You are reading the older HTML site
Positive Feedback ISSUE 65
The Memory Player 64:
Mark Porzilli's Memory Player Grows UP - Part 2
Hot news. Just got some
review notes from a fellow audiophile......
And this was only a partial evaluation of his latest and greatest front end. Yes, the superlatives for The Memory Player 64 from Mark Porzilli/Laufer Teknik are impressive, indeed. The descriptors above, however, relate to my friend's Ortofon MC Anna, his new cartridge.
All of which points to stunning advances made in analog and digital realms. In both areas, we are able to hear music closer and closer to the real thing. Will we ever get there? Of course not. Is the race towards perfection exciting? You betcha. And who is closer... state of the art analog, or state of the art digital? Ah, as the Bard said, "That is the question..." Which should be our path to nirvana? Where should we put our hard-earned bucks? And should our slings and arrows be directed at competing technologies, or should both flourish and enrich our pleasure in recorded music?
As the title of this review suggests, this opus will concern itself with the latest Memory Player. (Hey, is Wagner Wacko? Is he really going to liken Laufer Teknik's The Memory Player 64 to the Hubble telescope and to the finest of Impressionist paintings? Read on.)
In PF Issue 57, I promised that future iterations of the Porzilli-designed Memory Player would offer simplified operation while sonic performance would be maintained at the high level achieved late in 2011. Half right and half wrong. Yes, operation of The Memory Player 64 has been simplified in significant ways. However, and more importantly, the sonic performance has taken a giant leap forward.
I will relate my impressions of the sonic improvements a bit later. Additionally, just how The Memory Player 64 has been refined will be explained further on in this review by the designer, Mark Porzilli, in an interview with this author.
But first, let's take a look at the improvements in operation. Earlier versions of The Memory Player 64 read the CD in RUR (Read Until Right instead of Reed Solomon ECC) and placed the music in a digital archive. After a few drag and drops, that data was scrubbed and polished by IDEAS (Impulse Discharge of Events in Atemporal Space) and made ready for play. A bit of a nuisance, but even for a computer klutz like me (I grew up in the wrong generation), that process took about 45 seconds. Not good enough for many who have heard and admired the sound of The Memory Player 64. They wanted simpler operation; you know, 'punch and play'.
Not an unreasonable request. The Memory Player 64, today, incorporates a 'Juke Box', operating much as the one you may recall from 'coffee and' at local diners after a movie, or even what you might have fed quarters into at an Army PX in years gone by. Punch and play. Of course The Memory Player's Jukebox does not eat your pocket change and its basic memory will hold up to 500 Redbook songs, far more than its ancestors.
Changes to the songs in your Jukebox are easily accomplished. Jukebox storage in The Memory Player 64 is virtually unlimited except by one's pocket book. Larger Jukeboxes add to the cost. The Jukebox can corral all manor and sort of musical data... Redbook, hires, stuff from the clouds, even up to 64/768 if that ever comes along.
Before I go on to describe the remarkable sound produced by The Memory Player 64, (and finally get to the 'Hubble and the Impressionist painting stuff'), may I introduce my interview with Mark Porzilli, who will describe the history and design philosophy of The Memory Player 64 in his own words:
Besides income, what drives your continued interest in the Memory Player? In particular, I would ask you what it was that convinced you to focus so much creative energy on lowering jitter, and lowering it well below the generally accepted industry levels. Are you still seeking to lower jitter further?
Back in the Melos Audio years (1979-1999), we lived through the birth and early years of digital audio and the concurrent backlash in the form of the vinyl LP craze. To those of us who are old enough to remember, that period felt apocalyptic, as if the supply of good musical recordings was ending forever.
Many of those of us who disliked digital audio simply turned to LPs almost exclusively. But it would be disingenuous to suggest that we remained steadfast in resisting this new but awful sounding medium. Almost all of the manufacturers designed their version of improving digital audio sound. At Melos, we designed many tubed players and tubed DACs.
But the CD actually ignited a three-decades long mission to improve its sound. Admittedly, I was one of those vinyl hoarders... just in case!
However it wasn't until around 2004 that we tested our first "memory playback" device. Everything changed for digital audio when memory playback technologies surfaced.
The idea behind memory playback was, in its most generic of definitions, that if we copy all of the bits to a piece of memory large enough to not fragment the music files, we'd have the time and accessibility to clear it of jitter.
One of the paramount virtues of the CD was, the bit quantity being ensured by ECC. When you play a CD, you'll hear all of the bits. However although this is what is accomplished through ECC (Error Correction Codes), never properly addressed the problem of jitter. Of course 'bit perfection' ensured that all of the bits would be there, but what was profoundly different was that we would have the time to apply multiple jitter reduction methods that required far too much time to apply in a real time player (any CD player that plays the physical disc) .
