Vivaldi Upsampler toslink input hiccups

I’ve been noticing some peculiar toslink behavior with the upsampler. I’m a bit neurotic about lip syncing issues, so I’ve been trying all the settings and upsampling bit rates, and the delay seems slightly slightly different among them, and perhaps AES3 vs AES3+4 required output. The bigger issue though is when using 48->DSD or DSDx2 I occasionally get a “matrix” like sound as heard in the first few seconds of this video:
When doing DXD it occasionally gets muffled for a couple seconds like going from stereo to mono or almost hearing something underwater. Apparently my LG tv is known for occasional drop outs but I’ve also noticed the behavior straight from my xbox one x. I’ve tried different sources (Netflix / Prime video / YouTube) from the Xbox Apple TV and LG tv itself, two different optical cables, two different pairs of AES cables, just one clock cable with the upsampler set to W1 and dac M. I have not noticed it on Ethernet input for music which is the only other input I’ve used so far. From another thread it seems the upsampler optical in has high standards for signal purity, but if I leave it at 48khz output I don’t hear the “mono” like sound, at least not yet. I don’t know if dCS has any debugging instructions I could do.

On a tangent, I’m curious how many milliseconds the delay is on various upsampling settings, and if the dac master clock / upsampler wordclock1 setting adds a syncing delay.

Just adding a data point. Played Qobuz through MacBook Pro to a usb wyred4sound optical out. Exact same behavior. 48khz sources (Radiohead A Moon Shaped Pool) have upsampling oddities. I haven’t yet done long tests of other sampling rates->dsd or dxd over optical but it seems the upsampler does not like 48 over optical.

One more data point. Having both the Vivaldi dac and upsampler on “auto word clock” leaves both clocks on “?” but no matrix noise. With the proper universal master dac / word clock 1 upsampler I get occasional matrix noise on 48khz-> dsd2. Frustrating, to say the least.

Chuck, I think your clocking configuration’s isn’t quite what it needs to be :wink:

When using the TOSLink input on the Upsampler - in fact, for any Source which can’t be sync’ed to the Vivaldi’s system clocks - the clocking configuration should be;


  • Input: TOSLink
  • Sync Mode: “Audio” (not “Auto”). i.e. the Upsampler derives its clock from the TOSLink input
  • “Word Clock Out” port must be connected to DAC’s “Word Clock In3” port


  • Sync Mode: W3. i.e. uses the Clock supplied by the Upsampler (which is derived from the TOSLink input).

That way, there’s only a single clock in the system; derived from the TOSLink input.

Thanks Anu, I’m all ears. My config (unless experimenting) is from the manual:

But I see now that’s stated for network and usb inputs. So I should have another clock cable using out from the upsampler + toslink to in for the dac while usb and network upsampling needs universal clock out from the dac to in for the upsampler? Forgive my confusion, it does seem convoluted :slight_smile:

Yup, thats correct! :smiley:

Network & USB are asynchronous inputs, so the dCS system supplies the master clock. Whereas a TOSLink input is synchronous, and the dCS system needs to derive it’s clock from that input in order to avoid a clocking conflict.

I understand your confusion, it is a little complicated, but not unnecessarily so; the good news is that the dCS Vivaldi system is by far the most comprehensive when it comes to supporting various clocking schemes, no other vendor’s system comes close :wink:

With respect to documentation on TOSLink into the Upsampler Clocking configuration, there’s only one line in the Vivaldi DAC Manual that talks about it, on page 27;

Word Clock In3 is primarily intended to be connected to the Upsampler’s Word Clock Output to improve performance with sources that do not have a Word Clock input.

Additionally, if I recall correctly, James might have mentioned this very point somewhere among the various posts on as well.

Always fun to learn something new :slight_smile: I’ll swap my lone clock cable around for testing the toslink input before decoupling my tv for awhile. I’m trying out a dedicated tv room and music room, but want to ensure the upsampler can handle the optical in without issue. Thanks so much for the insight :+1:

1 Like

@Anupc hopefully last questions. If I eventually step up to the Vivaldi clock, from the clock I need two output clock cables to the upsampler and two to the dac. For synchronous inputs (toslink) to the upsampler does the upsampler clock out remain to the dac W3? If so does the Vivaldi clock help at all in this case or are we strictly off the toslink clock?

