This section of the Audio Crafters Guild site is for discussion and sharing of ACG's and Guild member's research activities. If you have information to share let us know and we will add it here or a link to your site.

The inactive 'R&D' link holder was on the very first ACG home page, content was a long time coming! It seems I would rather work on my bench than write HTML about it. Also much of the R&D type contributions from our friends has already found its way to the Articles & Essays page. And my best thinking about CD format digital audio reproduction took the shape of X-DAC 3.0 and X-DAC 3.0 Signature. After a period of maturing the CD standard technology we find ourselves entering an exciting new age of DVD enabled 'Super-CD' formats with higher bit and sample rates. As the proposed formats wisely allow coding at resolutions exceeding the current state of the art it is time to put on our thinking hats and get busy advancing the state of our art.

ACG's R&D project X-DAC 3.24

See August 2000 & April 2001 updates below.

The 96k/24bit upgrade for X-DAC 3.0 is up and running as a development prototype. A December 1999 update of activity can be found in this copy of a posting I made to the 'Joenet' mailing list. While this R&D X-DAC 3.24proto is not the form the 24 bit upgrade of X-DAC 3.0 will take it is representative of how X-DAC 3.0 is upgradeable to support post-CD formats and the direction our research is taking. The production upgrade X-DAC 3.24 will (of course !!!) be much neater in execution and use later generations of chip sets. (See CS8420 & August 2000 update sections below.) This section can also be of use to those who are planning to update other DACs to 24/96k status as a source of ideas.

Step one is to get the 24/96 S/PDIF data into X-DAC 3.24. The first solution was the CS8414 DIR chip. Crystal made it easy as the pinouts are the same as the CS8412. A minor difficulty is the CS8414 is only available in a surface mount package. As the following picture shows this can be corrected with a steady hand by soldering the SMT part to a 28-DIP sized component carrier. Tiny gauge buss wire, Teflon tubing, a magnifying glass and some fine workmanship and soon the CS8414 is ready to plug into a CS8412 DIP socket. Hardly ideal and far from elegant, but it works.

The CS8414 allows up to 24 bit 96k data to be received into X-DAC 3.24. When this R&D X-DAC 3.24 was being put together sample rate converters which support 24 bit 96k data were not available. Thus the X-DAC 3.0 premier feature Asynchronous Jitter Reduction had to be temporarily abandoned by jumpering data across the AD1891 socket and disabling the low jitter local clock in favor of the PLL recovered clock output by CS8414. In the picture below the dark red component carrier in the center accomplishes this while leaving the test DAC easily re-configurable. The mini-coax carries the 256 Fs master clock from the X-DAC 3.0 expansion header to the CS4390 DAC.

For the DAC a CS4390 24/48 DAC is overclocked to 96k. This plugs into the socket usually occupied by CS4328 on X-DAC 3.0. The CS4390 adapter also contains sockets for balanced input diff-amps. In the photos the diff-amps are removed in favor of the transformers I have been playing with. These are in the silver cans their long leads left untrimmed, these are much too expensive to risk cutting them too short before the design is finalized! Left long they also contribute to the nuvo rats-nest look of X-DAC 3.24 and its outstanding radio transmitting and receiving characteristics. J

Future plans for X-DAC 3.24 R&D project include:

• CS8420 DIR/SRC to allow the return of Asynchronous Jitter Reduction.
• Local clock evaluations for use with AJR.
• A 'DAC shoot-out' among the choice 24/192 DAC contenders.
• As designs firm up go from hand built rats nest circuits to PCBs.
• A new trick or two in power supply and regulation.

August 2000 update

The circuit design and PCBs for X-DAC 3.24 are complete. The design is released to the circuit board fabrication shop with delivery expected 1 Sept. 2000. Here is the board set (note that converting PCB layout files to GIFs gives some traces a rough appearance not present on the actual design):

X-DAC 3.24 upgrade board set

The X-DAC 3.24 board set consists of three designs executed as the XD4396 and XD1853 DAC boards and the DIRMA Input Receiver/Sample Rate Converter board. The functions these provide are:

• XD4396 implements the Crystal CS4396 or CS4397 DAC chip with analog stage based on the Burr-Brown INA103 audio differential amps.
• XD1853 implements the Analog Devices AD1853 DAC chip with analog stage based on their AD827 high speed op-amps.

