has anyone tried PS Audio perfect wave duo

I don't get it.... This unit has a one disc memory and plays the disc from memory, not directly from the disc.
In effect, this is a one disc music server.
Am I missing something or is this just the ultimate in audiophile marketing hype?

I don't think you are missing anything. Any lo-fi cd player, and I believe any $20 computer cd-rom drive, not to mention any audiophile-caliber digital transport, has a solid state buffer memory (aka cache) through which the data coming off the disk is passed, before being clocked out at a fixed rate which is not subject to the (very large) variations in the rate at which data is retrieved from the disk.

The only relevant difference that I can glean from PS Audio's literature is that their memory is 64mB, which is much larger than typical. But I fail to see how the larger memory size would convey any benefit, assuming that the disk-reading mechanisms are performing in a half-way reasonable manner. The 64mB, by the way, is nowhere near large enough to cache an entire disk (which may be up to 700mB or so). PS Audio's writeup describes it as being able to store "up to 3 minutes of music," which is in the right ballpark.

I have no doubt that it is a very well engineered and good performing player, but its descriptive literature is misleading and disappointing at best.

Regards,
-- Al

Hi Mapman -- My main point was that it is misleading to claim that other transports output data directly from the optical read mechanism, and that other transports are therefore subject to the timing variations which inevitably occur as data is read from the disk which is spinning at a varying, motor-controlled rate.

But re the larger cache, all that is necessary is to have a cache which is large enough to not become emptied under conditions of worst case variation of the rate at which data is read from the disk. A few seconds worth of cache should be way more than enough to accomplish that for any audio format, for any reasonably well performing read mechanism.

Best regards,
-- Al

Hi Mapman,

Yes, if the cache were not large enough to support the timing variations of the read mechanism, clear audible dropouts would occur.

If the optically-read data were conveyed to a dac, or to the internal dac of a cdp (transport plus dac in one unit) without any memory cache in between, I would expect the result to be a total mess combining dropouts with various distortions associated with the timing variations.

Best regards,
-- Al

Thanks for the good explanations, Shazam. You should replace whoever writes their literature!

I do agree that the file-based read, and the I2S interface (which avoids multiplexing the data and clocks together) sound like excellent approaches. Hopefully, as you say, the I2S approach, or something similar, will be adopted by others.

What I was taking exception to as misleading, though, is exemplified by this paragraph at their site, in the description of the transport:

In a standard CD player or Transport, the master clock is synchronized to the optical disc reading mechanism. This means you are basically relying on a mechanical spinning mechanism and all of its correction systems to give you a perfectly stable, fixed clock to feed the DAC. It does not work and it is not stable.

Here's the problem. Optical disc readers are constantly changing the rate at which the data is coming from the disc. Sometimes it comes faster and sometimes it comes slower than the fixed speed of an asynchronous clock. If that data is coming in faster than the clock, you get a traffic pileup and the system crashes. Too slow and nothing comes out.

The Digital Lens has a large and smart memory storage buffer. It's big enough to handle any speed variation of the optical disc reader.

That would appear to indicate that in other transports and players the timing of data to the dac is subject to fluctuation identical to the fluctuating timing of the data coming off of the disk. It was sufficiently misleading, in fact, to have apparently misled one of our most intelligent and experienced members, with whom I was having the discussion above. :)

Best regards,
-- Al

Shazam -- Yes, it makes (bit-)perfect sense! Thanks very much.

So the benefit (or at least a benefit) of the large cache is that it allows time for the large number of retries which may be attempted, at least with marginal disks, which in turn would essentially eliminate the need for error interpolation.

One question that brings to mind that I would want to assess is how well-controlled the acoustic noise produced by the drive mechanism is, considering that it is running at considerably faster than 1x rates, and it will occasionally shuffle around and backtrack for the re-reads. I'd assume that is addressed well in the design, but it seems like a relevant question to raise.

Another point relating to their descriptive literature, though, would be that claims that no error correction is used are probably mis-stated. EAC's site, in fact, states that with the EAC software "if there are any errors that can’t be corrected, it will tell you on which time position the (possible) distortion occurred, so you could easily control it with e.g. the media player." In other words, I would envision that the PerfectWave Transport does away with the combined error correction/error interpolation provisions that are provided by more conventional cdp drive units, but then makes use of the Reed-Solomon error correcting codes in its own processing, to correct all errors which can be bit-perfectly corrected. There would seem to be no reason not to do that. The difference relative to a conventional cdp is, if I am correct, that no error interpolation (i.e., estimating of what the sample value should be) is performed -- that is what is eliminated by the multiple re-reads.

FYI, re your statement about the PW using EAC, one of the pages at their site indicates that their MREC (Multiple Read Error Correction) process is "similar in concept to EAC."

The benefit of all of that, of course, will vary with the physical quality of the cd, and perhaps the age and condition of the laser. But there is no question in my mind that the I2S interface approach is vastly preferable to the conventional interfaces that multiplex and then de-multiplex clocks and data, and that is perhaps the most significant advance provided in this design.

Thanks again for the good explanations.

Regards,
-- Al

I suspect that it is not flash memory which is being used, but rather dynamic RAM (random access memory), similar to computer RAM memory except with much less capacity. Hard drives and flash memory retain what has been stored when power is turned off, while dynamic RAM does not. However, RAM is much faster, and also cheaper and more compact. Also, flash memory has a limited number of write cycles it can handle before failing, the number being fairly large but potentially a significant factor in cdp or transport lifespan. And it requires sophisticated memory management algorithms to minimize that limitation.

In principle a transport could certainly be designed to utilize either a hard drive or a flash memory, but it would add to the cost and complexity, without providing added value for many or most users.

Regards,
-- Al

Tbg -- From the owner's manual:

The memory for the PWT is not located in the rear SD card as is commonly assumed. The rear panel SD card holds the cover art and song title information as it is acquired. In addition, the SD card is used to program new firmware in the PWT.

The SD card is flash memory, of course, but as this indicates it is not what is used as the buffer memory for the music.

Regards,
-- Al