Also not to lose sight of my lip-syncing neurosis for dcs employees, I turned up my tv volume along with my toslink dcs driven speakers. When the upsampler output only requires AES3 and not AES3+4 outputs, it seems more accurate and not “thick tongued” dialogue from a slight echo of tv sources on both my speakers and tv speakers. I feel there is some delay when going dual AES. End of the world, no. Was just hopeful to have one set of settings for all inputs at their best.

Yup! Even with the Vivaldi Master Clock in place, if you’re listening to a Source that has no Clock input (like your LG TV), then the system has to sync to that input (the manual showing the full dCS Vivaldi stack should technically include that Upsampler Clock Out connection to the DAC WordClock In3).

Latency issues aside (I haven’t actually explored latency differences), you never have to select “AES3” only, you can leave it at 3+4 all the time between the Upsampler and the DAC even for low bit rate sources, the dCS takes care of it.

This got me to wondering. Is there a reason to prefer the Toslink input in the Upsampler over the DAC?

Good point. From a latency/lip-sync perspective, it may make sense to do exactly that. Clock Sync Mode on the DAC will still need to be switched to “Audio” though, but there wouldn’t be a need for the extra clock-cable out from the Upsampler to the DAC (albeit, there seems no harm leaving that always connected anyway :grin:).

1 Like

Hey Greg, a reason would be to use the upsampler magic on youtube quality concerts. For TV I’d choose lip-syncing accuracy over a modest improvement in sound. Probably the same for movies. Concerts though I’d choose music over lip-syncing accuracy, and most concerts aren’t filmed perfectly in sync anyway. With the upsampler things seemed to line up better with just passing along the 48k to the Dac over AES3 than DXD or DSD over AES3+4. I’m sure there’s a touch of latency in the upsampling, and a touch of latency recombining 3+4 on the dac end. Even doing 192 over one AES seemed to be more accurate than 3+4. Or maybe I’m crazy :slight_smile:
For now I moved the tv back to its room and keeping the living room music only. I just wanted to fully test out the new certified pre-owned DCS purchase so I wouldn’t have any surprises down the road when recombining a music+movie room.

1 Like

Thanks Chuck. Appreciate that extra info.

You’re not crazy Chuck, even dCS can’t break the laws of Physics :grin:, there will surely be added latency on the various Upsampling settings when compared to passthrough…

That said, don’t forget to turn off the Buffer on your DAC to minimize overall latency by 0.72 seconds (for Redbook sources), for 48kS/s sources that setting should reduced latency by around 0.66 seconds or so.

Hi Chuck,

With regards to lip sync, the latency on all synchronous inputs is the same. There is no difference in the latency between AES3, Dual AES 3+4 and TOSLINK for example. The asynchronous inputs (USB, network) are of course different but aren’t really used for video - say for AirPlay which has delay compensation anyway.

As for the group delay through the system, this will depend on the Upsampler conversion and DAC filter. However, assuming the Buffer setting in the DAC is off, this will be <10mS in all cases - well inside the range that is accepted to be imperceptible to the human ear/brain.

The other aspect to consider with regards to lip sync is the TV itself - modern TVs carry out all sorts of processing, which tends to ad a few frames of latency. There are some sources that claim audio arriving before video is twice as detectable as vice versa, so a delay there may be causing some problems.


James, I understand about the group delay associated with some of the [DAC] filters, but can you elaborate on the aspects of “Upsampler conversion” which have group delay characteristics? :thinking:

ps: Oh, by “Upsampler conversion” did you mean the selection of filters on the Upsampler?

Thanks to James and Anupc for the responses. If I’m reading this right, good to know on the Dac side Aes3 vs Aes3+4 vs toslink directly skipping the upsampler has the same latency or really non-latency. And again if I’m reading this right the upsampler, regardless of filter or upsampling setting, has no more than 10ms of latency. I did think passing on 48k directly synced better but there’s also the chance my tv has the couple of frames difference James mentioned above. If the latency of the filters in the upsampler or possibly dac is quantified somewhere that’d be nice to know. So I could upsample YouTube concerts with less latency.