Common to both XD4396 and XD1853 is support for 24 bit samples at rates up to 192k Hz. Both feature an advanced new power supply regulator concept I am going to be a bit secretive about until testing is completed. Both plug into the CS4328 socket on X-DAC 3.0 (or other CS4328 equipped DACs for the truly adventuresome). If other analog stage designs are preferred the DAC chip analog outputs can be accessed.

• DIRMA stands for Digital Input Receiver Multifunction Adapter. Plugging into the CS8412 and (optionally) AD1891 sockets on X-DAC 3.0 (or other CS8412 equipped devices) it duplicates and extends both the performance and functionally of those two chips. The DIRMA  front end processes double speed 24/96 S/PDIF serial or I2S parallel inputs and outputs, optionally upsamples 16/44 formats to 24/96, and performs Asynchronous Jitter Reduction. The I²S capability not only allows interfacing with the few devices presently using that format as a inter component interface but more importantly provides a port for future post-CD interfaces such as I²S,  I²S· enhanced, or IEEE-1394 Firewire. Note the small circuit board next to DIRMA above is a manual programming adapter used to select its many options. It has to be mounted remotely to fit on the X-DAC 3.0 PCB. For more details on DIRMA's possibilities see the CS8420 section below.

All three boards share the option of sending the digital data and clocks via coax using SMB format connectors or using the X-DAC 3.0 circuit board traces. This allows many potential future options and intends to investigate if using coax for such short intra device runs is beneficial. All also use high grade passive components. These an exciting mix of known audio specification parts such as WIMA and Black-Gate along with the latest Surface Mount Devices. SMDs crept into this design both of necessity (latest digital audio ICs are all SMD) and performance reasons (more compact PCB and lower parasitics with SMD passive parts). I am very anxious to hear these especially the new plastic film SMD caps from Panasonic. Note how using SMDs AD1853 fits a regulator, DAC, and 6 opamps analog stage into a board with only 6.5 sq. in. (42 cm sq. ) of space! You want compact direct layouts? We can give you compact direct layouts!

Once the board sets arrive from the PCB house the plan is of course to get all three built, de-bugged, tested and online ASAP. Then the fun really begins investigating how all this technology impacts our beloved music. High points of the testing include:

1. Checking the many DIRMA modes including upsampling vs. direct data transfer and Asynchronous Jitter Reduction of 24/96 material.
2. The DAC Shoot-out pitting XD4396 vs XD1853 for the crown King of the 24 Bit 192k Hz DACs.

Recall the purpose of X-DAC 3.24 is twofold, first to allow X-DAC 3.0 to be upgraded to at least 24 bit 96k operation and second to act as the R&D test bed for XD5. I am pleased to report X-DAC 3.24 is done and has succeeded in both points. Below are pictures of X-DAC 3.24.

X-DAC 3.24 showing DIRMA, XD4396, and Proto-clock modules

X-DAC 3.24 with INA103 Diff-amps

To the right we see DIRMA the Digital Input Receiver Multi-function Adapter. The board looks a little empty as the optional SMB connectors for data I/O are not installed in this prototype. At the top of the board is CS8420, its support circuitry, and the SMB coax which keeps the master clock input to DIRMA nice and clean. DIRMA has been tested in both synchronous and asynchronous modes and works in both modes. Crystal has had some teething pains with the CS8420. I am sticking with it because in my experience on a bad day with a headwind a CS8420 equipped DAC sounds better than CS8412/14 and the upgrade capabilities are too important to walk away from. DIRMA is ready for the beta testers and I have their three boards under construction.

Next over in the picture is XD4396. This DAC board is equipped with Crystal CS4396/7 DAC chip feeding Burr-Brown INA103 differential input amps. Support circuits include JC811 regulator, BlackGate and WIMA caps. It is the fact XD4396 sounds so damn fine that gives me confidence to move this design into production. When Crystal announced their new CS43122 super-DAC chip uses the exact same pinout as CS4396/7 it was like an early Christmas present for Norman and an even greater confirmation this design is on the right track. I have also tried a completely passive output for XD4396 using Jensen transformers in place of INA103 amps.

