I disbelieve in audible differences among well designed (non tweak) digital players and converters. All the various tweaks either do nothing or make them worse, I think.
However, that being said, I've long felt I heard differences, and in at least one case I want to know more.
Back in 2008 when I first started playing DVD-Audio discs on my first high end living room system since 1992, I knew the digital from DVD-Audio discs was limited by design to 16bit and 48kHz at the digital audio outputs of all players.
But I figured I could capture the full 24 bit resolution at the analog outputs by sampling it to 24 bit digital with a pretty nice A to D converter (the Lavry AD10, which I bought in 2009, replacing the theoretically similar but poorly regarded A to D converter inside my Tact RCS 2.0 preamp). Since I had started doing my crossovers, EQ, and everything in digital, I needed to sample all analog sources including vinyl records and FM radio to digital, so why not High Resolution DVD-Audio discs as well?
You could make various arguments against this. You might perhaps say that the noise level is increased by this resampling. It might even be increased to worse than 16-bit levels, around 98dB. But in fact my players, my line level preamp, and my converter are all rated at 117dB signal to noise ratio. So if each stage reduced that by 3dB, you'd still have 108dB, which is more than the maximum from 16 bits.
And what fascinated me since first thinking about it, is that in some fundamental sense the resolution is being preserved, even as a little noise is added.
Nowadays a lot of Audio/Videophiles would find this hard to believe. Virtually any form of copying (if it's even possible) makes a noticeable difference in video. Even just de-compressing and re-compression into the usual MP4 or other compressed video formats. But video is far different from audio. For one thing...the lossy encoded formats inherently lose real resolution on each re-enconding. You can easily understand this when you see that many different inputs could cause the same output--which is the very essence of lossy compression. And there are other reasons relating to the nature of scan lines and things like that.
Analog interphase copying or transmission of audio can be damned good. And it's weirdly reversed from video compression. Instead of many inputs potentially causing one output...now we have one input potentially causing infinite outputs (just being offset by different tiny amounts of time, noise, or other analog "errors").
So, I started doing this just to get the full resolution of DVD-Audio, or at least as much as my 24/96 Lavry AD10 can capture. (I continue to believe that 24/96 is as good as any human needs, and that there's no need for more "time coherent" digital systems like DSD which does get reasonably good when you get to high enough sampling rates like 128x, which is nice, but it's just a waste.)
But then I noticed that even CD's sounded better this way. In fact, it was beginning to seem like all CD's sounded better being played by the Denon 5900 and resampled into 24/96 by the Lavry sounded better than straight digital from the Denon 5900 or any other player.
And resampled in this way, you can also hear the differences among CD players, etc. (If there are any.) And so then I started collecting a bunch of different disc players that seemed like they were optimal for different kinds of media. Like the Pioneer PD-75 for CD's, the Sony DVDP-9000ES for SACD's (said to have essentially the same true 1-bit converters as the SCD-1), and the Denon DVD-9000 for HDCD's.
Now, I got called out on this practice by someone who pointed out that I could indeed capture DVD-Audio digital from the HDMI with a de-embedder. And he was right, I could do that with nearly all DVD-Audio discs.
I did do some tests and decided then that indeed the direct digital was better than re-sampled from the analog outputs.
And while I still kept my battery of different players initially, over time they got displaced by other bits of equipment I wanted to consolidate in my audio rack. Only the Oppo BDP-205 and Denon DVD-9000 remain. Also the laser in the Sony DVDP-9000ES died and needs to be replaced again (something that curiously happens to that player about every 3 months in heavy use, or longer if lower use, I think Sony modified the SACD copy protection requirements after that and the earlier SACD players had much bigger laser assemblies that weren't bothered as much by the heat or something).
So now I was planning to sell the PD-75 but decided not to, somehow the digital played through it (and other players???) seems to have more passion, character, or something, than even the straight digital being streamed by the Oppo into coax.
And all along, I've noticed that nothing at all sounds like the Santana Supernatural DVD-Audio played on the Denon 5900 into analog and resampled into 24/96. The straight digital over HDMI lacks bass, sounds dry, and has far less passion.
