me happy, and I'd gladly stand corrected.
Isn't ATRAC SP mode above 250 kbit/s? ATRAC3 is about as good as MP3 - probably worse because the encoder isn't tuned as well as LAME - but that rate is certainly high enough to sound acceptable.
It won't always be transparent so the ABX test should still be passable on extreme content, but it'll sound good enough. Anything else would also sound good enough though.
> Also, I've played in orchestras (big bands and symphonies), so my ear is somewhat trained to find nuances.
How's your high end hearing? It tends to go once you're not a teen…
It can go up to 352kbps as Sonic Stage tells me. Will transfer one album to my MD player to see how it converts it to ATRAC. SonicStage rips your CDs to ATRAC3 Advanced Lossless by default and re-converts them to appropriate format for your device during "burning" it to your MD player. I'll use the files transferred to MD for ABX test.
> How's your high end hearing? It tends to go once you're not a teen…
I assume it's not bad. Can hear pretty silent stuff pretty well, including very high pitched adapter noise (not the capacitor whine, but the switching noise of the adapter itself, the one supposedly bothers dogs).
For everyone interested in what I've told in the GP, can you please register via the form [0]. This Sunday took a different turn but, I don't want to leave this on the table.
