MDietrich

I think that´s it in a nutshell (not nutcase!!). Couldn´t have put it better myself... it´s strange: after years of handling 'virtual' media (-> files on the PC) I enjoy it tremendously to handle physical media again, touching, changing, using it. I mean I still use CDs but once they are on the PC they sit on the shelf doing nothing. Using physical media like MD (which forces you to deal with media & the music) creates a connection to music I didn´t know I was missing until I started to use MD again.

Ah, ok... apparently I´ve forgotten it. Now I´m ashamed, sorry. The error is that they are talking about the MZ-1 as the first device hitting the market in 1992 when it´s clear that they used a picture showing an MZ-N1. One might neglect the possibility of an error but I think that it actually is one because of their similar designations.

http://www.digitalspy.co.uk/tech/news/a455520/sony-minidisc-player-shipments-to-cease-in-march.html I wasn´t aware that Sony was still producing MD gadgets for the Japanese market. Oh, and note the error in the link above

The result depends on the music. Punk, Rock or Pop have very weak dynamic capabilities (differences between soft and loud are very small) whereas classical or jazz have more or less level differences up to 50 dB. The latter cases might reveal truncation artifacts (quantization noise) on soft parts while the former for certain won´t. However, if I´d do something like you I´d use dither ALWAYS. Dither is harmless noise, it shares many characteristics with tape noise (or shaped white noise) but its gain is much, much lower and I´ve never ever witnessed it becoming audible. You can safely use it all the time. But only once on your last processing step and you shouldn´t do any processing afterwards (e.g. resampling; not even volume correction), dither should always be the last step. No. Since ATRAC 3.5 the MD accepts 20/24 Bits through its S/PDIF input; all my MD recorders (in my signature) accept a full 24 Bit signal and encode it properly. But you won´t gain a benefit if you upconvert from 16 to 24 Bits. If you´d do that you´d only add zeros in the 16to24 bit space which add nothing, they are not even interpolated. Before you play original 24 bits material you should also make sure that your DVD player transmits it without changing it to 16 bits. Some DVD players do this because of (stupid) copyright issues, on many it´s configurable, on some it´s not.

Yes, true. Starting with ATRAC 3.5 MD used 20 Bits, starting with ATRAC 4.5 it used 24 Bits (how much bits are actually given out I don´t know). BTW, the inputs of your decks is independent from what you configure in their respective menues as they will always accept the highest bit-depth possible on their digital inputs: 24 Bits (except 3.5 & 4.0: 20 Bits... but with that bit-depth truncation doesn´t matter anymore). CD-Rs and Audio CD-Rs technically are exactly the same. The latter one only has an additional embedded code (in the ATIP area I believe) so that any purely audio-compliant recorder will recognize it, subsequently allowing you to use it. The 'normal' CD-R for PC only lacks this code. The only difference between CD-Rs and Audio CD-Rs is that royalties have been payed for Audio CD-Rs, royalties for expected audio use - that´s why audio CD-R recorders won´t allow you to record on 'normal' CD-Rs since no one payed royalties for them. In some rare cases however there existed (still exist?) Audio CD-Rs specifically constructed for lower write speeds. With basically all available high-speed CD-R it´s difficult to write at low speeds because the error count will be higher as they are constructed for high writing speeds (16x and above), using that they produce very few errors. I´ve tried this myself with four different drives and from my error counts I found this to be true. Slow burning speed (1x, 2x, 4x, 8x) was beneficial for quality 10 years ago - but it isn´t anymore. Sorry for veering off-topic but I had to explain why special Audio CD-Rs might still be suitable: they might be engineered for lower writing speeds necessary for older CD-R HiFi recorders.

May I chime in? If you record from 24/20 Bit MD to CD-R it is crucial to change the bit-depth of the digital output to 16 Bit. If I remember correctly the MDS-JA 333 ES then decreases bit-depth with dithering. If it wouldn´t do that the CD-R recorder would get the full 20/24 bit signal and simply ignore those additional bits which could theoretically create truncation artifacts. The CD can ONLY ever use 16 bits, nothing more, nothing less. But with proper dithering (done by the MDS-JA 333 ES) you can avoid artifacts AND capture almost all of the full resolution on CD. Think of it as a built-in Super Bit Mapping that can be switched on or off for your convenience. For recording on CD I´d switch it on. But if you output the MD signal to a DAC capable of 24 Bits switch it off.