I noticed that the Denon 3910 seems to sound much like the Denon 5900 and possible...even better than the Oppo BDP-205.
All this is complicated by level differences. It remains possible (and in fact objectivist audiophiles would predict) that once I assiduously match the levels, everything that has at least 16 bit resolution should sound identical.
I usually haven't done deliberate level matching (only new listening) at all. When I've done it, it's been pretty ad hoc. If I want to systematically compare a bunch of players and transmission methods, I ought to make this systemaic.
And so I begin by laying out the parameters for this. One can NEVER resample the analog back into digital and retain exactly the same peak level. One must ALWAYS allow a little extra headroom in the AD converter. The way I usually do this is by discovering the level (in 0.5dB increments) on the Emotiva preamp that lites the second to the last red light on the Lavry, but not the last light, so that clipping has never occurred. This is obviously disk dependant, so there's a certain amount of guess and try that must go on.
What I ought to do is play Supernatural on all the players that can play it, and find the maximum level they can play it without clipping.
Then after that, there has to be an additional adjustment to raise the level further to match the direct digital. If a digital recording is being made (as I intend to do for resamplings of Supernatural on every machine which will play it) the final adjustment can be done in Audacity. To to make the comparisons fair, every recording should be made at the just-below-clipping level. (I've never bothered with that nicety before.)
In my previous experience, I've heard maddening differences between the Denon 5900 and the Integra RDV-1. I figured the RDV-1 would be better but in fact only the 5900 had the special magic that made Santana sound wonderful.
Now you might say, if I'm going to be comparing the players after having been recorded to 24/96 digital and then being streamed all over...if streaming is "losing something" then it's still going to be "losing something" if I'm using it to compare renditions that were not streamed.
You might think that, but I've felt I've heard the differences between the players best in this exact scenario, where they've been pre-recorded and re-streamed on demand. It's great for comparing phono systems too.
I don't actually think the problem is that streaming is "losing something." I think certain players especially, like the Denon 5900, and perhaps other Denon players, may be "adding something." What they might be adding which is actually beneficial is something like upsampling. The player is converting to the "infinite resolution" of analog as best it can...and some players do this more nicely than others. The Denon players have AL24 which I've always suspected as being superior to what is used by others. Or at least it sounds sweeter, smother, more spacious, and slightly more bassy than other players in a particular way that just happens to enhance Santana Supernatural, and perhaps other things.
It may be nicer to have all digital upsampled to 24/96 in ways other than is done electronically by the ASRC's of the MiniDSP OpenDRC's that I use now. (That's where this job gets done now in my system.)
"Nicer" might even be not as electrically and mathematically accurate, as well. It could involve some kind of bandpass filtering, for example, that just happens to remove out of band garbage.
Supernatural played on the Oppo BDP-205 with all balanced connections requires a +1dB gain in the Emotiva for -1dB peak indication on the Lavry. A limited amount of listening while I was making the recording suggested that the HDMI digital and the resampled analog sound identical on the Oppo. The digital, limited to 16 bit and 48 kHz, sounds notably inferior. On the Denon 3910, the HDMI sounds identical to the Oppo HDMI. But the resampled analog sounds different, much as I described before, slightly more spacious, less grainy, and comfortingly powerful bass. It requires the same +1dB gain using an unbalanced connection to the Emotiva. All consistent with the idea that the magic is in the Denon AL24 digital filter, even though in this case it does no upconversion per se since the source is already 24 bit and 88 kHz. The blue AL24 light is lighting up letting you know that Denon is "improving" the sound somehow.
A confounding factor in a lot of ad hoc background listening impressions is the troublesome way there is slight bass cancellation at the listening position, and bass augmentation most everywhere else. Chairside bass boost control might be nice, perhaps automatically enabled and disabled by sitting in the listening chair.
I keep trying to think how one might fix this completely with FIR. I was thinking of modeling the bass source at the listening position. But this is not invertible, it seems to me. What might be is this: modeling a bass reduction in the far corner(s) of the room.
