Playing DSD .dsf files from the beginning on the Rossini skips the first 0.84 second of audio

Support
63

Here is a bit of context to understand my growing level of frustration.

I started my journey with dCS with a Puccini SACD/CD player and UClock that stopped working when I changed my Apple devices. Turns out the UClock was not compatible with newer Apple operating systems. So, I packed it up and sent it to them for an upgrade, $1,000 worth of upgrade. COVID hit, the factory shut down and I was dead in the water.

dCS America graciously intervened, and got me a very fair price on a DeBussy. I was so taken by the improvement in sound quality I bought a Rosinni DAC, and planned to use the UClock with it (I’d had it returned unmodified). Since the UClock’s issue was with compatibility with Apple OS, and the clock function worked perfectly, I was gonna have a pretty sweet system.

Well, the Rosinni was delivered a couple days before my Wilson Sasha DAWs were to be delivered and installed. I never opened the Rossini box. The Wilson dealer brought with him a Vivaldi DAC, clock and Upsampler, a couple days later which I bought.

I’ve been on the audio merry go round for around the last 30 years, grabbing for the brass ring every time I passed it. I’ve spent way more than I can remember on audio components. About 2 years ago, as I was about to retire I decided to get off the merry go round, bite the bullet, and just buy the best of everything I wanted/needed to enjoy my system and focus on other things.

One of those things is playing guitar. One of my goals was to build a system in front of which I can sit, and play along as if the artists are in the room with me. The system I have allows me to do that, other than the front end of songs being truncated. If you were at a concert, and the mixing board had this kinda glitch you’d probably not be too happy and might even leave.

A final analogy. Suppose you bought your dream car, a Mercedes SL63. It was the best car you ever owned, other than in second gear between 2,000 and 2,500 rpm the engine stalled. And then suppose this was a systemic problem and others were having the same issue. Do ya think MB would just shrug it off, or tell their customers a third party might be able to tinker with the software and make it better?

Here is the original thread on my Puccini problems.

just FYI, I just finished listening to Roxette Wish I Could Fly from my NAS in normal sampling and then my Mosaic playlist had Bush Come Down in Hi-rez via Quboz. Bingo, you guessed it, I missed the first three beats in the song.

Mitch

64

Manel, if you didn’t realise, Mosaic Control actually has no part in the sound quality at all. Once you select a track or album to play using Mosaic Control, the dCS network board will go and directly fetch the tracks from your MinimSever (or Qobuz, or wherever).

So, when tracks are playing, Mosaic Control only receives periodic updates from the network board on the progress of the track playback, and other system status information (such as volume level etc.). It has nothing to do with the sound quality.

1 Like
65

I understand. It['s] definitely is an issue that needs resolution. :+1:t2:

66

Roon definitely sounds better than the DAC alone when it skips that first 0.84 seconds.

It’s clearly something dCS didn’t take into account, and its frustrating as every DAC using off-the-shelf chips I’ve been able to test gets it right.

I’ve opined before that despite the promise above, I suspect dCS found they couldn’t fix it in software; those samples have to be stashed somewhere for the length of the time it takes the software to figure out if the data is PCM or DSD, and if it’s DSD it will have to play the track from the beginning of the buffer as DSD and play out of that buffer from that point on until the DOP bitstream ends, remaining forever 0.84s behind.

I suspect other architectures just always buffer their bitstream data for just that reason.

While dCS does have a buffer feature that can be enabled to avoid noise while the DAC determines whether PCM data is Dolby Digital or PCM, I suspect it’s not large enough to hold the necessary DSD data.

I further suspect the bigger issue is a fix was promised, but cannot be delivered due to hardware limitations.

Of course they could have fixed it on the Apex boards by adding additional RAM, but we won’t know until they write new software to take advantage of the additional memory which obviously the current version of the Rossini software cannot do, and if they release new Rossini software to take advantage of extra RAM on the Apex board it breaks their promise to fix the bug in the next Rossini software release.

I don’t know if any of this is true, but based upon my decades of experience in software development and the complete radio silence on dCS’ part, it seems the most logical conclusion.

67

Sorry you’ve fallen down the rabbit hole with me. You at least were able to get acknowledgement from dCS Ltd that it is a problem that needs to be fixed. I’ve yet to get any response which really bugs me. I’ve spent countless hours pouring the manuals, crawling behind my stereo rack to triple check all the connections and asking for help.

The dealer from which I bought the gear is stumped. Prior dCS America employees were stumped. All the experienced users on this forum who’ve graciously taken the time to try to figure this out are stumped.

dCS Ltd is silent in the issue. Leads me to believe it is an unfixable problem and a pretty glaring shortcoming for equipment of such cost and otherwise exceptional quality.

I’m sure you and I cannot be the sole dCS owners who switch between normal sampling and Hirez or DSD. Maybe everyone else just accepted it, reworked all their playlists to contain a single sampling rate and settled for that. In my sunset years my mantra has become “life’s too short to settle…”. Big decision point staring me straight in the face now…

Cheers,

Mitch

68

It becomes a question of, if Apex still does it, why reward dCS with another $9000 US?