X-DAC 3.24 with Jensen transformers

This sounds very pure and clean but the downside is it becomes very sensitive to loading from the pre-amp. Specifically if the load drops to 10k or so you can say goodbye to the bass! The news is not that bad as it leaves the door open for lots of good experimenting with op-amps, tube stages, and FET followers or whatever else your fertile minds conger. Now I know in some quarters the simpler-and-more-expensive-must-be-better Mafia will insist the $50 Jensen transformers MUST sound better than some $5 integrated instrumentation amp. Taint that simple McGee. I have come to regard the INA103 as one of the 'finds' of this round of ACG R&D. They sound oh so sweet, and are so quiet, and boast a decent +-40 mA output stage to boot! Of course once we get past the flavor of the month mentality we learn there are many great devices and ways to employ them to get high-end results. What it really becomes is a case of picking horses for courses. I believe a big part of why INA103 is shining so brightly in this application is the signal CS4396/7/122 is giving it to work with. Unlike many other 'single-bit' DAC chips we do not find tons of ultra-sonic noise accompanying the signal on CS4396/7/122. Like the classic CS4328 the raw output of CS4396/7/122 looks like a signal generator it is so clean and pure. Thus relieved of the need to cope with high frequency junk and noise INA103 can settle down to the job for which it was designed, processing pure clean audio.

XD4396 is released for beta-testing and production. I will be building the three beta-testers XD4396s to round out their X-DAC 3.24s. Concurrently with that effort I will continue pressing Crystal for samples of CS43122 for use ASAP. Now would be a good time to note the CS4397 and CS43122 both support SACD DSD data. Thus this solution leaves the door open for one day X-DAC 3.24 and XD5 playing SACD disks. NO PROMISES HERE STATED OR IMPLIED. Obviously until a standard to output DSD and/or DVD-A data to external DACs is established I can make no promises beyond a commitment to keep our options as open as possible. The real determining factor will be a political and business decision. If the format is a straight IEEE-1394 we are good to go. If they take the IEEE-1394 transport layer and wrap a name brand by license only shell over it (as Dolby Labs and DTS did to S/PDIF adapting it to surround sound) we are most likely screwed. Unless of course you help me sell enough XD5s to afford the $5,000-$15,000 license fee! :-) Here's hoping a more reasonable and small business friendly standard emerges. I have no problem with licensing to help pay back a company's R&D costs. I do have a problem with a license model featuring large up front costs as used by HDCD, dolby-digital, and DTS. These favor large established companies over the adventurous small entrepreneur. If a model including the license fee in the IC parts cost is established we get a level playing field and the diversity which it fosters. Guess we can not get through one post-CD digital audio discussion with a little politics so accept my apologies for slipping that in and we will get back to the technology update.

What about XD1853? I have yet to get satisfactory results with this promising part the AD1853. So small, so promising, such a complex analog stage required. I have decided to 'back-burner' this part for now. Part of that decision results from the fantastic results obtained using its competition the Crystal CS4396/4397/43122 and their DSD compatibility. In the following picture the XD1853 prototype is shown with many other test circuits.

A Rouges Gallery of Test Circuits

Consider most of these are hand built hard wired and one can see where some of the weeks went! There are actually two AD1853 circuits shown. The XD1853 PCB front center and it's predecessor a mostly hard wired circuit to the left, the one with the four coax cables running to a DIP adapter used to get data and clocks in. I spent some weeks listening to AD1853 on that circuit. It is a very nice sounding part, every bit the clean and supremely transparent reproduction one would expect from a top rank 24 bit 192k contender in the DAC race. What I learned from the prototype is AD1853 is very particular about its analog stage. Unlike the Crystal parts AD1853 is a current output DAC with no onboard analog filtering. It MUST have an essentially zero ohm termination impedance and extensive anti-alias (smoothing) filtering. The simple resistor to ground current to voltage converters I have used with previous current output DACs did not work in this application. It was only when I set aside the esoteric solutions and went to a three op-amp analog stage as Analog Devices application notes advice that a useable DAC was realized. This knowledge greatly influenced the design of the XD1853 circuit. It became quite a challenge to get the AD1853 IC, JC811 regulator, 6 AD825 op-amps and all associated support parts and circuits onto a board which can fit in the space available in X-DAC 3.24. To do this all passive parts and op-amps, with the exception of the BlackGate power supply caps, are Surface Mount Devices (SMD). The layout is VERY tight. I took this as an advantage allowing shortest possible signal paths. The downside became apparent when assembling the XD1853 prototype. The SMD Rs and Cs in the output filters are too close together for reliable hand assembly. The evidence of this is both the time required building the circuit board and the fact one channel is not reliable. It works for awhile then misbehaves with it's output 'sticking' at the positive supply rail. De-bugging this type of behavior in a 3 op-amp multi-function circuit occupying only 1.5 square inches of PCB is not trivial. I have 'fixed' it once only to have the problem return. The technician in me wants to get to the bottom of it. The ACG owner in me who has promised many X-DAC 3.24 and XD5 regrets we must set this challenge aside for now.