That might be changed by a frequency dependent phase shift similar to parametric EQ, with a notch phase shift at the mode frequency that excites the far corner, such that it cancels there but not at the listening position. Then, the parametric EQ that notches the mode out completely could be removed, restoring the full bass impact at the listening position.
****
Update: I've made full transcriptions of Supernatural on both Oppo BDP-95 and Denon 3910. I'm not at all sure I even hear a difference, or that it consistently favors the Denon.
Notably, when I crank up the level to +6dB on my kitchen preamp, playing in Audacity on my Mac, I can hear hum and a tad of noise in between tracks on the Denon 3910 transcription. At normal levels I don't clearly hear it at all, and I doubt it makes much difference. My SOP estimate is about -80dB from peak level. The Denon is sitting on my coffee table and I needed to run a 6 foot audio cable to reach it. I pulled out a genuine late series Radio Shack interconnect for this purpose, and it crosses power cables and such at close to 90 degree angles as the best I could do. The Denon is plugged into a different outlet strip as the Oppo (but the same ultimate power conditioner, the Panamax 1500). With more effort I could probably improve the hum issue here in obvious ways (but at considerable effort, like using the rack slot normally occupied by the Denon 9000 which is hard to move). I will try to better for my 5900 transcription. I could NOT make a Denon 9000 transcription because the 9000 I currently have in the rack is the one that only plays CD's and HDCD's, but which seemed to have the most magic sound of all. I'd need to replace it with the second Denon 9000 I have, and I'm not even sure if that will still play DVD-Audio's either.
Meanwhile the Oppo has a balanced connection to the Emotiva, which has a balanced connection to my AD converter, all with short cables. Cranking up to maximum on the Oppo transcription I hear nothing but silence.
So I know in at least one technical way, the Oppo transcription is better. But sometimes the Denon transcription sounds smoother and invites being cranked up in a way that the Oppo transcription does not.
It might be that "crank it up" impulse, when fulfilled, leads to the "magic" recollected moments.
When I think I hear this sort of difference in the Denon transcription, it seems mostly like a "simplification." With the Oppo, I can clearly hear how overdubbing is used extensively in this album. The more "detailed" rendition it provides exposes all the trickery. Meanwhile the Denon seems to bring things together, better expressing the musical intent without peeking behind the curtain. In particular, the Denon emphasizes the musical swells more than technical minutae.
But they are so close I'm very doubtful I could reliably hear a difference in double blind tests, without considerable "training," and perhaps even then. And also quite often the Oppo rendition is nothing short of spectacular, and I have little doubt it's "technically" better.
(A friend always looks to polarity differences as the culprit when things sound "too detailed." It was easy to check this in Audacity. It's visible from the initial waveforms the transcriptions have identical polarity. I never knew I could do such comparisons so easily. But I have no idea if both are "correct" or both are "incorrect," only that it probably doesn't matter much.)
Which is one thing I hate about my modified (for total transparency) QSC ABX box that I currently use only for amplifier comparison testing. It gives no feedback until you are finished all the trial rounds. I think everyone needs a training setup, where you get immediate feedback, and perhaps no score is even tabulated. I need to get up to speed on more recent ABX programs and apps.
For the purposes of ABX tests with the Dynaco 410 last week, I added a line level switch which could shut off the signal to the Dynaco. That enabled me to check my guesses right away without messing with the QSC tabulation. I identified the amplifier (Hafler 9300 vs Dynaco) correctly 4 out of 9 times.
I'm glad to have made these transcriptions for future listening and analysis. But I'm doubtful it's worth keeping the Denon 3910 simply to "improve" the sound of discs, if it even does that. I want to get comparable transcriptions from Denon 5900 and Denon 9000 as well, among others. I have long planned to keep the 5900 because of how magic it sometimes seemed in the past, but it hasn't been hooked up to my living room system in over 5 years now.
I ultimately decided to "amplify" every transcription almost to maximum on Audacity, leaving a mere 0.01dB of unused headroom to be sure no clipping is occurring. And that way they can be compared directly without further level matching. And likely they can be compared with a digital files version as well, if I ever get one (as I have been thinking to do).
