Experimenting with digital transports...

[feckn_eejit]

I have a handful of redbook CD digital transports. For the purpose of this post, I'm going to concentrate on three: the Sonic Frontiers SFT-1, Micromega Drive 3, and Parasound C/BD-2000. It's interesting to note that both the SFT-1 and Drive 3 use the same Philips CDM 12.4 drive mechanism, while the Parasound of course is a belt-driven CEC jobbie.

I have always felt they all had their own sound, however I was extremely happy with the Micromega Drive 3.

Either way, I bought a Genesis Digital Lens for yuks to see if it would make them all sound the same, because I really like using the Parasound thanks to its unique user experience of sliding open the door and placing the lead puck on the spindle... kinda vinyl like in a way. Further, I hoped it would make my Airport Express a great sounding digital source (which it does).

Prior to the insertion of the Digital Lens, the SFT-1 and Drive 3 sounded demonstrably different. The SFT-1 simply sounded much more "digital" than the Drive 3 - much colder and harsher in the highs, and light in the loafers in the bass. The Digital Lens closed the gap to the point where I couldn't tell the difference between the two.

Confusingly, the Parasound still sounded demonstrably different. After this experiment I'm fairly certain my unit is actually broken, but we'll get to that... the Parasound always had woolly bass, rolled off highs, and a recessed, almost distorted midrange but an incredible imaging and great sense of "air" around the instruments... the bass and treble extension certainly improved through the Digital Lens (though did not approach the performance of the CDM 12.4-based transports...), however it retained its strange-sounding midrange.

To try and figure out what (if anything) was really going on, I decided to hook these three transports up to the S/PDIF input of my RME HDSPe AIO sound card, and record the output of a CD track. I'd then do the same again, but with the Digital Lens in the chain. I did all of this twice to see if I'd get identical results in each configuration over two passes. Finally, I'd rip a copy of the track using cdparanoia. As a bonus, I used iTunes to stream an ALAC-encoded rip of the same track to a latest-generation Airport Express and recorded the results of its mini-toslink output both straight into the RME and through the Digital Lens.

As the RME doesn't have BNC S/PDIF jacks (.......yet.....), I used a 1.5M Kimber D-60 RCA-->RCA cable to feed the RME from either the source transport or Digital Lens. I used an extremely shitty 0.5M RCA-->RCA coaxial cable between the transports and Digital Lens. For the Airport Express, I used a mini-toslink-->toslink adapter and a $1/foot 6' toslink cable. The Digital Lens was configured not to do any dithering. Monitoring was done with the balanced XLR outputs of the RME feeding a Pass Labs X250.5, feeding Stax SR-X Mk.III/Pro cans through an Illusion ESC-1001 transformer box.

The test track was Steely Dan's "Negative Girl" from their Grammy-award (best album!)-winning 2000 effort, _Two Against Nature_ (hard to say whether I hate this or 2003's _Everything Must Go_ more, musically speaking...), picked for a couple of reasons. Firstly, it's a completely digital recording so I wouldn't have to contend with tape hiss taking me off digital zero when trimming the offset. Secondly, it's an audio benchmark track for me - particularly Tom Barney's bass track, played on a 5-string MusicMan StingRay through a gorgeous Eden rig. This is a reference bass guitar recording as far as I am concerned - the nuances of his fantastic performance are captured in a level of detail I've rarely heard from other highly-produced multitrack recordings. Finally, while I do take pretty good care of my CDs and this one shows it, it is not without imperfections as it has logged considerable car time. I wanted to use a less-than-perfect CD in attempt to coax more read errors out of the transports to magnify any potential differences.

After some trimming to get the offset right on all the tracks (simply got rid of all the lead in and out digital zero in WaveLab...by hand...), it was time to run some comparisons!

I used both WaveLab's compare function and EAC's just to confirm.

The results when compared to the ripped track, as reported by the number of "differences" in WaveLab's compare function:

Straight into the RME:

SFT-1 - 91 differences (averaged between two takes)

Drive 3 - 75 differences (averaged between two takes)

C/BD-2000 - 4625 differences (averaged between two takes)

Airport Express - identical

With the Digital Lens in the chain:

SFT-1 - identical (both takes)

Drive 3 - identical (both takes)

C/BD-2000 - 5040 differences (averaged between two takes)

Airport Express - identical

I haven't ABX'd it, but casually, the tracks recorded from the Parasound transport sound demonstrably different on playback, while everything else sounds the same.

Though it could easily be down to prejudice due to previous experience, the various configurations exhibited similar sonic flavours through the monitoring system during recording compared to what I'm used to feeding my DACs.

I've come away with the following conclusions:

1) The difference in sound between the two CDM 12.4-based transports is down to their differing jitter signatures -- a difference which is eliminated by the Digital Lens

2) The Parasound transport is either a) seriously broken or considerably shitty

3) The Airport Express sounds shitty straight into a DAC because it's a jitter monster, which is solved by the Digital Lens

I remain confused as to the reasoning behind why the CDM 12.4-based transports produced errors going straight into the RME, but were bit-perfect going through the Digital Lens.