I don´t think so - the ATRAC ICs are completely different, they are not only different hardware revisions but completely new chips. To know for sure a look at their respective Service Manuals should clear things up.

I´ve written a new article where I describe hacking the firmware of three recorders so that they might use the penultimate ATRAC version. Most of you probably already know that the R-500 / 700 can be hacked to use DSP Type-R - but I bet not one of you know that the same can be done to the MZ-R 900! Read all about it: http://marlene-d.blogspot.de/2012/11/to-hack-or-not-to-hack-sony-mz-r-700-mz.html

Same here in Germany. But your infatuation with the command line could come in handy. At least you know what a 'switch' is when someone talks about using them with mp3, flac or wavpack.

Hello! I´ve just acquired a Sony NW-HD3 and I would like to convert my music with SonicStage to ATRAC3Plus, using a bitrate of 256 kBit/s. The problem is: I´d like to use the High-Quality-encoding setting! I can do this with Sound Forge - but Sound Forge then looses the tags. Sonic Stage seems to be able to use the High Quality setting too - but only when ripping a CD directly. Is there a way (perhaps through a registry hack) to enable this also for direct transfers from the library to the walkman device?

Don´t confuse dithering or noise shaping with ripping Both things don´t have anything to do with each other. A CD with resolution better than the standard is like any normal CD. The signal is just contained inside the, let´s say "container", of the standard 16 bit PCM coding. To do this one needs a trick: dithering and noiseshaping. The resolution of 16 bit is basically limited by the quantization noise. The 96 dB possible dynamic were derived from the point where the quantization noisefloor starts. This noisefloor consists of nothing but quantization noise equally distributed throughout the frequency (has the same gain at 100 Hz, 1.000 Hz, 10.000 Hz etc.). In reality this noise isn´t really noise, it´s a distortion. You see, noise (like tape noise, microphone noise etc.) is normally random and not disturbing to our ears. Quantization noise is not random and our ears are very able to pick it it up. So dithering does something very clever: it adds randomized noise to hide the quantization noise. Just like the quantization noise this noise is equally distributed. With that you reduce the dynamic of course to roughly 90 dB. And that´s where noise shaping comes into play: the noise is moved towards higher frequencies where you can´t hear it. The genius thing is that the quantization noise that´s still hidden within the dithering noise is moved along with it! So instead of equally distributed noise you now have a mountain of noise starting at 15.000 Hz - but no noise (not dither, not quantization) at lower frequencies. Et Voila: at those lower frequencies you now have a dynamic of 130 dB - and all inside 16/44.1! You do this at the same time when decreasing bit depth from 24 to 16 Bit. If you wouldn´t you would loose any information below 96 dB as it would turn into quantization noise. Dithering and noiseshaping just avoid this by shaping the noise / distortion that would normally eat up musical information towards higher frequencies. I´m sure you´ve seen CDs having a SBM print. SuperBitMapping is exactly that: noiseshaped dither. Just like Apogee's UV22, Telarc's 20 Bit process and countless others. A dithered and noiseshaped CD is just like any normal 16 bit with 16 bit nominal resolution - with a 24 bit information inside. So it doesn´t matter if you use EAC, Nero or iTunes - it will always be a 16 bit CD with quasi 24 bits inside. You assume that the aforementioned programs add noise to the ripped CD - but let me assure that they don´t. A CD is a digital medium, you can copy it a million times and in theory it will always be the same (apart from jitter - but that´s something else) with no added noise, distortions or anything else.

I have a small problem with my MZ-R 900. Playback works nice, recording doesn´t though. Recordings made with it create problems like silence for our Kenwood DM-5090. In extreme cases it will just jump to the next track. I opened it to see how the Kenwood reacts to such discs and watched the mechanical gear that moves the laser sled turning back and forth, as if the wobble groove of the MZ-R-900-recorded MD would be flawed or inconsistent. The MZ-R 30 behaves the same during playback but apparently it has better mechanics/optics; it doesn´t show any errors. Of course, this problem happens with any disc so I assume the mechanism of the R 900 has some problems. Any idea how to fix this or if it´s even possible?

Yeah. The scale for increased resolution is logarithmic. You´re also right about an SNR of -120 dB (more precise: maximal possible dynamic). With 24 Bit it would be -144 dB for SNR. Theoretical of course. My ASUS Xonar for example reaches an SNR of roughly -114 dB which would be roughly 18-19 Bits of resolution. Almost no microphones or microphone amps are as good. Today, digital recording resolution gets limited by the other things used to make a recording. BTW, through proper dithering the CD can have a resolution close to that (it´s true). 96 dB are the calculated standard from times when the format was introduced. Through dithering and noiseshaping one can reach a dynamic of roughly 130 dB - all within 16 Bit and 44.1 kHz.