The answer is, of course, if the sound improvement is worth the money, then perhaps it is.

Anything that can insert a delay when sending a DOP stream can work around this; JRiver is one app that can, and of course Roon, so there are “hacks” if you will.

So can I live with it? I can, but it’s just so disappointing on so many levels, kind of like buying an expensive car that won’t start when driven somewhere for a short trip; it’s easy to say “don’t do that, short trips are bad for a car anyway” but it’s still disturbing.

Don’t get me wrong, bugs happen.

It’s just the no update on it since August 2020 that had me thinking “they can’t fix it” as I outlined in my previous post.

69

I never said Mosaic is responsible of sound quality, minimserver is the source and Mosaic a necessary Control point.

Roon has its sound signature, that you may like or not.

Theories aside, minimserver-Mosaic sound different that Roon-Mosaic.

Play a string quartet and you can hear the difference.

70

They all are bitperfect, and both come from the same NAS and network, wich is been worked to optimize results.
Out or topic, but I would like to share with you guys, that the Innuos PhoenixNet switch next to the Vivaldi turned out to be an audible improvement.

71

This is my reference when I talk about sound quality :wink: :

2011-07-10 00.01.02
2011-07-10 00.01.021920×1434 251 KB

1 Like
72

Imagine sitting in the audience, and the singularly distinctive 4 beat opening of Beethoven’ s Fifth were truncated to just 2 beats. You’d get up, leave, and expect your money back.

73

Very good.

I actually played with this a bit and instead of hearing the famous

da da da dummmmm…

You actually hear

… a dummmmm…

Of course when you complain, you will be assured that if you just listened via the symphony’s subscription-only app you would hear the full performance in a way that is sonically identical to listening live.

Of course the symphony also has a new upgrade that costs about 1/3 of what you paid for your season tickets that will make the parts of the symphony you actually can hear sound better, but will still need to subscribe to the extra cost app to hear it all.

74

Heavy sarcasm follows:

We could go to every recording artist in the world and ask them to insert 5 seconds of dead space at the front of every song so our dCS equipment can keep up.

End of sarcasm.

I really enjoy having friends over and showing off my system. The song I like to play for them to really experience the system’s clarity, and power is Pneuma by Tool, available on Quboz only as a hi-rez recording. So, before they come over I have to have the system running, and before they arrive, make sure the last song played is also in hi-rez, otherwise I have to explain why the first second of Pneuma is missing.

“Mighty fine system ya got there Mitch, too bad we missed the beginning of the song…”

75

It doesn’t actually. It’s just an asynchronous error-corrected packet transport software. Once the UPnP or RAAT packets are “unwrapped” and re-assembled into PCM or DSD, the dCS Streaming board puts them out as identical I²S bitstreams to the rest of the system.

3 Likes
76

+google. . . …

77

I heard from dCS through my dealer; they are aware of the problem and will fix it. No timeline provided so a bit more patience is required.

In the meantime I will rework my playlists to contain a single sampling rate so no switching is required. unfortunately, my favorite song at the moment, Pneuma by Tool, is available on Quboz only as a hi-rez recording, so I’m gonna have to find the shortest hi-rez recording in the world to precede it in that playlist. Or, I could buy a Vivaldi transport and a CD of the album…

As an aside, I had a guitar lesson today (more like a jam session along with the stereo) and neither my instructor or I could get the correct rhythm going on KWS Lay it on Down since we lost the first 3 beats of the song. Sorta like having a drummer in a band who is always a couple beats behind…you’d go find a new drummer…

Here is hoping I don’t have to harp on this again.

Mitch

78

Mitch, maybe I’m missing something here :thinking:

This thread is specifically about the problem associated with missing initial bits when playing a DSD track immediately following a PCM one (ie. format change specifically from PCM → DSD). There was never an issue with just PCM tracks playing one after another regardless of the bit rates, or even after a DSD track.

But you seem to suggest having problems playing PCM tracks off Qobuz. That’s news to me. Can you describe the problem exactly? (Maybe this deserves its own thread actually).

79

Anupc,

Poorly worded summary on my point:

If I play a “normal” sample track (either from my NAS or Quboz), followed by a DSD or hi-rez track, the first second of the second track is truncated. BillK took the time to figure out it is actually the first .84 seconds.

The converse is true as well; if I play a DSD or hi-rez track, followed by a normally sampled track, the first .84 seconds of the second song is truncated.

This is a portion of the response I got from DCS Ltd:

“…this is a systemic problem. It is not a setup issue that is causing the beginning of DSD tracks not to be played. As you’ll see from the Forum thread, it does also impact Rossini, but the extra steps in the chain with Vivaldi (the Upsampler’s output rate for example) has added more complexity in getting the issue nailed down”.

I got another response from DCS America that also stated the DAC goes into mute mode while the Upsampler is switching sampling frequencies and as a result the beginning of a track that requires a new sampling rate is truncated. Once the Upsampler adjusts to a new frequency and sends the signal to the DAC, the DAC comes out of mute, and the first part of the track is lost. At least that is how I understand it, but I might be in error.

