Blast from the Past: MiniDisc, Wide Bit Stream and 20 Bit

I cannot really tell you, dear reader, why I´m so interested in ancient digital technology. I suspect several reasons but none of them seems clear enough for me. But I know that it seems to be interesting enough for others too - otherwise I couldn´t explain the 3000 monthly pageviews I still find surprising given those geeky topics I usually write about. Today I will not only write about a nerdy subject but also about an example of now ancient technology that quickly became technically obsolete: the MiniDisc. Some months ago I reviewed several portable CD players I had acquired; in that same article I mentionend in passing my first MiniDisc recorder bought in 1996, the Sony MZ-R 30. I also published several measurments I made with this unit and talked about some of the technical features enabling the MiniDisc to function. Ever since that article I had the nagging feeling that I needed to elaborate a bit more on the technical aspects of that system as well as writing about its sound in detail. Well, here it goes.

MiniDisc logotype (copyright: Sony)

The MiniDisc was introduced to the market in 1992, at a time when sales CDs reached an alltime high and the Cassette Tape approached its final evolution with the introduction of Dolby S. By then it already was apparent that the Tape was going to be replaced by CD technology, just the same as Vinyl was. Additionally, Sony and Philips had already developed and standardized the CD-R format in 1989. Back then however the technology needed for CD-R (lasers and media) was alarmingly expensive & still in its infancy so both companies developed something else supposed to be equally capable of digital recording but much less costly. Philips decided on the ill-fated DCC format while Sony decided to use something more cutting edge: MO-discs. They also determined the new format to be much smaller than anything else released before; in order to achieve that they had to use a lossy compression scheme. The result was an MO-disc shrinked to one fifth of its former size combined with a newly developed lossy compression called ATRAC, all of it encapsulated inside a nifty package looking like a tiny floppy disc. This new format appropriately named MiniDisc was - just like DCC - released in 1992 in form of the portable recorder MZ-1. From the minute they presented it I was fascinated by it because it offered some traits completely new to the audio world. Beside being able to record something digitally the MiniDisc was highly portable (remember: we are talking about a time when flash memory was still in its infancy and had yet to be released). But it was the possibility to edit its content to your liking that catapulted the system into the limelight: a user was able to move around or to erase a particular track or several tracks, one could also divide or combine these tracks. In short the perfect personal "Best-of" was now possible - all in flawless digital quality with then unknown portability. Or so Sony thought.

Sony MZ-R 30: have a look at the "Erase" and "Track Mark" buttons - a first back then

Since Philips was going to unleash its DCC system in 1992 Sony had to follow suit in order not to loose marketshare. The engineers of the first MD recorder however knew that they would need at least another year to perfect the compression scheme ATRAC. Have you ever heard the first mp3-encoder from the Fraunhofer Institute? I´m old enough to remember that I did; it sounded just awful. I encoded some track from an Enya album with 128 kBit/s and after decoding it back to .wav I noticed a strange pumping effect and distortions like flanging and pre-echos. Clearly, mp3 wasn´t ready to be used - and the same happened three years prior when the MZ-1 was released. Early tests noticed some strange artifacts sounding like crackling fire on top of sounds in combination with a generally metallic and dull sound; the first version of ATRAC just wasn´t ready for public release. The biggest leap in quality was supposedly achieved with ATRAC 2 one year later: much less metallic, frequency cutoff extended to 18 kHz (before: 15 kHz) and less audible crackling noise. ATRAC 3 from 1995 came close to DAT in a DBT, the crackling noise had almost disappeared. But only ATRAC 3.5, released first in August 1995 with the home component MDS-JA 3 ES would be virtually indistinguishable from CD without metallic sound, noise or anything else. This story serves to illuminate the course of technical improvement being quite visible to the general public. Especially people in Europe were very critical of the new system: the MZ-1 was very expensive, MiniDiscs were equally expensive and all of it sounded just awful. Additionally Sony made an incredibly stupid marketing error: Instead of concentrating on the systems' recording ability they advertised it with pre-recorded MiniDiscs, leading many people to believe that the MD was the CDs successor. Back then people hadn´t yet accepted lossy quality and so they stayed with Tapes a bit longer. In Japan however the MiniDisc was successful from Day One.