There is also the issue of costs and capabilities between XD4396 and XD1853. Using parts of equal quality the raw parts cost of XD4396 is only 68% of XD1853! Perhaps if time and R&D budget were not of such a concern the XD1853 design could be brought down in cost by using multi-amp op-amp devices in place of my preferred singles or other cost engineering measures. Designing down to cost has never been the way at ACG, rather I design up to a standard then see what it costs. Then add the fact XD4396 offers the possibility of DSD support and upgrade to CS43122 with proven reliability and the decision to move ahead post haste with XD4396 while setting XD1853 aside for now becomes a no-brainer.

As stated above the X-DAC 3.24 beta testers will be getting their upgrades as soon as I build and test them. Their three DIRMA boards are 75% complete and work will start on their XD4396 boards this week. I will also be surveying X-DAC 3.0 owners to determine how many will be upgrading to X-DAC 3.24 to begin production planning for this design.

The status of XD5 is the design is very nearly complete. Schematics run to eight pages thus far. The initial version will be similar to X-DAC 3.24 in that it will use the CS8420 and CS4396/4397/43122 chip set. Thanks to the modular nature of XD5 this can later be configured to work with other DACs that prove worthy such as the PCM1704. But again if this is to see the light of day I must guard against feature creep. It is a delicate balancing act keeping the design open enough to fulfill its mandate to be modular and updateable while limiting it to a point where we can get on with a physical realization in the real world.

While time and patience brought X-DAC 3.24 to fruition as a very sweet, dynamic, and transparent device the wacky world of Post-CD music reproduction marches on. When I began this project the question on everyone's lips was "who will win the format war, DVD-A or SACD?" As I see it this is becoming largely a non-issue. I make that statement because in 2001 neither DVD-A or SACD will emerge as a major mass market force in the way LP or CD did decades before. Twenty first century consumers have Cell Phones, PDAs, Play Station 2s, Internet appliances, PCs, Laptops, DVD, DVD-R, CD-ROM, CD-R, CD-RW, CD, dolby-Digital, DTS, digital radio, Direct TV, Dish Network, PDR, analog & digital camcorders, digital still cameras, MP3 (on the net and in portable devices) all competing for attention with DVD-A and SACD. DVD-A and SACD are not even on the typical consumer's radar screen! Which is really OK for enthusiasts like us. In this age of digital content delivery both consumers and providers are growing used to the reality of multiple delivery media and formats. While we may long for the 'good old days' of whatever music we want being available on LP or CD the upside here is we have a chance to sample two quite different proposals in DVD-A and SACD and vote with our pocket books for one or both. During this time of uncertainty some have retreated to old known formats as a safe refuge. Others wait for that mythical day when once again one single universal format reigns supreme. What X-DAC 3.24 as taught me, aside from the many technical lessons, is that while DVD-A and SACD establish footholds one can enjoy greatly enhanced CD playback while being equipped to taste the sweet joys of 24 bit playback. As I write this X-DAC 3.24 is playing the Classic Records 24/96 DAD of Holst's "The Planets" and I can assure you it beats the hell out of waiting for "the next big thing"!

DAD, DAD, Daddy-O & X-DAC 3.24proto !