Yes, it´s extremely difficult to assess all information about S/PDIF... what bitrate, what samplerate, etc. If I´m correct not even high samplerates are originally support. I may be wrong there though. You´re also right about 100 dB being too much for 16 Bit. Though I have the Kenwood DP-5090 (a CD player) which measures as having almost 99 dB of SNR and dynamic while still being only 16 Bit. BTW, the best measured player yet... I mean, the best I´ve "measured" myself with RMAA. Interesting, isn´t it? The Kenwood measures better than the theoretical max. Either it does something odd or RMAA measures wrong. Or I did something wrong during measurment. The article at nwavguys blog is extremely good! Very recommended reading for those who intend to use the software... why didn´t I think about the link? Thank you very much! As for the loopback test... that´s the one test I avoid like the devil avoids holy water. You don´t know what you´re measuring: the input or the output since both are connected. Which one is the noisy one? Which one distorts? One would never know... For CD or MD players I always use just one soundcard: the E-MU 0202 USB. For the E-MU 0202 USB and the Creative Soundblaster X-Fi HD USB I´ve used the ASUS Xonar Essence ST since it shows extremely low noise and distortions. Of course, NOT doing a loopback creates the possibility of grounding errors (probably the reason for my measured differences with USB cables)... so not everything is peachy.

That´s why I asked. The receiver we own is a Sony STR-DB 830. Its digital input will accept 24/96 signals but it´ll never show the bit depth, only the samplerate. So maybe your receiver just accepts it without questioning. Or maybe it´s "just" a 20 Bit signal which the receiver cannot show but will accept anyway.

The software I use for testing is called RMAA. You can get it here: http://audio.rightma...index_new.shtml - in its basic version (the one I use) it´s free. Well... not completely. ATRAC - just like every other lossy codec removes high frequencies by principle (depends on the codec and the bitrate of course). For the ATRAC LP4 it doesn´t make any difference if there are frequencies or not, it will remove them anyway (it cuts away everything at 13 kHz - well within audible range). Going from a lossy format (for example mp3) to ATRAC would be worse for quality in theory. When a lossy codec takes something away the thing that is left is more or less clever hidden quantization noise (beside a lowpass, joint stereo, etc.). Imagine a lossy source for ATRAC to encode: it not only has to encode the things anew that already have been encoded, it also has to encode the audible rest AND the quantization noise. We might not hear it, but ATRAC does, it even hears things we will never ever hear. When going from a lossy source to ATRAC it has to recompress all the data that was left during the first encoding. And depending on the bitrate this may not be that much. If it still sounds transparent that does only mean that us humans can be fooled easily. Please, Stephen, you don´t need to apologize. Nor does anyone else. I´m the one who confuses people... You know... I don´t want to complain but it´s funny. Scientists accuse me of dumbing down my articles so that regular people (aka scientists) might understand them which is my goal. I write because I want to clean up misconceptions about vinyl, high res, lossy codecs, etc.. But these regular people state that my articles are too complex, require too much foreknowledge or are too scientific. If I would be scientific I would do a complete study where I first would have to create a null theory to begin with, would do many measurments to have as much empirical and statistical data as possible, would incorporate probability (statistical probability), would avoid personal opinion completely, would transparently explain my setup (well, that I mostly do) - and I would do real measurments. The software I linked above has many flaws, one of them is that it isn´t entirely objective. It cannot measure everything and it interprets things wrong. The numbers are just numbers already interpreted. These numbers it presents don´t have any real expression. Now my dilemma: I have to use it because I cannot afford true measurment equipment (which would be bulky, many different things and horribly expensive). The graphs RMAA creates are just one thing: nice looking pictures. I use them to make assumptions just because I don´t have anything else. But in reality they are only pictures. The charts are just what RMAA presents to me - and since no one knows how RMAA works on the inside (it could easily lie for all its worth) it isn´t reliable. A different set of measurments could come to a completely different outcome. I try to work around the flaws of RMAA by stating electrical charts of the things I "measure" so that anyone knows how to compare them. The problem with RMAA is that many use it: private people, computer magazines... and those results are completely worthless because no one usually states how these measurments were created. For example: I´ve read a review for my Soundblaster X-Fi HD USB from a big magazine. They measured it using RMAA and had horrible results making the card much worse than it really is. From my experience with Windows I however knew that a misconfigured Windows audio engine (samplerates mismatch) produced these results. But they didn´t know - and they are supposed to spot this error! But RMAA didn´t tell them that they made an error - it doesn´t know that. Could also be that they were paid for the review to look bad by ASUS or another audio card manufacturer, happens all the time. The funny thing is that not one of their readers noticed the errors. I repeat my "measurments" as much as possible to create a relatively expressive result but I cannot really know. Sorry for answering this long. This thread has derailed and its my fault. BTW, would anyone want to see measurments for LP2 or LP4? I can use Sound Forge to create pure ATRAC3 or ATRAC3Plus files (not ATRAC2 though). EDIT: I will write an article about codecs. Actually I´ve been planning this but now I will do it. If every coded works the way I expect it to work I will show what has been erased and what is left - dynamically alternating quantization noise basically. With a bit of effort I will also include ATRAC2, ATRAC3, ATRAC3Plus, WMA, OGG, AAC, MP3. EDIT II: I´m just listening with a Sony MZ-R 55 I´ve acquired a few days ago. I know everyone hates it... but I love it! It looks so beautiful and feels so nice to the touch... I hope that it doesn´t break to soon. And boy, it is fast! It sounds so well, it easily beats my MZ-R 30 and comes close to the MZ-R 900. Of course it´s WideBitStream capable during digital recording even though it cannot give it out analogue (16 bit D/A converter) I´ll plan to write an article about it and a player I now have (Sony MZ-E 60). I just had an idea: Stephen, you have the JB980, yes? I would think that it does indeed puts out a true 20 Bit digital signal. I´ve seen the Service Manual for our Kenwood DM-5090 and there the optical output goes straight from the ATRAC chip to the output. 20 Bits seem to be the standard output for ATRAC. Or there´s something I´ve overseen. You could test it yourself with RMAA (would be good enough for that) and a Soundcard having an optical input using 24 Bit and a matching software to record that.

