Im considering to replace my current digital front-end with a Vivaldi stack, and wonder in what way you should set this up to get the most out of the setup. Currently, I’m using a Nuc/Rock connected trough the network (Raat) to a dac with a built in roon endpoint. Is that the recommended setup with the Vivaldi, i.e using the upsampler’s networking capabilities, or are there other options that might be better?
In terms of DSD, what is supported “natively” (DOP I guess?) from roon to the Vivaldi - DSD128? 256?
Final question, how is DSD treated in the Vivaldi? is it converted to PCM before DA, or is it true dsd before conversion to analog?
The best setup with the Vivaldi is to use the Upsampler and its network interface. I’ve yet to find a setup that provides better sound quality or overall integration.
The Vivaldi DAC accepts up to DSD128 using the dual AES inputs or USB input. The Upsampler accepts up to DSD128 on its network interface and USB input. DoP is supported in all cases and required on the USB input. In the case of Roon DSD bitstreams are sent natively to the network card in the Upsampler.
The dCS DACs use a proprietary internal data format called Ring DAC format. In the case of the Vivaldi 2.0 it’s 5 bits at 5.64MHz or 6.14MHz (depending on the base sample rate of the incoming data stream). This is the format that is absolutely optimized for the DAC hardware. It’s neither PCM nor DSD and is best thought of as a hybrid of both and as such can exploit the strengths of each.
It’s difficult to compare formats and pick out a clear best in terms of quality. Ultimately, digital audio is all about making trade-offs in order to accomplish the final goal. PCM is simple and logically straightforward. It’s also nearly impossible to create a truly linear 24bit DAC in hardware (physics gets in the way). DSD is much easier to implement in hardware and is a more efficient transmission mechanism for a given quality level, but it has the trade-off of being very difficult to process (DSP). Due to the nature of the format and its fundamental need for noise shaping (DSP) it’s a difficult format to work with beyond simple decoding.
Ring DAC format is a best of both worlds approach. It doesn’t have the same noise problems inherent in one bit systems, and doesn’t require one to violate the laws of physics when it comes time to build the hardware (and achieve the dynamic range promised by 24 bit PCM). Furthermore the DSP required to prepare the datastream for decoding is less resource-intensive than DSD and maintains much of the flexibility of dealing with PCM data. This has been our approach for 30 years and we’ve perfected the technology over time through constant refinements of the algorithm and hardware.
I am posting under this thread since it also pertains to my Vivaldi and Roon integration:
I am currently running Roon on a Mac Mini (MoJo upgraded with external PS and sound dampening), connected via USB to my VUP upsampler, which is connected to the Vivaldi DAC (latest software v2.01) and Vivaldi Clock via dual AES and BNC.
The Mac Mini (Roon) pulls its music from a Synology DS216+ii on the ethernet network.
Since my current VUP upsampler only supports up to DSD64, I have decided to upgrade to the VPP board. Hoping to receive it within 2 months.
When I get it installed, do you recommend I keep using the Mac Mini via USB ? Or have Roon installed straight on the Synology NAS and run from it through the Ethernet port on the VPP upsampler?
Keep in mind my music is mostly DSD64 rips, PCM44/96, with an ongoing DSD128 rips of my vinyl collection.
Once you have the VPP I highly recommend using the built-in Roon Ready support via the network interface. Given the hardware you have on hand at the moment I would recommend leaving your Mac Mini as the Roon core and just having it communicate with the VPP via the network. It’s possible to use the Synology as the core, but I’ve done this and it isn’t pleasant as the CPU is well below Roon’s minimum spec and there simply isn’t enough RAM to allow Roon to function correctly.
When you get your VPP upgrade done be sure to have your DAC updated to v2.02 as this release includes some code that speeds up the sync time between the DAC and Upsampler. There’s nothing else important in 2.02 so there’s no need to rush out and get the update now.
I have a Vivaldi stack that is connected to a network router via an Aidioquest Cinnamon ethernet 7 cable. Music can be played via Audirvana located on an IMac on the network. I also run Roon headless on a sonictransporter I7 also connected to the network. My experience is that Audirvana has a fuller, crisper, more defined sound quality than Roon does. I use Roon for a graphical interface into my library and Tidal but play music thru A+. FYI, The Vivaldi stack plays thru a Boulder 2010/2060 and Wilson Alexx speakers with Transparent Reference cabling. Anybody with similar experience?
Excellent answers - thank you Andrew!
A few follow-up questions
Is the limitation to DSD128 hardware or software-related? i.e. is there a possibility that 256 will be supported in the future w just a software upgrade?
I record quite a few vinyl 45 rpm records and R2R master tape copies using my Tascam D3000 that supports recording to DSD 128 or PCM 192/24. With my current dac (PS Audio Direct Stream Dac) I prefer DSD for quite obvious reasons. Would you say that decoding with the Vivaldi, DSD and PCM would generate similar results, or is either format better suited?
