Sound quality differences with optical outputs - why?

For years I haven´t been bothered with digital S/PDIF connections. I basically stopped using them in 2001 when I got rid of my MD deck. This deck had been connected using its RCA digital input, only because everyone was saying that these gave superior sounding results. I never actually checked out if this was true with the result that I ignored optical Toslink in-/outputs back then. But my recent encounters with the Sony MZ-R 30 and the Sony MZ-R 900 forced me to use them again - they are the only way with which I can achieve digital copies with these MD recorders. Thankfully many of my vintage portable CD players are equipped with an optical TOSLINK interface, sometimes I also bought them for this reason. With an optical cable I can connect my PCDPs to an MD recorder to create a digital copy even including trackmarks automatically set by the MiniDisc recorder.

Optical mini-Toslink on the Sony D-335

During the last few weeks I recorded roughly twelve MiniDiscs. Some were recorded with the respective CDs played back by my Sony D-465 or the Sony D-335, some were recorded with the Creative Soundblaster X-Fi HD USB as their the digital source. To my surprise the CDs copied with those two CD players sounded different to what I´m used to, it was apparent that the sound signature had been changed somehow. The ones made with my Creative sounded pristine instead. To use an analogy: recordings from my CD players sounded like a decent analogue copy while the ones from the Creative sounded like the originals. But how can this be true? In both cases I used an optical cable that transferred its data digitally, effectively creating a second original (ignoring the lossy ATRAC compression). All the digital outputs I´ve used so far are bit-perfect meaning they won´t change the digital signal in any way. I´ve confirmed this with one of my CDs from TELARC, a rare example of a DTS-recording they released some 15 years ago. Our Sony STR-DB 830 QS is able to recognize the digital signal as DTS-encoded, it doesn´t matter which device is doing the playback. So why do I hear differences between digital signals? It just doesn´t make sense.

My optical fiber cable (copyright: Philips)

The cable you can see above is the one I´ve used for all my tests involving MD recorders and my Creative. Build quality is good... the gold plated "connectors" don´t make sense of course; the signal is pure light and cannot be influenced by the conductivity of certain metals. People seem to expect it though... what the heck, it looks nice doesn´t it? Back to the question of why I´m hearing differences. The S/PDIF protocol necessary for digital Toslink transfer has several disadvantages. The most significant is that the receiver of the signal doesn´t control it, the sender does. Because of that it must synchronize its internal clock (necessary for D/A converting) to the one contained inside the signal itself which is generated by the device transmitting it. Now in case of jitter (you knew that I would mention this, didn´t you?) introduced by an unstable or polluted clock the receiver (D/A converter or MiniDisc) won´t have a stable clock reference, in effect making jitter audible upon conversion to analogue signals. Research into this subject has been going on for years, for example from members of the AES. This influence can be reduced of course by reclocking the incoming signal. Both my MiniDisc recorders do exactly that: they reclock the signal with their built-in samplerate converter, effectively creating their own clock reference. Another impermeable wall for jitter is the MiniDisc writing process itself because it is another reclocking step. All of this should in theory eliminate jitter completely. Would it be able to survive it would result in distortions not unlike those found in intermodulation distortion and in a compromised stability and clarity of the stereo image. How are my MD recorders doing this reclocking? They probably use a phase-locked loop circuit (PLL) which you can imagine as a huge flywheel averaging out speed variations. Depending on the quality and implementation of the PLL jitter reduction is more or less successful, according to an article from Soundonsound it cannot ever be eliminated completely though. Attention: even with jitter the transferred signal will always be the same, it still contains the same 0 and 1, albeit at different places (sort of).

Jitter test signal on MZ-R 900, played back by Creative Soundblaster X-Fi HD USB

Jitter test signal on MZ-R 900, played back by Sony D-465

Look at the two pictures above by clicking on one of them, then scroll through them with your mousewheel. They were created employing the Sony MZ-R 900, its analogue output recorded with my E-MU 0202 USB. I set the MD recorder into record-pause mode in order for it to serve as a D/A converter only without the result being distorted by the ATRAC encoding. In both cases I used exactly the same signal, it´s the jitter test I usually use for my vintage portable CD players. In case of the Creative it was played back by foobar2000, utilizing the WASAPI interface of the soundcard - without resampling of course. Exactly the same file (this time burned on a CD-R) was then played back by the D-465. Level matching was unnecessary since both results have the same gain. On the pictures you can see that the spread of the signal is wider with the Creative. With the signal played back by the D-465 however there are additional sidebands absent with the Creative. The example below also exhibits a noisefloor higher for roughly 1-2 dB. While I assume these tiny differences to be insignificant, they (or something else) do have an audible effect on music. Recorded from the D-465 music has an unsharp and compact image, sounding a bit bloated and less crisp. Piano for instance sounds a bit distorted while strings loose their gloss. All of this was restored to normal when the Creative was used as a digital source for recording. I even did a DBT, not only for you but also for myself to make perfectly sure that I wasn´t hallucinating these differences:

foo_abx 1.3.4 report
foobar2000 v1.1.13
2012/06/25 20:54:42
File A: D:\Coast Guard Rescue Creative.wav
File B: D:\Coast Guard Rescue D-465.wav
20:54:42 : Test started.
20:56:15 : 01/01 50.0%
20:57:00 : 02/02 25.0%
20:57:29 : 03/03 12.5%
20:57:48 : 04/04 6.3%
20:58:08 : 05/05 3.1%
20:58:26 : 06/06 1.6%
20:59:45 : 06/07 6.3%
21:00:00 : 07/08 3.5%
21:00:06 : 08/09 2.0%
21:00:41 : 09/10 1.1%
21:01:01 : 10/11 0.6%
21:01:08 : Test finished.
----------
Total: 10/11 (0.6%)

For this DBT I recorded the same track ("Coast Guard Rescue", The Perfect Storm) twice, again with the D-465 and the Creative. Level matching was unnecessary as both had the same level, they needed precise alignment of course in order for the DBT to be successful. By this example you can see that different digital sources might have an impact on sound, even when they are bit-perfect. I´ve subsequently tested several other players and their digital outputs: all my portables appear to be susceptibe to something that distorts the digital sound while still being able to transfer a bit-perfect digital signal. There was only one portable player coming closer to the clarity of the Creative transferred signal, the DE-J 725. It didn´t sound bloated but flat and uninvolving (strangeley mirroring its analogue signature). The player almost sounding like an exact clone of the Creative was my Pioneer DV-610 AV which is nothing more than a simple DVD player. Are the reasons for this jitter? Or is it something else? I´d like to know if you, dear reader, have made similar experiences. Maybe my portables do have an unstable clock source or their signal gets distorted by other factors unknown to me yet. In either case the Creative once again proves itself to be a very reliable device of high quality. Additionally a clean digital signal from the source seems to be vital for perfect sound. To all the doubters out there: a bit-perfect realtime signal doesn´t mean that it´s free from some kind of distortion! You might now think that everything is a bliss - you are mistaken there. Recording to MD with the Creative is quite tiresome, trackmarks are not transmitted and therefore I have to set them myself after recording which essentially means that I have to divide a single, 60-minutes track into several pieces. But I´m obviously not a person that loves convenience that much (Exhibit: using MD recorders) so my desire for quality usually gets the upper hand and in the end I´m rewarded with perfect sound quality from MiniDisc. Before I forget: my reviews for the MZ-R 30 and the MZ-R 900 have been done with recordings from the Creative.

Last update: 10.01.2013