My articles are not scientific. I frequently address this issue during most of my articles. And I made it clear that it is my personal opinion by saying "And I mentioned LP4 is horrible (it really is, I can´t stand it)" - I think "I can´t stand it" should make it obvious that it´s my personal opinion. But I´ve just edited my article to make it more clear that a particular statement of mine is my personal opinion. I wrote: "And I mentioned LP4 is horrible (it really is, I can´t stand it) because I listened to it. With music at least. But I haven´t tried anything else, Audiobook or anything else." You quoted: "And I mentioned LP4 is horrible (it really is, I can´t stand it) because I listened to it. With music at least. one should be able to listen while being in a noisy environment." The part that has been misquoted in context has been underlined by me. The recording quality should be even worse then because it is basically a transcoding from one lossy format to another lossy format. But I never did such a recording so I cannot comment on this. I´ll accept your observations. I´m sorry, I only wanted to offer a possible explanation.

I don´t mean to be disrespectful - but this is an opinion you would have to accept as my own. I´m not making general assumptions. Furthermore, I would never have combined senctences out of context in the way you quoted them. Besides offering a sound that is unacceptable to me (for me, which doesn´t necessarily mean to anyone else) it also creates intermodulation distortions so high that they are readily audible.

Of course the sound of the recorder/player I use is important. But every player - good or bad - needs a perfectly encoded source I think. A good player won´t improve an encoding that hasn´t been encoded well in the first place. The player only hides possible errors in that case. And I don´t like to be fooled. And I mentioned LP4 is horrible (it really is, I can´t stand it) because I listened to it. With music at least. But I haven´t tried anything else, Audiobook or anything else. No, never. It has never been repaired. The chip had a name: CXD-2652AR. That was surprising to me since it´s exactly the same ATRAC chip used in my MZ-R 30 and I thought it to be a special chip engineered for portable use only (-> energy conserving). The reason for that is simple: the Sharps have much more powerful headphone outputs. Most MD recorders or players from Sony have 5 mW on their output while many Sharp have 10 mW. With the Sharps one should be able to listen while being in a noisy environment.

