This month our reviewer Louis Challis was able to test and listen to the new MDS-JA3ES MiniDisc Recorder, the first of Sony's 'third generation' MD products. Its features include 20 bit performance, improved ATRAC digital compression/expansion processing and an impressive six second recording delay, and Louis predicts that it's likely to trigger an immediate re-evaluation of the way the market ranks current audio recording technologies...
Three years ago when I received Sony's first generation MZ-1 MiniDisc Recorder, I was impressed by its flexibility, convenience and by its friendly features. Notwithstanding my impressions, though, it appears that relatively few others shared my enthusiasm. In the ensuing three years, to its chagrin, Sony has faced a less than captivated market place. Sales of MiniDisc hardware, and more disturbingly the all-important software, have apparently been disappointing.
Around about a year after the release of the MZ-1, Sony produced its second generation MiniDisc decks which were smaller, smarter, and considerably cheaper than the first generation. But otherwise not much changed, and the response of most consumers might well be described as a big 'yawn'. I can well imagine how depressed Sony's marketing personnel must have felt at that time.
Now if it were you or me, we would most probably have dropped the whole concept and moved on to what we perceived as being greener pastures. That however, is not the way Sony responds. With an unbridled tenacity, its engineers initiated the development of their third generation MiniDisc deck as much as to say, 'we are right and you are wrong'.
If you've started yawning again, I recommend that you don't. It would be very wrong to think that this new-generation MD product is like its predecessors. The MDS-JA3ES is smart. It's precisely the sort of technology which can induce (if not seduce) a reviewer, or a purchaser to change his or her mind about what's smart, what's innovative - and most particularly, what he or she would like in his/her Christmas stocking...
As a serious user of sound and music recorders for more than 30 years, I guess I have had a better opportunity than most readers to evaluate almost all of the different types of tape and magnetic recording systems that have been developed during that period. Whilst I have seen a number of transient arrivals, and even more rapid departures of recorders during that period, none of them has had the attributes and performance capabilities of this current new generation of MiniDisc decks.
In June of this year I would have assigned the following ranking for the four most important consumer audio recording systems:
Sounds wacky? Well, I guess it might. And of course as always, there has to be a catch, when you think about it, how can any recorder anticipate what you want to do? Well, the catch in this case is that in order to achieve that goal, you need to have taken a previous step in which you pressed the RECORD button, and the recorder was waiting in the PAUSE mode. The recorder can then sit in that state for as long as you like, awaiting your activation of the remote control, or the button on the front of the deck. While it is waiting, it is continually cycling six seconds of data through what is in effect a modern day complex digital version of what used to be described as a 'bucket brigade' memory. In other words, it always retains six seconds' worth of audio in its recording buffer...
At first I was sceptical, and queried the objectivity of the claim. But after a few carefully contrived tests, I confirmed that this was no idle boast; the recorder does precisely what is claimed.
You can actually record the data you have just heard coming over the radio or your baby's first words, and avoid the need to sequentially record hours or days of a special occasion for a critical, transient 'never to be repeated sound', that previously you had no other convenient way of recording. I can't begin to count the number of occasions on which I needed that capability, and simply did not have it available in the recorder with the dynamic range, flexibility and convenience that the JA3ES now offers.
Many is the time I've wished I had this feature available, and Hallelujah - it's now available. It's also supported by a multitude of equally exciting features.
The JA3ES deck uses Sony's new CXD-8505Q D/A converter chip, which was developed for their latest generation CD players. This chip accepts a 20-bit digital data stream, which is converted by CXC-8505Q's first stage into an equivalent 22-bit signal, before passing it to the second stage which converts it into a stream of one-bit PWM pulses.
The PWM output of the chip is an audio signal with a spectacular dynamic range and a degree of fidelity which convinced me that any previous claims of lack of fidelity and commonality of quality between Mini Discs and CDs has now become a technically invalid issue.
Although Sony claims a theoretical 131dB signal to noise ratio, I believe that what they were actually aiming to achieve was the linearity of an 18-bit system (i.e., 108dB capability). On delving further, that belief was confirmed by my measurements.
The recording side of the JA3ES deck employs a CXD8539M Sigma-Delta over-sampling analog to digital converter. This chip incorporates a 'decimation filter' plus a high pass filter which effectively eliminates clock noise. As I will describe more fully later, I am satisfied that the system works as well as claimed.
The way Sony has achieved this is through the adoption of a canny floating point coding procedure, in which the position of the signal data is selectively scaled up or down by means of the DSP scaling feature.