With the initial 96k/24bit upgrade for X-DAC 3.24 complete I was of course anxious to hear for myself if the buzz surrounding 24 bit audio was true. A Pioneer DV-414 DVD player with 24/96 S/PDIF output enabled is used to play the following collection of DADs (Digital Audio Disk).

Out of all the DADs listed above only one disappointed (the Pictures/Planets CDAD 1027, sounds like a so-so master tape; why did this go on DAD?). My overall reaction is that I like high bit rate digital audio, I like it very much. However, it is not the universal panacea it is being touted as in some quarters. For example some would claim a 24/96 capable DVD player like my Pioneer is all one needs both for advanced CD as well as DVD, & DAD playback. I disagree. From its analog outs it sounds like what it is a mass market consumer electronics product built down to a price. Compared to X-DAC 3.0 its easy to hear X-DAC 3.0 sounds better playing 16/44 CDs than the DV-414 does playing 24/96 DADs! One cannot take several steps backwards in the basics of high-end electronics design and expect 24/96 to make up for a wealth of other sins! So the real test is how it sounds as a system with DV-414 feeding 24/96 data to X-DAC 3.24.

X-DAC 3.24 shows the improvements brought by DADs are all in the positive direction. This is not a 'knock-you-over-the-head-sonic-blockbuster-awesome-dude' effect. Rather a subtle yet quite meaningful refinement best appreciated by the connoisseur. Most of the time ones attention is on the music not the sound as it should be. Then along comes a little detail which is reproduced with a delicacy never heard before often leaving even this hardened audiophile slack jawed. Nuance of high frequency percussion, the 'texture' of instrument's individual sounds, bass instruments and room sounds are all just that much more natural and real. The all-important communication of human emotion many of us are drawn to music for is also enhanced. On the 'I'm Bad Like Jesse James' cut from John Lee Hooker's 'Boom Boom' the menace Hooker conveys comes through loud and clear. The jazz disks place the groups just right there in front of you. I wish I could say they sounded more like the 'absolute sound', but the sad reality is its been a damn long time since I for one heard live jazz presented with sound this fine. The Gershwin and Philadelphia Orchestra orchestral disks are both just really special treats. I love Gershwin so it is easy to enjoy his music on any medium. On DAD the lovely tone of the melody and inner complexities of the orchestration are presented in such a detailed yet completely natural fashion, what's not to like?! The Philadelphia Orchestra on Waterlily is in a word a triumph. For those of us who dream of actually reproducing the sound of a symphony orchestra in our home and frequent the concert hall to keep that sound fresh in mind (and one would hope to drink deep from art's heady elixir) a recording like this is an all too rare event. If within minutes of cueing up the first track you are not telling yourself "this is what this hi-fi thing is all about" then all I can say is "get thee to a concert hall!"

A couple other observations and notes on DAD:

• Three of the above disks have video as well as 24/96 audio (Mangione, Sara K., & Hart). I find the enhancements of 24/96 audio matter little if I play these with the video on. Maybe I am too visually oriented but the visuals seem to 'override' the aural. Instead of luxuriating in the sound I find myself critiquing the piece.
• The Mick Hart 'Planet Drum' disk is a 'DVD Single'. Even if you are not a fan of the 'world beat' style of music on this disk it is interesting in what it shows is possible in the new media. The disk contains video and three different audio tracks, the stereo 24/96, 5.1 dolby digital, and 5.1 DTS. Thus it is of interest for comparing the three major DVD audio delivery options.

ACG's R&D project Tube-DAC X-9

Somewhere at the bottom of Tube-DAC X-9 one can find parts of the original Tube-DAC upon which my Glass Audio article (issue 1/94) was based. The basic theme of a simple resistor to ground for I/V conversion and a tube based output stage remain as does the case and the pre-regulators; everything else is changed. The areas this proof of concept circuit seeks to investigate are:

1. Use of paralleled R²R current output DACs.
2. Transformer coupling between DAC and analog stage for isolation and filtering.
3. Tube stage based on 5842 last generation high-mu triode.
4. Tube based DACs for 96k 24 bit hi-fi digital to analog conversion.

Initial results have shown the first three areas of investigation are fertile indeed. Despite the far from optimal wiring and layout the X-9 DAC as pictured here sounds fantastic.