It might also be interesting to do CD and SACD transcriptions from various players as well. But I'm going to wait until particular discs excite me enough to try.
To be clear, I think both the Oppo and Denon transcriptions are magic sounding on the Supernatural album, but in slightly different ways that could be only sighted listening bias.
It's possible the RDV-1 transcription I once made which sounded relatively dull had some issue that didn't really represent the best of the RDV-1. In particular it's very easy to mistakenly play the "Surround" tracks which the disc defaults to (regardless of player settings) instead of the dedicated Stereo tracks which sound far better.* I want to check that out too. The RDV-1 should have among the best sounding outputs, and I was using it for several years as my "DAC."
(*I put the disc in, and it soon starts playing the multichannel version, which plays very softly, about 12dB down, because on the stereo outputs I'm getting a mix-down. I have to select top menu on the remote to select the stereo tracks. Then, maddeningly, if the disc automatically repeats, as it tends to do on many players, the repeat again defaults to multichannel version. This sort of thing helped kill DVD-Audio. From the beginning SACD played according to pre-selected mode, but not DVD-Audio.)
*****
Now I discover to my dismay (and forgetfulness) that Roon is also adjusting playback levels. As I had only adjusted the levels of the middle section of the album to match (at 0.01dB peak), and since I have album-wise leveling selected in Roon, that might explain why the 205 was getting 5.1dB of reduction, whereas the Denon only 4.9dB. So I began tossing out the other album sections into my home folder and asking Roon to re-scan. That, most strangely of all, made matters most worst, with 4.3dB reduction for the Denon and 4.9dB now for the Oppo. So I put all the other tunes back into the Denon folder, hoping to bring it back to 4.9dB as well, and now it stubbornly insists on attenuating by 4.8dB, so a mismatch I've managed to reduce from 0.2dB to 0.1dB by radical measures. As the difference gets smaller, they seem more alike.
****
I continued to play around with moving the first and last sections out of the 205 and 3910 transcriptions of Supernatural until finally I got Roon to apply the same level matching to both album transcriptions. It seemed weirdly history dependent, but finally I got the job done by attenuating the first track section to -0.6dB peak for the 3910 version, and leaving this section entirely out of the 205 version. If it were up to me, I'd leave this track out of both transcriptions and that was my plan for awhile I don't like to play it at all. But the first track seems to drive the level matching more than any of the other tracks, and somehow Roon estimates the 205 sourced tracks as louder by about 0.5dB, even when I've adjusted their peaks to the exact same value. So fiddling with the first track let me get the level matching correct for the other tracks I really care about playing, and those tracks have peaks set to -0.01 dB for consistency.
This alone suggests there is something weirdly different between the 205 and the 3910 sourced versions, even as they are now sounding virtually identical after level matching and tricking Roon to play album sections 2 and 3 at the exact same attenuations for testing by leaving a specially doctored track one in only one of them.
(Some might wonder why I still put up with Roon with all it's weirdness. But I like Roon for a large number of reasons, particularly my ability to create playlists for it. And for automatically generated playlists of full albums, there is nothing better than the album-wise level matching option Roon provides. This makes it possible to enjoy one album after another. When there is no level matching one quickly becomes very fed up with automatic play. So level matching is something I need, I only perhaps wish I could switch it off on an individual album basis, when some albums are actually just "tests" of different transcriptions like I've been doing here. For test purposes in other words. But meanwhile I can see why Roon makes it hard to change this. It's generally the secret sauce that makes everything sound good. And audiophiles like me are prone to leave things in test conditions by accident and then foul everything up for awhile. Such as leaving speakers out-of-phase when one was testing that. I've done this sort of thing numerous times, often taking amazingly long to figure out anything was wrong. That's in large part to my mostly-background music listening. I only rarely sit in the "hot" seat, sometimes not for weeks when I've been especially busy. I consider it a feature, not a bug, that I can pretty well enjoy my living room stereo anywhere in the house. And often it's even somewhat cool the bass is exaggerated in most such locations. I only I wish I could revert some of that excess to the listening position.)
No comments:
Post a Comment