There were always some dissenters, besides us, who were unable to agree on what the limits of the term 'bit perfect' were and/or should be. For many to this day, remarkably, it is a panacea in the most global sense.
The ignorance cited above means little though. The primary ignorance of ECC is it's power limitations and it's byproducts. It is not 'free'. Yes, ECC recreates 100% of the lost bits, but the sonic cost is annoying to catastrophic. ECC has no timing mechanism. ECC is always late. Late Bits (being played late) = JITTER.
But others looked beyond ECC. Without getting too encumbered by the verbiage itself, digital audio code theory is about determining the location of a bit in a given time frame. In code, it's known as the Probability of Vicinity. Being knowledgeable of computers offers literally little advantage here. It's the shell of the egg. It's physics time.
However others did and do understand the theory and physics of digital audio. Others looked beyond ECC, knowing that ECC will raise jitter to intolerable levels.
The most prominent and respected approach to the problem was MIT's venerable "Paranoia" rereading code. Unconvinced that the bit perfection attained by ECC was a panacea, they wrote the most important CD rereading code of all time.
We embraced Paranoia in the beginning, and then wrote our own code later.
Called Read Until Right (RUR), it used expanded powers in the control of the laser and the rotation of the CD. I won't get into that right now, as it's not relevant, but I raised the point because the critics of rereading, or other alternatives to ECC, attach a significance to ECC that it never made claim to.
And speaking of claims, if anyone tells you that "ECC is almost never used" I have sixteen words for them: Measure the jitter on the outputs of the Reed-Solomon decoders, not just the audio outputs!
Most other error-correction devices use ECC as their kernel, even for basic tracking. It's the cornerstone of the playback process. It's not that ECC does not work, it's that ECC does not show us the entire picture. After all, ECC is a BURST code so anyone who understands the audio codes, knows all burst codes will almost always raise the jitter on the bit level. I feel that this is the inevitable first step in designing a progressive, digital audio device. One MUST understand the AUDIO CODES, not necessarily computers. Anyone can design a "computer audio" or "music server". But no great progress can be made in this way. One must understand Reed-Solomon, Hamming's, Parity et al. If not, it's just a computer playing music.
The ECC codes used on CDs and DVDs do not have the ability to resolve information on the bits, only the number of bits. If bits are lost (dropped bits), the bits are replaced. However, one must be mindful of the audio code's limitations.
The replacements can be late, inverted from noise, and/or even out of order (all sources of jitter) and still, 'bit perfection' will be reached.
So, it became an imperative that we include this 'blind' part of the process in our analysis for jitter. To that end, we arrived at a purely theoretical position of the existence, or the predictability, of two origins of jitter. The first one is output jitter, which is commonly read at the DAC and has become more or less the standard, and the second type is jitter which we can predict, but have not been able to directly measure, called "packet jitter".
If one understands the code, not digital audio engineering, not computers, but the digital audio code itself, it becomes apparent that the opportunity for even very high jitter during playback enjoys the sunny side of the probability curve. The reason is that a different form of jitter takes place whenever any digital data is moved from one point to another.
At this stage, the data has not yet been converted to the familiar format of: 16/44, 24/192, etc., but rather the data is in the form of a stream of bits held inside a "packet". The contents of these packets are not transferred individually, but rather the entire packet moves in accordance with each tick of the master clock. Therefore, no amount of improving of the clock's accuracy can help reduce jittering of the bits inside of the packets because the clock ticks for the packet, not it's contents. This is because this type of digital data being moved to any destination, is a burst code which by nature, it is blind to bits within the packet (or a byte).
The concept behind what is called packet jitter is well known in other areas, such as computer networking and high-speed internet (as in QOS), but it has yet to contribute to the evolution of digital audio, as far as we know.
Recently, Clement Perry wrote about us in his latest review of the MP in The Stereo Times, where he stated, "performance and measurements are on his side... I don't know what the future holds but I know who holds the future". We can quantify packet jitter and it can be measured, but not with a casual connection to the audio output, but rather, in stages. The jitter in two, possibly three stages must be measured, raised to an equivalent platform and summed, in order to quantify all of the system's jitter.
The new MP's subjective results outperformed everyone's expectations tremendously. Really, none of us knew it would be this audible. After all, 700 femtosecond jitter is pretty short, but literally everyone who tested it heard it, and of course, we measured it. It was about that time when we received a nomination to digitize the Lincoln Center tapes after their engineer auditioned Fractionalized Memory. He was amazed and very honest when it defied the bit perfection theory but required a complete analysis to present to Sony before they adopt the fractionalization of memory.