Sony quietly acknowledges that this dazzling chip was developed in conjunction with another 'unnamed semiconductor manufacturer'. Precise comparison between ATRAC and a conventional linear system is difficult, because ATRAC converts signals on the time axis into the frequency spectrum before it transmits data. Using the highly accurate computational techniques of the floating-point method, the dynamic range is wider. The concept behind this is described below:
A 16-bit linear recorder eliminates the lower bits of middle order signals, no matter how wide the input bits are. This results in some degradation in the input signal. In comparison with this process, the MiniDisc maintains the precision of 16 bits as they are received, and raises the level if the lower bit signals are well down in the dynamic range, recording their maximum values as a SF (Scale Factor) function. When these signals (with their 6-bit SF data) are subsequently replayed, their absolute level is returned to the original (relevant) level, at the time of recording. Therefore, so long as they are correctly computed, and correctly scaled, the 16-bit precision resolution can be maintained for signals encompassing a 22-bit dynamic range.
When recording a low-level signal, a conventional 16-bit linear recorder eliminates signals whose peak levels fall outside (below) the conventional 16-bit dynamic range. When that occurs, you have no signal output. The JA3ES however, records those signals by increasing their level (and their scale factor), in order to incorporate them into a semi-conventional 16-bit signal range. On play-back, they are reduced to their original value but the overall dynamic range achieved is better than 18 bits, and based on my evaluation, most certainly approaches 19 bits.
The advantages of this data re-scaling approach are displayed on the attached graphs (Fig.1). As you will observe, by recording the scale factor with the data (and ignoring the impact of residual system noise which is still significant), the attributes are both positive, and as I later discovered, provide a superb unexpected clarity and fidelity.
As I delved further, I found other attributes in the system. The JA3ES incorporates improvements in its 'Bit Allocation Algorithms'. Although Sony don't provide a detailed and comprehensive description of this function, it appears that this improved algorithm achieves a higher bit resolution for both low and high frequency signals than is provided by the previous generation of MiniDisc recorders and decks.
The critical advantages of this algorithm are an improvement in the signal to noise ratio with a simultaneous reduction of both total harmonic distortion and noise. This was most certainly confirmed by my measurements.
Where the signals have a regular spectrum with a uniform signal density between the low and high frequencies, as typified for example by a squarewave, the remaining bits in the algorithm are allocated to the middle and lower frequencies. This provides an immediate perception of reduced distortion, and is tied in with the adoption of a basic 20-bit processing signal in the type CXD-8505 digital to analog converter.
The JA3ES also incorporates a direct quartz synchronisation system which is tied back with a conventional phase locked loop (PLL), to match any of the three different sampling frequencies with which the digital signal inputs may be encoded.
One feature that becomes very obvious when you start to use the MiniDisc deck is the ease with which it can be used for direct recording from a CD player, a DAT recorder or any other digital source. As long as the other digital source has either a conventional coaxial digital output or an optical digital output, the MiniDisc deck will accept that signal. More significantly, it will interpret the signal correctly, and will record its own signal with a 44.1kHz sampling frequency, independent of the original sampling frequency.
During the subsequent signal processing the deck will encode each successive track with either the original track number, or the track number corresponding to the incremental track number. If the source stops, or pauses, the Mini Disc deck will similarly stop or pause.
If the source material comes to the end of the disc or tape, the MiniDisc will similarly stop and await your instructions before proceeding further. It adroitly avoids all of those nasty functional problems which you and I have experienced, when the source material stops, and the interconnected recorder continues recording blank material.
The first, and less than obvious technical feature, is one which bas been adopted by a number of other Sony Extremely High Standard (ES) series consumer products: a neat and remarkably effective R-core Power Transformer. The 'R' stands for round, as this power transformer uses an elliptical cross section for the iron core.
The advantage of the almost round profile is that it facilitates the use of a high wire winding tension. The primary attribute of this approach is that it reduces transformer vibration. Sony also claims that it simultaneously reduces magnetic flux leakage, although I was unable to confirm this.
There are a number of other neat technical features visible inside the deck. The first was the extent to which special multi-lead ribbon cables and electrically screened ribbon cables have been used to interconnect the seven separate printed circuit boards and the MiniDisc drive. Some of these ribbon cables have as many as 30 separate conductors, and one of the boards has as many as six sets of multi-lead ribbon cables connected to it. The high technology in this deck is located in the LSI chips, and with one exception, they all appear to be manufactured by Sony.
An examination of the outside of the deck reveals that the only controls which differ to any significant degree from your existing CD player or cassette deck is this deck's adoption of the 'Large Multi Jog - Dial'. The dial provides users with quick and convenient access to the table of contents (TOC) display. Having selected the correct TOC track number, you can either play it, edit it, or over-record a single track or all tracks, as and when required.
The other relatively unusual controls and functions provided on the front panel are firstly the provision of a pair of conventional 6.5 mm diameter tip and sleeve microphone sockets, which will accept either a single or two microphones, to provide the inputs to the two channels. The second unusual feature is the provision of a 20dB input attenuator, so that either high output microphones or other inputs may be fed to the microphone circuit.