If I build a play list that is exclusively one sampling rate (all normal, or all hi rez or all DSD) there is no issue, unless the track played immediately prior is a different sampling frequency, in which case the first song of the playlist at the new frequency is truncated. All else that follows is fine.

I don’t know if you are able or willing, but build a 2 track playlist in Mosaic (bypass Roon). First song normal sampling and second DSD or hi-rez. You’ll hear the switching occurring in between tracks and lose the first part of the second track.

Is this the end of the world? Obviously not. I’m reworking all my playlists to be exclusively one sampling rate and that seems to avoid the issue. There are many songs however that are only available as hi-rez on Quboz, so I have to compromise and have those only on a “hi-rez playlist”.

BillK has a lot more experience with data transfer than I so he can chime in and correct anything above I’ve mis-stated.

Cheers,

Mitch

80

Anpuc,

I did start a new thread on my specific issue about a year ago:

I hate to beat a dead horse, but here’s a hypothetical:

Suppose you had around $80k to spend on the best digital streaming front end you could buy. A dealer sets up two rigs: a Vivaldi stack, and a Boulder 2120 with an Aurender W20SE. You have two songs you want to hear on both to assess which one is most resolving and faithful to the original recording. One of them is a normally sampled track, and one is available on Quboz only as hi-rez. You listen to the Boulder/Aurender combo and it is awesome. Next up ya listen to the Vivaldi set up. It sounds appreciably better, but the second song is truncated. The dealer explains this is a systemic fault due to switching of sampling frequencies (the Boulder/Aurender does not have this issue). He’s not sure when/if dCS is ever going to fix this. Which system do you buy?

dCS made a commitment to me to get this fixed, and I’m known for my patience so I’ll wait. But then again, I woulda been a doctor but I did not have the patients.

Cheers,

Mitch

81

That pretty much covers it.

There are two issues here as mentioned before:

  1. The DAC takes time to determine the data it is receiving is DoP DSD and switch to DSD mode from PCM

  2. The dCS units discard all data received when switching into DSD mode, which causes the loss of the first 0.84 second of the DSD track.

What makes this complicated is the unit makes different assumptions about whether the datastream will remain in the same format under different circumstances.

I mentioned earlier in this thread, when playing DSD data:

Via USB:

Played continuously starting at Track 1:
Track 1: First part cut off
Track 2: Plays normally

Jump to Track 2:
First part cut off

Jump back to the start of Track 2:
Plays normally

Via Network:

Played continuously starting at Track 1:
Track 1: First part cut off
Track 2: First part cut off

Jump to Track 2:
First part cut off

Jump back to the start of Track 2:
First part cut off

This is because:

  1. When playing via USB, the DAC remains in DSD mode when playing to the next track, so there’s no PCM/DSD switch. When jumping to the next track it does reset to PCM mode, causing the drop, but when jumping back to the beginning of the same track, it stays in DSD mode.

  2. Via network the dCS hardware closes the connection at the end of the track and resets to expecting PCM when the datastream stops as it awaits a new connection for the next track - whether it’s due to a track ending due to completion of the track or jumping to the next track or to the beginning of the current track.

There is something else occurring as well, as the DoP spec says a DAC should take no more than about 180 Âľsec to determine a data stream is DoP rather than PCM, so it apparently also takes the dCS hardware around .75 second or so to switch to DSD mode and unmute.

Now the conditions above only occur when DSD data is being pushed to the DAC; if the tracks are instead pulled from the DAC using Mosaic, the issue does not occur because Mosaic knows it is requesting a DSD format track and changes DAC modes ahead of time.

It’s not just a bug fix; it’s a design choice that needs to be made as well, as if the DAC buffers the data then properly starts DSD playback from the buffer after changing modes it will be forever .87 seconds behind, though that’s not really an issue as it’s not much different than PCM playback when the PCM buffer is enabled. It can’t cause lipsync issues as the buffer option does as no video sources emit DoP except perhaps when using a device like the GeerFab D. BOB to provide DoP from an SACD/BD combi player to a dCS DAC.

I have to admit that I haven’t tried to see whether there is a DSD to PCM delay, but it may be different on the Rossini due to the lack of a separate upsampler as is present in the Vivaldi stack or Vivaldi One.

Were I dCS I would probably make the buffer option control both PCM and DoP buffering, if they wanted an option to shut it off.

As before I still wonder if they can fix it, as the buffer memory used for PCM may be wired directly to the PCM data bus and the DSD decoding engine may not be able to “see” it. Or, the decision to just drop DoP data on the floor when switching the DAC to DSD mode may have just been a poor technical decision or one made because they needed to get the code released before they could solve the issue.

Regardless, as it’s been eighteen months since a fix was promised to me, as stated I have to wonder if one is actually coming.

I do not expect one to come with the APEX mod, as it was promised that the fix would come in the next software release, and the Rossini APEX is labeled “2.0,” which is the same software revision as the current Rossini, so I expect the new hardware to be installed but the software will remain the same, pending the “future” update.

82

Bill,

Thanks, I learn something new from every one of your posts…I’m an engineer, but a civil one (so I’ve been told). You wanna know anything about moving dirt - I’m your man…

Cheers,

Mitch