So we feel we've opened a world of possibilities to not only narrow the gap between digital and analog musical recordings, but to a potential to expand digital audio recordings to a level of fidelity that is a literal copy of the original master. It's not solely about bit counts, but bit timing, bit ordering, bit and byte jitter, decompression of the packet data to a given audio format, these all take time, jitter...in all aspects. No, we're not there yet. But unlike bit perfectionists, we are looking. We drew the map.
What improvements are being considered?
As you know, we have a Remote Repair system in place. I think that the fact that digital audio components become antiquated every six months or so has always been a significant negative, creating an environment that can be both inconvenient and unaffordable. One of the design requirements was to defy becoming antiquated by using programmable firmware and software throughout, and the internet.
For example, through our Remote Repair system, the MP can be programmed to play any music file with a peak resolution of 64/768, if digital audio is ever released that high! For the foreseeable future, it seems to top out at 24/192 and the similar SACD rips reaching 384.
And would those improvements be in the form of freeware, installed remotely from Los Angeles?
Every MP owner may enjoy free repair, and both free and purchasable upgrades done entirely over the internet. We already keep the MPs around the world up to date and at peak performance remotely, even programming the chips remotely to become more advanced, with no shipping and down time, taking only hours instead of weeks. The intent behind this is that the MP owner will never have to be concerned that he's stuck using old technology. The MP64 should never go out of date.
Remote Repair has been fantastic for us. In 2012, we only saw three MPs shipped back, over the entire world! The Remote Repair system has enabled us to provide a powerful and convenient way to support customers' systems, preventing obsolescence, and avoiding down-time and shipping costs.
What improvements are under development?
We are planning to offer "IDEAS for Audio", an internet-based service which will be available by subscription. One will be able to log on and reduce the jitter of his music library dramatically. Owners of music servers or even computer audio collections, may log on and run IDEAS on their music collection, with often astounding drops in jitter. If the player plays files and connects to the internet, IDEAS for Audio can reduce jitter to spectacularly low numbers, with often transformative sonically results.
What single aspect of MP design is most responsible for its excellence?
Well, in addition to the above, we would be remiss if we did not briefly discuss how we were able to negate the importance of the master clock's accuracy. In doing so, we surpassed the jitter levels of the finest atomic clocking players that retail between $70,000 and $80,000. We call it Fractionalized Memory, and it's about to be awarded a U.S. patent. You may read the patent from a link on our website.
Fractionalized Memory copies the music data to a memory structured to group the data in packets 256 times larger than a normal byte.
Since IDEAS has de-jittering power deep into sub-byte, at the actual bit level, we can clear the memory of jitter within each packet. However, The Memory Player enjoys a powerful advantage of requiring only one clock tick, whereas a conventional system will require 256 clock ticks. The probability of having a perfectly accurate clock tick 256 times as opposed to one time is far, far greater, with more than 256 times fewer opportunities for jitter. It yields measurable and simply transformative sonics. Fir the first time in my experience, is confused with analog. For me, the obvious second highest compliment!
This is why atomic clocking and reclocking devices hit a wall, since the clocks tick for the byte, not the bits inside of the byte (Or if earlier in the chain, it ticks for packets). They do reduce jitter, but only peripherally. The timing and speed of all of the devices in any audio player cause jitter. And if the jitter is in a packet (downloads, CD rips, etc.) the new original file has jitter baked into it; jitter that no clock can see, and therefore cannot remove.
Fractionalized Memory is so powerful, I can't believe it will not enjoy ubiquity once it's well understood. Almost anything played on Fractionalized Memory sounds better, regardless of player type or clock accuracy. Fractionalized Memory is so powerful, that by simply placing the music file on it and playing it, the global jitter can drop by 200% - 300%. In fact, in blind tests, still pictures and digital video files appear higher in resolution when displayed on it!
Are there other aspects of MP64 design you would like to comment on?
For listeners who are skeptical about how this works, there is an inexpensive and easy way to hear the jitter created by ECC. A simple 'self-test' for audio jitter may help people learn the sound of jitter by adding jitter created by ECC as opposed to removing it.
Burn two CDs on your computer of any music you use to test with. Burn them identically, using exactly the same type of blanks.
Now, use sand paper or an emery board and lightly scratch one of the CDs on the laser side where the information is written (opposite the label side). Don't scratch it too much, or it may not play at all! Just with just some scratching so you see light scratches all over the disc. Blow the dust off.