The two pictures above show the complete X-9 DAC. To the left are a pair of pre-regulators serving the DIR/filter and DAC boards. In the center are the DIR/filter (front) and DAC boards (rear). On the right is the 5842 tube based analog stage, its regulator, and input transformers.

X-9 side view. Three military connectors in left foreground carry DC inputs from external power supplies.

Digital Input Receiver (DIR) and 8X over sampling filter board. This is the 'old man' circuit board on the X-9 as I built it 4 or 5 years ago as evidenced by the patina its bare copper ground plane has acquired. It was my last major digital audio project on a hand made circuit card before going to CAD layout of etched PCBs. Based on a blank un-etched piece of FR4 circuit board drilled to accept sockets and solder turrets for the parts. Connections are made using a combination of the top ground plane, wire wrap, coax, shielded twisted pair, and heavy bus wire depending on each circuit node's function. Some of the twisted pair and coax is on top of the board and can be seen in the picture. The DIR installed is the UltraAnalog AES20. The design is such that when the AES20 is removed another socket is exposed into which a Crystal CS8412 can be fitted. Thanks to Crystal using the CS8412 pinouts on their CS8414 96/24 DIR upgrading the DIR to higher rates is a plug-in operation. The 8X digital filter is the NPC SM5842. This device is still state-of-the-art for 44.1/16 sources. An adapter for a 96/24 filter will have to be built once they become available.

The DAC boards. Among the sea of capacitors are two double sided DAC boards each holding 4 paralleled Burr-Brown PCM1702 20 bit DAC chips. The PCM1702s are powered by separate +-5v regulators for their analog and digital supply pins along with the gaggle of capacitors filtering these voltages. The white twisted pair wires carry the clocks and data from the DIR/filter board to the DACs as signal/ground pairs. The two purple and violet wires on the right are XLO cable which carries the analog DAC output to the next stage. Built into their connectors are sockets allowing different values and types of resistors to be tried for the current to voltage R.

The analog output stage built around one 5842 tube per channel. Partly visible on the left of the picture the regulator section of a Joe Curcio head amp provides B+ and heater power. For the circuit a perfboard was covered with copper tape to make a ground plane and most of the circuit is wired on the tube sockets. Mounted at 90 degrees the board carries the 5842s, their B+, and cathode bypass caps horizontally. Jumpers allow the 5842s to be biased using either conventional RC or Ni-cad battery bias. The dark cylinders in the center are Component Research Corporation output caps, these mil-spec parts a lucky surplus store find. The small perfboard and silver cylinders on the far right of the photo are the transformer isolation experiment. The DAC board outputs go to sockets on the perfboard and then onto the 5842 inputs. The socket allows quick comparisons between direct wire links vs. filters vs. transformers. The transformers shown in circuit are the damn fine Jensen JT-11P-1.

Tube-DAC X-9 future plans.

1. Parts are on hand to upgrade the 5842 RC bias components to audio grade parts. Then compare the RC to battery bias schemes.
2. Shunt regulator for the 5842 analog stage.
3. Upgrade DACs from 20 bit PCM1702 to 24 bit PCM1704.
4. Upgrade oversampling filter to 24 bit 96k capable part.
5. Rework packaging and wiring.

I would very much like your feedback on Tube-DAC X-9. As it now stands this is a personal project for my own technical curiosity and listening enjoyment. It could be developed into a kit or circuit board set for sale through ACG. The issue I see with that route is cost. While I have not costed out this design I know such a kit would easily run $2000-$3000. A printed circuit board set (DIR/FILT, DAC-L, DAC-R, Analog-L, Analog-R, 3x SSPS) would be $500 in the quality of boards I demand. Assuming Tube-DAC X-9 delivers state of the art performance would you be seriously interested? And by serious I mean pay cash dollars for one, not spend the next six months talking about it on the Internet! Would you prefer the luxury (and cost) of the current design with separate boards for each subsystem or combining subsystems into two or three boards to save cost at the expense of flexibility? E-mail me your comments to ntracy@galstar.com.