A view inside the JA3ES deck The main digital processing board is to the left of the MiniDisc mechanism (upper centre). Note the compact R-core transformer at lower right
With analog input the measured signal to noise ratio is -98dB(A), and -89dB unweighted. An examination of the noise level recording reveals that the only significant intrusive components are a dominant peak at 100Hz (-90dB), with less pronounced peaks at 200Hz and 300Hz.
The digital record to replay spectrum, is significantly cleaner than that using the analog input, as they contain no trace of mains hum or harmonic components at 100Hz, 200Hz and 300Hz.
An examination of the noise thresholds with our FFT analyser using an ultra-pure 1 kHz sinewave signal at -63dB reveals that the residual noise components are at least 50dB below the fundamental signal component, and are thus totally inaudible. The combined effect of the ATRAC system and the frequency scaling capabilities of the deck ensure that low level signals maintain their fidelity and an exceptional signal to noise ratio performance, even when they are 60dB below the peak recording level.
The crosstalk between left and right channels with analog inputs are -90dB at 100Hz, -93dB at 1kHz, and -76dB at 10kHz, which is a particularly impressive performance. The record to replay linearity is within 0.1dB all the way to -70dB, is 0.2dB high at -80dB, and 0.5dB high at - 90dB - a performance which few CD players can currently match.
The distortion characteristics with analog input remain at insignificant levels, which are basically below the noise threshold of the equipment (determined by the ATRAC encoding), all the way down to -90dB, at which point the residual system noise still remains the dominant factor. All other significant parameters including wow and flutter, and transit speeds were impeccable.
The first test was to record selected samples from a demonstration disc prepared by Sony, entitled 'Super Bit Mapping Demonstration Disc' (SBM 1). I used the coaxial digital output from my CD player to provide a direct digital input to the MiniDisc deck.
The SBM 1 disc contains some unusually low level test signals, whose transient peaks lie in the range -70dB to 80dB.
Signals whose peak levels are as low as that, and whose average levels lie in the range 80dB to 90dB, constitute particularly difficult material to rerecord with an adequate signal to noise ratio. To my surprise and pleasure, the JA3ES had no difficulty in achieving that goal, and we were incapable of detecting which was the original CD and what was the copied MiniDisc material in a double blind test.
Reference Series Disc (10063-2-F). This disc contains superb transients and exciting music from both the violin and the piano. We simply could not pick the difference between the original and the copy.
We progressed to a third test disc, which I have previously used, featuring Yo-Yo Ma and Bobby McFerrin in 'Hush' (Sony Masterworks SK481 77). For the comparison however, I used a Sony Masterworks pre-recorded MiniDisc (SM481 77), incorporating the identical tracks. Again my test panel were unable to pick which was the CD or the MiniDisc, and my esteem for the new MiniDisc deck incremented by yet another notch.
The last disc which I recorded was Ludwig Van Beethoven's 'Works for Chorus and Orchestra' (Koch Schwann 3-14~5-2). This disc provided us with some excellent choral material, and although the individual voices were less effective in terms of vocal identification, we were still able to draw the same conclusion namely that we could not tell the difference between the original CD and the copied tracks on the MiniDisc.
Each of these discs contained superb musical content, and epitomised the wide band vocal, musical and transient test signals which tested our hearing and our discrimination, as well as the equipment, whilst performing multiple direct A-B/A-B inter-comparison tests. Those tests used a time lag between CD player and MiniDisc player of three seconds, so that we could repetitively switch from one source to the other, and hear precisely the same material.
We ran this A-B testing during two separate sessions encompassing a total period of three hours. After the testing was completed we were satisfied that we could neither identify, nor could we hear any difference between the digital original and the digitally recorded MiniDisc, or the manufacturer's own pre-recorded version of that same disc.
After carrying out numerous additional recordings of my own (on my own), I came to the conclusion that the JA3ES is currently the most convenient, and outstanding digital audio recorder (deck) that I have ever had the pleasure to use, or to audition.
With a selling price of $1999 it is certainly not cheap. However, when I review its performance, its convenience and its special capabilities, and most particularly that magical six seconds of pre-recording time capability, this is undoubtedly the most outstanding music and consumer recorder that I have ever used.
The Sony MDS-JA3ES Mini Disc Recorder Deck measures 430 x 345 x 125mm (W x D x H), and weighs 6kg. It comes with a IR remote control unit (RM-D2M), operator's manual and cables.
Further information is available from the Consumer Products Group, Sony Australia, PO Box 377, North Ryde 2113; phone (02) 9687 6666 or fax (02) 9887 4351.