Listen to both. One will sound hard in the midrange and harsh in the treble, and even bass may become blurry and poorly defined. What you are hearing is jitter created by ECC, trying to fill in the lost bits where the scratch appears. Both have exactly the same amount of bits, ECC filled in the bits you scratched away. What you are hearing is ECC jitter.
To see how ECC, noise and bit-reordering can affect the bits in the shell of a byte, please see the following illustrations.
First Jitter Source: Uploading of the music to a cache, so the ECC may analyze and recreate lost bits on the cache, takes time because uploading takes time. Even the cache itself has a speed limit (slew rate and settling times), which makes it take still more time.
Second Jitter Source: A bit may be read out of order within the byte once ECC 'repairs' the missing bits.
Third Jitter Source: Finally, although it is not as common, but occurs frequently enough, there is unidirectional bit inversion (0 to 1) which can be caused by noise (house power, radio, appliances, internal RF and magnetic fields in the player, etc.) where the noise fills up the '0' where there should be no sound.
So, if bit perfection had indeed been comprehensively 'perfect', then why would it have sounded different at all? Why would the two CDs in the test above sound different? Why do green paint on CDs, anti-static sprays or anti-microphonic feet and many other treatments and tweaks often improve the sound? Not just on players, on digital drives with no analog to treat.
In addition to the above, there are some obvious myths that impede the path towards inaudible jitter. Contrary to the rhetoric, electricity does not move even close to the speed of light through transistors and chips. In addition, nothing is instantaneous in mathematics. These codes are math, and mathematics abhors generalizations. The speed limitations are determined by the speeds of the transistors and chips, and even these vary, depending upon the state of duty that they could be in, at any given moment. Even FIFO (first in, first out) are byte codes, which cannot ensure FIFO for bits. Another myth is that we've already reached inaudibly low levels of jitter. It's a total cop-out, because as low as we've gotten jitter, 1000x lower than some of the most expensive players I've seen reviewed, it's still not as low as we can hear. We know there's much more ahead.
There are at least three different sources of jitter in digital audio players of all types, and only one has historically been addressed. To my knowledge, the only globally comprehensive jitter reduction technology that examines and repairs jitter all the way to bit levels (sub-byte) is Fractionalized Memory. And when used in concert with a jitter reduction software suite like IDEAS, it can reduce the total jitter in the system to the femtosecond region.
Although ECC replaces the bits that the sanding obscured from the laser, it also jittered them. Packet Jitter and ECC take time, and Time = Jitter.
So, there you have it, missing the proprietary details of course, but otherwise the design philosophy of Mark Porzilli as it relates to The Memory Player 64. Were you able to grasp all of what Mark wrote? I was not and I suspect I have a lot of company. Even those of you who are quite knowledgeable may be confounded, even made skeptical by some of the ideas presented. Especially since a good deal of what makes The Memory Player 64 sound as good as it does is based more on mathematics than on computer savvy. In particular it was mathematics, rather than computer knowledge, that convinced Porzilli to go to '64'. It's no sales gimmick. Mark explains it this way, "The virtue of using 64-bits is to be able to narrow the space and time where a bit should be, to a precision that makes 16-bit look silly. With 16-bits, we can only look in 1 area per bit. With 32-bits we can look at 17 million areas instead of 1. With 64-bits it's 290 trillion. A bit in a given space, at a given time. That's the whole reason for 64-bits."
If all of the foregoing makes your head spin, join the crowd. May I propose a simple solution? Listen first. Tech talk later. I've done the listening... months of joyous listening. Still trying to grasp the technology.
As you can imagine, Porzilli's restless mind continues to improve the software of The Memory Player 64—improvements that arrive from time to time by internet. How nice to wake up in the morning to find that Santa Mark has updated your player with his latest and greatest insights.
How best to convey what one might hear from Laufer Teknik's The Memory Player 64, now sporting the many software refinements described in Porzilli's text and including solid-state memory?
First I should explain that The Memory Player 64 is a very functional but otherwise simple box housing a transport, a DAC, and output stages. One needs only a pair of interconnects of choice and an amplifier and speakers to make music. To be more specific, you may use balanced and/or single-ended outputs, USB ports, Ethernet input, etc.
The music you hear, depending of course on the quality of the recording, is capable of coming as close to the sound of live music as I have ever heard. At times, for example, a simple recording, 'Tico Tico' from David Chesky, makes your room a stage with some of the most believable sound you could hope for. Other Chesky recordings like L'Histoire de Soldat (A Soldier's Tale) detail the voices and instruments perfectly while presenting awesome depth, space, focus nuance and dynamics. Deep bass, in Desplat's 'Birth', EMI cut 11, is rendered with ease.