Interestingly, while to me the Drive 3 sounds the best without the Digital Lens in the chain, the Digital Lens reports it as having the least accurate clock speed - 162ppm slow (enough that the second PLL in my Assemblage D2D-1 won't lock onto it, though no other receiver seems to have a problem...), while everything else I have is within 40ppm of "correct". I assume this means that while it's slow, it's mighty stable, while the SFT-1's clock is not as rock solid though on average closer to correct. *shrug*

I have a Micromega Drive 1 and Drive 2 as well, which again use the same CDM 12.4 drive. I haven't tried but I bet they would produce the same results as the Drive 3 through the Digital Lens.

[Grahame]

Have you tried using Audio DiffMaker

It's use lead to some interesting conclusions here

Attention Audio Mythbusters! - Squeezebox : Community : Forums

[feckn_eejit]

Haven't tried that but seems it does the same thing as WaveLab's compare function - you end up with a file which is the difference between the two files being compared.

In my experiment the only ones that stood out were the files that came out of the Parasound - the difference file is basically a bunch of ticks and pops and doesn't sound like music in any way shape or form......

[Pars]

The CEC drive is belt-drive, correct? I can see belt-drive on a turntable but I thought CD drives change speed depending upon where on the disc it is reading? This is something that I would not think belt-drive would be good for. There are probably a lot of other differences in the SPDIF interfaces, etc. as well.

[Grahame]

"

Audio editor packages such as Wavelab already offer 'audio file comparer' functions that simply subtract one file from another. This is handy for checking that two files are identical (to see if S/PDIF I/O is bit-transparent, for instance, or that the bypass button on a plug-in really does what it claims), but Audio Diffmaker's Extract function is far more sophisticated. When evaluating analogue gear it can also compensate for tiny audio-interface sample-rate drift between the 'before' and 'after' recordings, or gain changes due to different cables or gear warm-up, so you only hear more subtle modifications. Still, for best results you should warm up analogue gear for several hours before starting test runs."

from Happy Birthday Vista: How Was It For You?

[feckn_eejit]

Sounds like an interesting piece of software. Not sure that it's any more useful than what I've done with WaveLab in my particular case, as I'm dealing with things entirely in the digital domain here where they are either perfect or not... it does look like it'd be interesting for comparing the analogue output of a DAC with the various transports hooked up to see what the difference is. An experiment for another day...

The CEC drive is belt-drive, correct? I can see belt-drive on a turntable but I thought CD drives change speed depending upon where on the disc it is reading? This is something that I would not think belt-drive would be good for. There are probably a lot of other differences in the SPDIF interfaces, etc. as well.

Yes it's belt drive. It certainly takes a long time to spin up/down when changing tracks. The CEC bumf was always that the belt drive and 1lb puck on top of the CD smooths out unevenness in the motor and provides a more steady stream of bits to the laser ;P

[grawk]

I think most modern cd transports read at a constant motor speed, and just buffer the data.

[feckn_eejit]

Define "modern"... all of my CD players/transports from the 90's-early 00's vary the motor speed depending where you are at on the disc...

[luvdunhill]

I thought CD drives change speed depending upon where on the disc it is reading?

There are two main approaches here. Maintain a constant linear velocity, in which case the disc spins slower as you move outside on the disc, thus keeping the overall transfer rate constant. The second is a constant angular velocity in which the disc spins the same speed, but the transfer rate changes depending where on the disc you are located. In fact, I'd argue that the later are better for audio, as since the speed doesn't change, the mechanisms can be made to be significantly quieter. Furthermore, since data transfer speed isn't all that important in audio it's a viable solution. The later is indeed what the CEC belt driver players use and the former is what most modern CD/DVD drives use.

[Dusty Chalk]

I think most modern cd transports read at a constant motor speed, and just buffer the data.

I'm pretty sure very few of them buffer the data. Think about lipsync issues with video -- I realize that's DVD not CD, but still...the only ones that buffer are the portable ones with anti-skip protection. I realize it's trivial to implement, but hardly any manufacturers -- that I know of -- do.

[grawk]

I'm not sure how you'd do async conversions without a buffer.

[Dusty Chalk]

Like what, like what dCS do? They're one of the exceptions. Unless you're talking about something else.

[feckn_eejit]

There are two main approaches here. Maintain a constant linear velocity, in which case the disc spins slower as you move outside on the disc, thus keeping the overall transfer rate constant. The second is a constant angular velocity in which the disc spins the same speed, but the transfer rate changes depending where on the disc you are located. In fact, I'd argue that the later are better for audio, as since the speed doesn't change, the mechanisms can be made to be significantly quieter. Furthermore, since data transfer speed isn't all that important in audio it's a viable solution. The later is indeed what the CEC belt driver players use and the former is what most modern CD/DVD drives use.

The CEC-based mechanism in the Parasound C/BD-2000 is most certainly CLV...