Fair points. But as I´ve written in my article, DSP Type-S and DSP Type-R are the same, the only difference being the decoding part for MDLP data. DSP Type-R was introduced as soon as 1998, back then only with stationary decks. So compared to the Sharp and its presumed latest ATRAC version the Sony ATRAC has an advantage even though it´s five years older in development. Furthermore, the Kenwood DM-5090 doesn´t use ATRAC 4.5 but ATRAC 4.0 - I´ve confirmed by looking at the ATRAC chip directly below the drive. Kenwood was lying back then about the ATRAC version. But for playback this doesn´t matter. I doubt that distortions or dynamic will change if my test MDs would be played back with a post ATRAC 4.5 codec. Considering all of this one might now understand how disappointed I was when Sharp's ATRAC ended up like that. They had five years to come up with something better - yet they didn´t.

Just an update: I´ve reviewed the Sharp MD-MT 180 & the Sony MZ-N 510 and their respective ATRAC versions. I´m afraid that Sharp's ATRAC isn´t very good. Read all about it here: http://marlene-d.blogspot.de/2012/09/more-md-recorders-sharp-md-mt-180-sony.html BTW, am I right to assume that the Sharp uses ATRAC 6.0?

Me too, I´ve never heard an ATRAC 1.0 encoding. It would really be nice if someone could record something with an MZ-1 and post it here. If it´s really that bad 30 seconds of material should suffice (so that it stays legal). This is a great suggestion btw...

Thank you! I´ve never been able to read much about those decks. Probably better to stay with MD... but I miss my old tapedeck http://www.gersic.com/blog.php?id=54 http://www.hydrogena...06 http://www.hydrogena...showtopic=92974 They tell what I wouldn´t do - since it can be misleading: mp3 doesn´t have a bitdepth. Just like AAC, OGG, or WMA (even though MS tweaks this a bit). The input doesn´t matter to the mp3 encoder, the decoder is the important thing: will it put out the 32 bit floating point or won´t it? foobar2000 will use the 32 bit, jriver mediacenter will too. For converting I´d recommend dbpoweramp, there is an option in its configuration menu to switch on decoding to 32 bit floating point. Which, btw, is how I made my test files (which prove that mp3 really achieves 24 bit quality).

True, this thread is getting out of hand - something of which I can be accused surely. Regarding the upsampling (don´t need answering soon): decompression and upsampling are two completely unrelated signal processing events. After LP2 is decompressed it behaves exactly like every other digital audio signal, it doesn´t matter if bits are missing or not. After that you can do with it anything you want, the outcome will be what you will treat it to. Theoretically the outcome should even be better since LP2 contains no frequencies beyond 17 kHz, meaning there is nothing left close to the edge of the passband (-> 22.5 kHz; assuming 44.1 kHz sampling rate) what could be aliased back into it. And today fractional resampling rarely poses a problem. It takes a bit longer to process (because of increased number crunching) but that´s all. You seem to assume that the resampling process somehow knows what it is resampling; that´s not the case. A resampler is a robotic tool that doesn´t think, it just does what it´s programmed to do: resample audio data. Resampling is one of the most transparent DSPs known to me - compare that to an equalizer: with that you would expose the missing bits easily.

All in the playback? Hm... ATRAC3 seems to be an asymetrical codec; it probably was designed not to be too demanding on the decoding part (-> less power consumption). So the major part is done during the encoding itself. If ATRAC Type-S manages to get quantization noise down when decoding then it´s probably because Sony simply improved the decoding bit-depth. Everything else is highly unlikely. Which in turn means that while the old ATRAC was quasi-floating point ATRAC3 was not. This is all is not based on actual information though... ATRAC3 was never made public so no one really knows about it. And I have to ask: why should it be bad to upsample LP2? The result would be the same as upsampling CD, ATRAC, or mp3: upsampled audio. The output quality only depends on the resampler. A good resampler shouldn´t make the audio material better or worse. Organ should be easy to encode. Organ pipes usually give out pure sines which are easier to encode. However, depending on the organ and the venue where it is housed a recording could have a lot of reverberation which might be hard to encode, especially if there are additional wind noises (usually very audible with older organs). BTW, I´d like to go back to your JB980: have you ever actually tried to record its output? Maybe it puts out true 20/24 bit audio through its digital outputs, just because Sony didn´t mention it in the manual that doesn´t mean that it isn´t able to do so. Remember, my Kenwood puts out 20 Bits too, even though it isn´t mentionend in the manual. Maybe you are transmitting true 24 Bit data to your receiver without knowing it. Do you have a soundcard with an optical input, capable of recording bit perfect digital signals?