Here's another goody... the New Music Consort-Pulse (New World 80405-2). If you're into classical music of more recent vintage, the chamber and percussion pieces on this disc are near perfection when played on The Memory Player 64. I sense the recording was made on a budget because one can hear the simplicity of the miking. If a console was used, it surely must have been simple as well. The percussion, drums through to triangle, is convincing beyond words. Trust me on this one.
Ditto for Winston Ma's First Impression Music's version of "All Star Music for Percussion". This one is lots of fun... Bizet, Beethoven and Berlioz in bells... absolutely comes to life... wonderful decay into the space between the instruments.
To be fair, other players can do some of these things well. Yet there are many special qualities of sound from The Memory Player 64 that I have not heard prior: Tonal purity, the quality of space, and ultimate quiet in particular. In these respects, I have heard nothing that matches The Memory Player 64.
Tonal purity, the presence of the clarinet (and, oh, those drum sets) in Jazz at the Pawnshop. The clarion voice of Anna Moffo in her 50-year-old RCA recording. Heifetz's violin in the Sibelius concerto, also on RCA. (Note: Here I'm citing the tonal purity of a violin on a digital transfer from an analog recording made with tubed electronics on tape. What could be a more demanding test for a digital playback system?)
One of my favorite recorded performances is Schubert's Trio, D929, played by the Beaux Arts Trio on Phillips. This is a CD I carry with me to shows and to friend's homes, hoping to hear it come to life. In the past, it has never quite made it sonically, although the quality of the composition and the performance has made it a favorite of mine. The Memory Player 64 has made the necessary leap. Bernie Greenhouse's cello, Daniel Guilet's violin, and Menahem Pressler's piano now sing. It is no longer just a favorite performance. Schubert's music has come to life. It has presence. Beautiful.
And to build a case that I am not a musical snob... try "Animali in Marcia", Gianluigi Trovesi Nonet. Round About a Midsummer's Dream. Cut #9. An international hit. Betcha can't listen just once.
Now to the qualities of space and quiet—(and finally to the Hubble Telescope and those Impressionist paintings.)
The Hubble telescope operates in outer space. Terrestrial telescopes, on whatever mountain of whatever height, must reach through some of the earth's atmosphere. Just as earthbound telescopes struggle to minimize the deleterious effects of errant air, even very good players have not reduced jitter and other artifacts to the degree achieved by The Memory Player. In Porzilli's interview he cites the levels to which jitter has been reduced, levels far below those considered indiscernible by the listener. And yet those 'indiscernible' levels can be heard as a subtle hash that fills the space between instruments. In a live acoustical setting, that particular hash is not present. (Unless, of course, you note coughing, an overly energetic air conditioning system, or a subway rumbling underneath.) Assuming ambient sound is under control in the recording venue, one hopes to hear music, the music minus the hash, when playing it back at home.
But wait a minute, there is more to music than just the music from the instruments. If the removal of jitter and artifacts leave the space between the instruments uncluttered with detritus, what can fill that space? Ah hah—that space is filled with all the harmonics, echoes, reverberations, etc. from all the instruments along with the sound of the recording space (the concert hall) itself. That space should not be sullied with digital artifacts and hash, however subtle. The Hubble telescope renders outer space as no terrestrial telescope can. Likewise, The Memory Player 64, reproduces recorded music as none other.
Please bear with me as I carry this idea further. You see, the space in a recording operates much as negative space in an Impressionist's painting. Sure, the painted image has subjects—trees, figures, etc—which function much as instruments in a recording. But that same painting has space and shapes between those objects, sometimes even the bare canvas left exposed (Manet), the configuration and textures of which are a vital part of the finished work. So, too, in a recording, the negative space, between instruments, is a critical part of the end result. The Memory Player 64 reveals the beauty and intricate detail that lie in that vital space between the instruments
Now, finally, to the cost of The Memory Player 64, list price $17,950. That includes a 2 terabyte archival memory which swallows 1000 or more complete Redbook CD's, plus 64 gigs of 'Juke Box' memory for approximately 500 songs. (Of course hi-rez can be accommodated as well.) As mentioned before, more gigs of Juke Box memory is available at extra cost. The Memory Player consists of these 'memories': A CD transport, a tubed DAC, and a tubed output stage plus a host of other features. One needs only a pair of high quality interconnects to reach your amp/amps. Done.
Suggestion: Put the bite on Sam Laufer to help with the cost of an iPad to operate The Memory Player 64. Then settle into your cushy center seat and listen to what is likely to be a stunning recreation of the music you love as you have never heard it before.