Sony MZ-R 30, using the Wide Bit Stream capable ATRAC 4

But all of that changed with the MDS-JA 3 ES. The major reason for improved quality was the decision to switch to quasi-floating point operation within the ATRAC encoder/decoder circuit. Sadly it is an almost publicly unknown fact that virtually every lossy compression scheme (mp3, AAC, OGG) internally works with floating point arithmetic. Integer processing isn´t suited that well for audio since it only can support a limited range of values whereas floating point can represent an almost unlimited amout. The first three ATRAC IC designs processed by using integer values only, limiting their range. When Sony got rid of this restriction ATRAC was ready to work with improved quality, subsequently becoming able of capturing high resolution material. Using their patented "Block Floating Operation" the MiniDisc now was ready to process and store information with higher resolution than CDs 16 Bit. And instead of reducing the bit depth to 16 Bit again during en/decoding Sony just used 20 bit capable A/D-D/A converters in order to make sure that the improved processing depth would be preserved. Meaning: the MDS-JA 3 ES was the first truly 20 bit capable consumer recording device. In 1995 this was a small revolution - not one recorder had been capable of delivering a resolution that high up until that point. Sony therefore took the wise step by advertising it as such while making sure that every internal processing step matched this 20 bit capability, calling the entire process "Wide Bit Stream". On home consumer devices a resolution this high makes perfect sense, it´s an entirely different situation with portable recorders however. Even if a recorder like my MZ-R 30 would work within this bit depth it would be pointless; its combined A/D-D/A converter Asahi Kasei 4515 is a true 16 Bit converter, rendering the internal processing power useless as a result. Likewise, digital 20 Bit sources weren´t existing back then so this impressive processing precision effectively only made sure that CDs were encoded without flaw.

But since the day I first read about it I have been interested in finding out if Sony spoke the truth. Could a lossy device like the MD be capable of delivering higher-than-CD resolution? All those years ago I couldn´t do revealing tests myself only because I lacked the equipment for it. But hey, we now have 2012 and I own the Creative Sound Blaster X-Fi HD USB! As you know it features optical in/outputs delivering true, bit matched 24 Bit signals. My boyfriend also owns a Kenwood DM-5090 (1997) which I suspected to be Wide Bit Stream capable. BTW, despite being advertised featuring an ATRAC 4.5 IC the Kenwood only uses the same ATRAC 4.0 IC from my MZ-R 30. For measuring pure ATRAC performance I devised a  test routine that would protect the digital signal integrity at every step during my procedure without any analogue intersteps. I created an RMAA generated test signal with 24/44.1 and did the following:

SB X-Fi HD ---> Sony MZ-R 30 (recording) ---> Kenwood DM-5090 (playback) ---> SB X-Fi HD

Recording was done digitally using the MZ-R 30

Playback was done digitally with the ATRAC 4.0 equipped Kenwood DM-5090

The results were surprising to say the least. My old MZ-R 30 turned out to be a true Wide Bit Stream capable recorder - only when using digital connections of course. Just imagine you now would find out that a recorder you´ve known for 16 years has been capable of capturing high resolution all the time. For Christ's sake, this thing is from 1996, a time when not even recording studios were using 20 Bit regularly! The Kenwood DM-5090 was equally surprising since its digital output delivers a true 20 Bit signal. A) for a consumer device it would make much more sense to transmit with 16 Bits only and B) my resulting files were actually 20 Bits inside a 24 Bit .wav file (as indicated by the Bit-meter in WaveLab). Look at the results:

Entirely digitally transmitted 16 Bit test signal yields 16 Bit performance using ATRAC

A 24 Bit test signal yields 20 Bit performance thanks to WideBitStream

For comparison: mp3 320 kBit/s CBR (Lame 3.98.4) with 20 Bit output

Sony told the truth indeed: even a small portable MD recorder is capable of recording true 20 Bit quality (Dynamic Range would need to be roughly 10 dB better though). The 16 Bit signal results in 16 Bit quality while the 24 Bit signal yields 20 Bit quality. It would be interesting to measure the output of an ATRAC 4.5 equipped device since that particular ATRAC version was the first to use full 24 Bit processing. Furthermore you now can see how mp3 behaves when it´s properly decoded, allowing it to deliver the full resolution of its internal 32 Bit floating point data (for fairness decreased to 20 Bit in the example above). The mp3 measurment also serves as an example how much lossy codecs have improved during the last 16 years. Granted, results are a bit unfair since ATRAC uses 292 kBit/s whereas the mp3 test signal above was encoded with 320 kBit/s. The most important measurment for all three examples is Total Harmonic Distortion plus Noise (THD + Noise), ATRAC measures 8 dB worse than mp3 which might lead to more audible compression artifacts. But considering its age, its limited processing power and implementation ATRAC fares impressively well. Of course, with every lossy compression scheme only the output with real music counts so measurments might not reveal any sonic shortcomings. For that I not only recorded test signals; I also copied several music examples the same way as described above to find out if ATRAC sounds different to the originals. To determine if there was an audible change was indeed quite difficult, proof to how well Sony designed the whole process. It could very well be a placebo effect but I was under the impression that ATRAC widens & deepens the stage impression very slightly leaving tiny "holes" and a not so stable stage. It also seemed to sound either too harsh or too mellow depending on the material - but keep in mind that the perceived effects were so slight that I´m not sure I´ve even heard them. Impressive, don´t you think?