DVD and the Future of Digital Hi-Fi

"I don't think they play at all fairly," Alice began, in rather a complaining tone, "and they all quarrel so dreadfully one can't hear oneself speak-and they don't seem to have any rules in particular; at least, if there are, nobody attends to them-and you've no idea how confusing it is all the things being alive; for instance, there's the arch I've got to go through next walking about at the other end of the ground-and I should have croqueted the Queen's hedgehog just now, only it ran away when it saw mine coming!" from Lewis Carroll's ' "Alice's Adventures in Wonderland"

Like Alice attempting to play croquet with a live flamingo mallet and live hedgehog ball all the while the Queen of Hearts screaming "Off with their heads!" I do not think they play at all fairly. They in this case are the rogues' gallery of consumer electronics companies, record companies, computer companies, and movie studios who are driving DVD technologies. Conspicuous by its absence is any concern for the consumer's rights or ease of use (hint to any corporate flacks reading this the consumer = your customer not your enemy). Like the Queen of Hearts' croquet game the rules and even the goal posts are constantly moving. I have some of this documented on 'The DVD Audio Soap Opera' page. I moved this off out R&D page because I have grown so very weary of it and to separate the technology from the politics.

Digital to Analog Converter Chips for 96k/24bit conversion

24-Bit DACs to Improve DVD Sound article from Nikkei Electronics Asia web publication.

16 to 24-Bit DACs comparison table compiled by Norman Tracy (send updates, suggestions, & comments).

Of course as anyone who designs for high end audio will testify the numbers do not tell the whole story. Nor can they be ignored rather they form part of the 'big picture' one must consider. Surveying the current state of the art in audio frequency DACs summarized in the above article and table my conclusions include:

• Doubling the conversion rate and reaching for 24 bit linearity can actually yield lower measured performance. This is expected as faster clocks and data lines make noise breakthrough and control a larger and more contentious issue.
• Several of the parts being offered are obviously aimed at mid-fi markets not even reaching the performance levels of good 44/16 DACs. Nevertheless one is already seeing components being marketed to the high end hi-fi market containing '96/24' DACs which even the manufacturers own specifications reveal as more like 16 bit devices.
• The best of the new 96/24 breed are exciting indeed. In the PCM1704 Burr-Brown sets new standards for R²R devices. The other parts are all variations on the PWM theme. Exciting among these are devices like Analog Devices' AD185x line and Burr-Brown PCM1728 which attack the jitter problem right at the DAC architecture level. This is important because simple scaling of 44/16 techniques results in a jitter requirement for the master clock beyond the current state of the art if not physical possibilities!
• A new challenge for DIY builders and small manufacturers is that with one exception (Crystal's CS4390) all new DAC chips are only available in small surface mount packages. The upside of this is miniaturized SMD packages offer better signal fidelity thanks to lower LC parasitics. The downside is added difficulties in building and testing with the tiny packages.

Super Symmetric Power Supply

To deliver on the promise of 96/24 formats a super clean ground and power subsystem is required. The Super Symmetric Power Supply (SSPS) sets out to provide the necessary foundation. A second generation version of the existing X-Pwr used in X-DAC 3.0 Signature SSPS is being developed with the invaluable input of John Camille. The following graphic is the prototype. The production SSPS details are now online here.

A drawing showing the use of Super Symmetric Power Supply to power the Philips CDPro transport is here (4,7M).

The drawing is rather large thus it is here in PKZIP format (2.5M).

Digital Audio Input and Future Compatibility

The disjointed and rambling introduction above is disjointed and rambling because standards for the post CD era of hi-fi digital audio are disjointed and rambling! In fact rather than universal standards we see the technology being shaped by the politics of large multinational corporations jockeying for position. Currently the options include:

• AES3/SPDIF as used in CD. A few makers of DVD players (Pioneer, Theta, CAL) are allowing 96/24 data streams out of their machines on SPDIF ports. Others (Sony, Toshiba) limit SPDIF output to 48/16 even while playing 96/24 DVD disks bowing to record company pressures.
• I²S and I2S· enhanced are promoted in some quarters as the high end low jitter solution for hi-fi digital audio links. Its features of breaking out the clock line from the data in a parallel multiple conductor cable and allowing the master clock to reside in the DAC are quite desirable. The need to isolate more signals and potential sources of RFI between components is a potential technical pitfall. More troublesome from a standards point of view is that the US high end audio companies promoting I²S lack support from major consumer electronics companies and worse yet seem to have allowed the I²S· enhanced "standard" to have fractured into two slightly differing specifications.
• IEEE-1394 Firewire is the vaporware solution. Favored by major consumer electronics and record companies for a promise of greater copyright security and computer to entertainment electronics convergence. Despite utopian hopes of simple serial interconnection of everything immediate advantages for the audiophile are hard to see. In the very complex interactive audio/video system the two way asynchronous and isosynchronous link is desirable for the ability to implement 'plug and play' setup, operation, and deliver both audio and video rates. In the context of the simple user active audio only systems still favored by audiophiles IEEE-1394 brings a layer of complexity to the component design with possible sonic consequences. Basically another microcontroller in the DAC churning away pumping packets back and forth means one more noise source to isolate and control. As of this writing it is a moot point as I know of no hi-fi devices yet using IEEE-1394.

Thus the challenge becomes moving forward while attempting to keep options open. A few SPDIF/AES3 Digital Input Receiver chips designed to run at the 96kHz rate have appeared. These are an option and we are seeing product using them appear. While pondering how to best add 96/24 input capability with the flexibility needed the answer appeared in the form of Crystal Semiconductor's CS8420. The CS8420 combines input and outputs on both AES3/SPDIF and parallel formats with a selectable 96k capable sample rate converter. These elements are combined as shown in the following functional block diagram.

CS8420 details

Note especially the switches connecting the various CS8420 internal components. These allow great flexibility and let this one part serve many functions. In the world of hi-fi digital to analog converters one can envision using the part in the following ways:

1. A 96kHz 24 bit capable SPDIF DIR (Digital Input Receiver) to serial output. Equivalent function of the industry standard CS8412 used for 44/16 formats. In this configuration the AES3/SPDIF decoder is enabled, the sample rate converter (SRC) is bypassed, and output sent to the three wire Serial Audio Output port. The exciting things about this mode are the 96/24 capability and improved jitter performance on the recovered clock. Quoting from the CS8420 data sheet: "the PLL has been designed to only use the preambles of the AES3 stream to provide lock update information to the PLL. This results in the PLL being immune to data dependent jitter effects, since the AES3 preambles do not vary with the data." The PLL in CS8420 begins filtering jitter at 1kHz in contrast to the earlier CS8411/12 which only filtered above 20kHz. Finally the PLL external filter RC components can be varied to further tweak the PLL performance optimizing the PLL lock vs. data source accuracy tradeoff.
2. An I²S input for the DAC. Using the Serial Audio Input port instead of or in addition to the AES3/SPDIF DIR I²S can be accommodated. With additional interface components the I²S could come from CD/DVD transports so equipped. Should a completely new (to hi-fi audio) interface become a standard, such as IEEE-1394, the CS8420 Serial Audio Input port makes a perfect input gateway for a IEEE-1394 or other interface upgrade.
3. A 96kHz 24 bit capable SPDIF DIR plus Sample Rate Converter to serial output. This enables Asynchronous Jitter Reduction and 44/16 to 96/24 upsampling in our hi-fi DAC. With X-DAC 3.0 we enjoyed great success using Asynchronous Jitter Reduction to enhance CD playback. While the use of SRC is controversial in some quarters I view it as just another DSP operation. That is correctly executed and applied SRC can help, improperly done it will hurt. Given the plethora of sample rates and bit depths allowed on DVD the SRC-phobic had best get over it! In addition to the known advantage of allowing the use of a low jitter local clock this configuration will allow the intriguing possibility of experimenting with converting CD format 44/16 data to 96/24 before conversion. While information theory tells us such an operation will not create any new musical information experience in the real world shows upsampling has often yielded audible benefits. A configuration using the Big Buck$ dCS SRC to upsample caused a buzz among attendees at the last Stereophile show. With CS8420 we can try it for considerably less.

Norbert Bayer's website www.diyaudio.de has great content with a special emphasis on hi-fi audio DACs. He shares the results of over 6 years work on the subject and is currently planning his fourth DAC. Below are pictures of two of his previous efforts. As a big fan of the Crystal CS4328 I especially like the one featuring it with a tubed output stage including Tango output transformers.