Final question: despite the obvious one box versus three, are there any difference between a Vivaldi One or a separate Transport + Upsampler + Dac?
This is all managed in software. Due to the architecture of the platform we can pretty much make it do anything we want. DSD256 is definitely a possibility and is something that we have discussed internally. The issue comes down to prioritizing resources in balance with market demand. We have a number of projects in the pipeline now and at this point we haven’t seen enough demand for DSD256 to justify the resource expense. We monitor these things continuously and add items to our roadmap when they gain enough market momentum.
The sad fact of the matter is that there’s almost zero content out there in DSD256.
With the Vivaldi neither format is given preference. We treat all formats equally and don’t ever want to create a situation where we end up with a product that sounds better with one format vs another so we optimize both independently. To that end 16/44 is absolutely outstanding on the Vivaldi and a common comment is, “wow, I never thought Redbook could sound that good!”
If your preference for your vinyl rips is DSD128 then by all means continue with that format.
The two products are functionally equivalent and are very close to each other in terms of performance. The Vivaldi separates do have the advantage of more functional isolation and more dedicated power supplies due to the separate boxes. On the down side they’re more expensive, require more space, and lots of cables. In a comparison of the two systems (4 box Vivaldi vs. Vivaldi One w/ clock) the two have nearly identical sonic signatures. The separates have an edge when it comes to the last bit of harmonic and dynamic resolution.
If one can afford the full stack and cables of commensurate quality then he can purchase the full stack knowing that there simply isn’t anything better. On the other hand, should budget not allow it (either in terms of money or space) then the Vivaldi One with the clock is going to get him most of the way there with a significant savings.
If I were spending my own money I would likely choose the Vivaldi One as I have some space limitations that make the separates hard to deal with in my cabinet. I would have zero regrets in making that choice.
Yes, the mapping algorithm can be changed via a firmware update. The entire platform is software defined so we can update functionality, DSP, and just about anything else through a firmware update. This is how MQA was integrated into the platform.
There is a plan to release enhanced mapping algorithms for the Rossini, but we’re still trying to sort out the timeline (we still need to finish MQA first ). Hoping for the latter half of this year, but we have a number of things that need to get out the door first.
For DSD I tend to stick with F1 or F2 on the Rossini. Each successive filter removes more and more of the ultrasonics. F4 is the most severe and is really only there for those amplifiers which become unstable when fed with ultrasonic signal. Filter 5 on the Vivaldi is a completely new approach and is some of the “special sauce” that’s unique to that platform.
I’ve stuck with F1 since the sound is lovely with it. I have not tried any of the other filters (other than M1 - the MQA filter, selected automatically - obviously). My SET amp doesn’t really have any ultrasonics…
Use whatever sounds good to you. Each setting is a little different and it’s going to come down to personal preference.
With DSD I like Map 1 on the Vivaldi DAC and Filter 5. You may also want to try upsampling to DXD in Roon (use the Max PCM Power of 2 setting) and then use Map 3 and either filter 3 or filter 6 on the Vivaldi.
This is somewhat tweaky and tinker-y, but I’ve had good results with using fiber ethernet media converters to optically isolate my dCS Network Bridge (and other network endpoints) from the rest of my home network. I brought the gear to do this into my local dCS dealer and they were impressed enough with its effects on their Vivaldi stack (as well as an Ayre QX-5 Twenty) that they ordered the same components to use in their stores.
I did a lot of experimentation, and at least in my setup, I discovered that running the Network Bridge (and UltraRendu) ethernet interface at 100Mbps sounded better than gigabit, and that 5V seemed to be the quietest supply voltage for the media converters. The power supply at both ends made a difference, but the “receiving” end (closest to the endpoint) made the biggest difference. Also, single-mode fiber seemed to be much quieter than multi-mode.
We’ve investigated these solutions and found that they’re generally more trouble than they are worth. In some situations they resulted in connectivity issues and in controlled testing we found that they made no appreciable difference in performance (and in some cases they actually made the sound worse). While customers are absolutely welcome to use them they are officially unsupported by dCS. Should a customer have a support issue we’ll request that the fiber converters be removed so that troubleshooting can be performed using a standards-compliant Ethernet cable.
A completely understandable stance. In my own experimentation, I ended up with a lot of variations on this general idea that ended up sounding the same, worse, or just different. In my own situation, the hassle was worth it; in both systems in my apartment, the fiber cables replaced long runs of copper ethernet cable that ran parallel to in-wall mains wiring or power cords for several meters. Optical isolation seemed to get rid of some kind of bad interaction between the ethernet cables and mains wiring that didn’t effect data integrity, but had an effect on sound quality. Also, I think running the ethernet interface on my dCS gear at fast ethernet speeds, rather than gigabit, improved sound quality regardless of the involvement of optical isolation.