PCIe to USB bridge update:
While the guys are working on casting the XO's in their Panzerholz anti vibration mounts, machining the copper brackets for mounting the PCB's into the Extreme and soldering the wire harnesses, I have been getting some hands on quality time with this final production level revision of the board with some interesting and also a few, admittedly, unexpected results.
First of all our effort to produce a card that does not change the general voicing / sound signature of the Extreme have endured. It really "sounds" the same in general terms. What it does do is remove a significant amount of hash, grunge and even distortion of the stereotype "you had no idea it was there until it's gone". The result of this is a much more relaxing sound with a better musical "flow", a larger dynamic range and a significant increase in separation, especially noticeable when using TAS is for example the differentiation of a piano piece where you can now clearly separate the hammer striking the string, the string and tone board resonances and all of its subtle decays fading out over time, all combined into a singular event with correct placement into the soundstage, each individual part with the right relative volume, making it just much more believable and natural. The whole listening experiences becomes both more relaxing and exciting at the same time. The impact is fundamental, straight to the heart of musical enjoyment.
An unexpected side effect I did not anticipate is it dramatically changes the equation on the networking side of things. Where we managed to achieve a significantly reduced audibility of network components used in TAS, the USB card changes the equation. My favourite (currently available) switch for example is the Telegartner Gold M12 with its supplied CAT7 network cables, however with the USB card in the Extreme, it deteriorates performance compared to a standard Netgear GS108 switch with a standard CAT5/6 unshielded network cable, muting top end extension, decreasing overall clarity and detail levels and reducing dynamic impact and sound staging abilities. This is a complete reversal of what used to be a reasonable predictable scenario. Even more surprisingly it applies to BOTH TAS and Roon to a similar degree. We will have to verify these results with some other "audiophile" switches and network cables, our resident Ether Regen has unfortunately passed on to the digital after life, cause unknown, it did run very hot. We will verify the impact on Fibre networking next either running direct from our router or by using FMC's. We may have entered a stage where the de-tweaking philosophy of forum member ddk (to achieve a more natural sound) may have become a reality for digital playback, at least on the networking side.
A partial technical explanation for this may be found in the noise performance of the USB card. I am hesitant in sharing as many design details as I used to as we have grown from a garage/home based 3 man part-time team to a full time 7 man team operating from a 20.000 sq ft facility with about a dozen subcontractors. There are over 2000 engineering hours that went into the design, excluding listening evaluations which I did not keep track of, all components have been selected by a combination of simulated and measured performance, suitability for circuit design and listening tests. So we do have some intellectual property to protect at this level of involvement and scale of operation. With all of that said, noise performance clearly matters, over a very wide bandwidth, even from a very low 0.1Hz up to 100MHz and possibly even higher. Every single component has a noise signature, often subject to external influences, and even if this is exceedingly low, it can have a surprising impact on perceived playback sound quality. Voltage regulator performance is of paramount importance. Nowadays you have "ultra low noise" IC regulators like the popular LT3045, although these are pretty good on their own, they do not reach the performance of a well designed discrete regulator. We do employ a pair of LT3045s, but they are only used in a less critical location as a pre-regulator, followed up by lower noise regulator designs. On the subject of clocking, you could view an oscillator as a noise regulator, a necessary evil in digital playback. I think there is no disagreement on those having an impact on sound quality. The large Can OCXO's which have become popular in digital audio are available in many shapes, forms and different performance levels for different applications. An OCXO is made up from a crystal, cut to resonate at a specific frequency. The stability is subject to variations in applied voltage, surrounding circuitry and external influences like temperature and vibration. The crystal in an OCXO is cut to perform optimal in a specific temperature range, a controlled heating element (oven) is added to ensure the crystal remains within that temperature range to reach maximum stability, typically between 70 and 85 degrees Celsius. Commercially available packages include the crystal, oven and surrounding circuitry into a single "tin can" which you can often drop into an existing design replacing less stable oscillators. The surrounding circuitry can have quite a few components, capacitors, resistors, chokes, logic IC's, regulation, filter circuits etc. This will all be operating at relative high temperatures, of course the parts used are rated to operate in that range so there is a reasonable life expectancy, though performance does change over time. We have some experience with this using OCXO's in our first models, and performance does change noticeably over a 3-5 year timespan, even diverging off spec to the point where the system starts producing errors. We have had to replace the OCXO's in our 2015 model on several occasions already which is simply unacceptable at this level of expenditure. The XO design we use has been redesigned from the ground up to operate optimally in a ~30-45 degree Celsius temperature range which is the typical internal operating temperature of the Extreme (~10 degrees above ambient). The design goal was to have a competitive phase noise performance to even quite costly OCXO's and being able to evaluate all components used in the XO circuitry for their influence on Sound Quality. The crystal is cut to work optimal in that temperature range, it turned out quite well, the discard rate is a very low 5.3%. The circuitry components tolerance is very tight and the series production performance variation is negligible. I will not disclose specs but it exceeds the performance of a reference OCXO which on its own exceeds the entire USB card retail price, without the use of an oven!
It just turned out even better than I had hoped for. I would carefully like to call it a very significant advance for computer based playback / streaming but the proof will have to come from you, our valued customers, listening evaluations.
The retail price has turned out higher then what we originally aimed for at EURO 1600 excluding sales tax where applicable, but I do feel it's performance more than warrants its price tag.