Sony MZ-R 30

But how does the MZ-R 30 sound? I´ve already talked about this before - but I didn´t rate its sonic quality with my new rating system. The sound of the MZ-R 30 is fairly balanced, blowing up upper bass while missing high frequency detail. Despite this it displays slight loudness-switch characteristics, emphasizing bass and treble frequencies around 10 kHz, efffectively draining sonic colour. Dynamics seem to be held back with the result that the MZ-R 30 always sounds "nice"; bass is soft and misses punch, transients in general are too soft. Staging is fragmented yet flatter and convoluted, guessing size and position of instruments is becoming difficult for the listener. It´s like hearing a fuzzy bubble without much depth and width. What suffers most however is resolution; details are rendered far to casual, more often than not high frequency detail turns into simple noise. High frequency intelligibility really isn´t very well rendered on this device, yielding a voluptuous and too warm sound. Don´t get me wrong, it sounds moderately well, definitely better than 40% of my portable players. Without colouring the sound too much it presents music very pleasant, you can therefore listen with it for extended periods without suffering from listening fatigue. Its headphone output on the other hand - although powerful enough - sucks big time. Using it one looses much crispness, it sounds surprisingly dull.

Sonic Balance:

Dynamics:

Resolution:

Stage / Ambiance:

Character:

A personal Note

Of course it´s pointless to use such an ancient device in present times. Recent portable players can be filled with music in a matter of seconds, offer theoretically improved sound quality (using lossless FLAC and improved technology) and are much smaller in size while offering reasonable running times. With musical gadgets convenience always is key while quality plays a secondary role. Therefore technological obsolescence plays a major role in the death of the MiniDisc. 

But there are other reasons as well... up above I´ve talked about CD-R - since they were so expensive Sony estimated it would take a long time for them to become affordable. They were wrong: recordable CDs became cheaper than MDs in 2000/2001, rendering the MiniDisc an inconvenient and slow recording device with a compromising quality approach. The mighty giant Sony reacted with altering the MD format by introducing an advanced version of ATRAC called ATRAC3 (not to be confused with ATRAC 3) which allowed longer recording times by lowering the datarate (64 kBit/s & 132 kBit/s). Needless to say that such recordings were incompatible with older MD devices. 

Around the same time a second problem arose: portable mp3 players. Sony tried to counter these developments with another new feature released in 2002 called NetMD where you could - using your PC - upload your music to a recorder equipped as such via a built-in USB link. Of course, you had to use the proprietary ATRAC compression scheme, meaning that you had to re-encode your collection of mp3-files, lowering the quality even further (lossy re-encoded to lossy). For all consumers it was as clear as day that this was highly inconvenient and extremely impractical - but Sony found nothing by it. Subsequently many consumers parted with Sony and the MiniDisc. 

Kenwood DM-5090 - an MD recorder from the systems' heydays - and not even from Sony

The last nail into the coffin of MiniDisc however was the release of the iPod in 2001 and its subsequent success. Again Sony completely misinterpreted unfolding developments and tried to revive the almost deceased MiniDisc corpse by improving the recording quality. They introduced the Hi-MD in 2004 - for the first time in its life the MiniDisc was able to record and playback music in true lossless 16/44.1 quality, all of that combined with NetMD "comfort". Yes, they dropped the Wide Bit Stream feature entirely. It´s ironic: to infuse the system with new life they not only sacrificed something that made the MiniDisc unique (admittedly: many people didn´t care) but also compromised backwards compatibility. Sony subsequently lost its "coolness", the MiniDisc was just one of many decisions made in the first years of the new millenium that stank of stupidity. The once shining company that produced legendary components now is just a mere shadow of its former self, making more profit with finances (!). Remember: Sony is at heart a manufacturer of electronics and not a bank. Today Apple is hip while Samsung is market leader, both traits were personified by one company just ten years ago: Sony.

Sony may die a slow death - but thankfully their components are still around and obviously built to last for quite some time. I for one have decided to go back to the MiniDisc system for now, even though production has ceased in 2011 and I myself abandoned it in 2001. Why am I so stupid and unreasonable? I don´t really know... but the first time I touched the Sony MZ-R 30 I knew that I was in love. This never happened when I received my Sansa Clip+, it was just another device playing music. Furthermore I never trusted mp3 - without a logical reason really. By now lossy codecs have reached a mindboggling quality, with higher data rates they in fact are transparent. Still, to me mp3 always sounded a bit dull and compact. I know, I know... completely stupid and very likely a placebo. But I can´t help myself, I always perceived the MiniDisc to sound better. It also looks and feels very nice, it has a haptic mp3 cannot reach since it´s immaterial. Some weeks ago I read an essay in the German news magazine Der Spiegel called "You cannot love files". This article described how soulless downloaded music appears to many customers´and it argued that today you can buy music in a second and have it on your portable player the next. But you cannot hold it in your hand (like a CD , MD or even Vinyl), leading to a feeling of being disconnected from it. You don´t have to occupy yourself with it anymore, it´s just data confined within impersonal files. I´ve always wondered why I "remaster" every CD I buy (which are many), maybe it´s because I want to "connect" with it before it ends up in pristine quality on my HDD where it´ll only be a collection of files. My recent interest in vintage portable players or the MiniDisc forces me to do the same, I have to bother myself with the music, I have to spend time with it. In any case I have not understood the reasons for this yet, I´m just starting. Maybe it´s much simpler and I´m starting to get old. Equally possible but you know what? I don´t care because I want to enjoy music in very good sound - and the MiniDisc allows me to do that despite (or because?) its shortcomings.