|
|
|
12-07-2018, 10:52 AM
|
#1
|
Mortal
Join Date: Jan 2006
Location: Wickenburg, Arizona
Posts: 14,047
|
RME Digiface Dante
Is RME Digiface Dante real dante or does it use DVS?
|
|
|
12-07-2018, 11:10 AM
|
#2
|
Human being with feelings
Join Date: Apr 2015
Posts: 51
|
It’s real Dante to usb
|
|
|
12-07-2018, 11:23 AM
|
#3
|
Mortal
Join Date: Jan 2006
Location: Wickenburg, Arizona
Posts: 14,047
|
In order to really get anything in and out of it, you will need some sort of madi to adat converter and then mic pre, facilities to adat right? Is there a box that provides at least some basic I/O as well?
|
|
|
12-08-2018, 12:47 AM
|
#4
|
Human being with feelings
Join Date: Aug 2018
Location: Maastricht
Posts: 92
|
Quote:
Originally Posted by pipelineaudio
In order to really get anything in and out of it, you will need some sort of madi to adat converter and then mic pre, facilities to adat right? Is there a box that provides at least some basic I/O as well?
|
To get anything in or out of it you'd need a (any) Dante audio interface, no need to involve Madi or Adat. IMHO the Digiface Dante makes sense if you'd want to make use of the RME Totalmix feature, you need the Madi to Dante interface or your machine has no PCIe slot, the PCIe cards are less expensive and can handle more channels. The PCIe cards can also be installed in a Thunderbolt expansion chassis.
As for devices that provide I/O capability and Dante networking, I only now of the Focusrite RED series but these are pretty expensive and only for Mac (and they have fans in them that cannot be switched off :-( ).
In the end there's no way around the fact that setting up a Dante network costs a lot of money but in return you get incredible flexibility and expansion capabilities. One of the things I like very much is the fact that all the channels are always 'there' whether or not the interfaces are actually powered on. The DAW only sees the PCIe card and will never complain about missing channels.
|
|
|
12-08-2018, 12:53 AM
|
#5
|
Human being with feelings
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 14,688
|
Quote:
Originally Posted by pipelineaudio
In order to really get anything in and out of it, you will need some sort of madi to adat converter and then mic pre, facilities to adat right? Is there a box that provides at least some basic I/O as well?
|
I understand that Dante <-> [other digital standard] conversion only makes sense if you already have a lot of precious "legacy" equipment you don't want to dump, and Digiface is provided fur such purpose. On the PC site I understand that Dante best just uses a standard GBit Ethernet interface.
-Michael
Last edited by mschnell; 12-08-2018 at 01:34 AM.
|
|
|
12-08-2018, 01:14 AM
|
#6
|
Human being with feelings
Join Date: Aug 2018
Location: Maastricht
Posts: 92
|
Quote:
Originally Posted by mschnell
I understand that Dante/ADAT conversion only makes sense if you already have a lot of precious legacy ADAT equipment you don't want to dump. On The PC site I understand that Dante just uses a standard GBit Ethernet interface.
-Michael
|
Indeed, you only need Adat to Dante conversion for the case you describe and in that case the Ferrofish Verto series or the Focusrite RedNet 3 are good options.
Using the Dante Virtual sound card (DVS) or Dante Via software you do not actually need a PCIe card but you'll never get the really low latency and high channel count that can be achieved by using a PCIe card.
|
|
|
12-08-2018, 01:30 AM
|
#7
|
Human being with feelings
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 14,688
|
Quote:
Originally Posted by Gerrit
Using the Dante Virtual sound card (DVS) or Dante Via software you do not actually need a PCIe card but you'll never get the really low latency and high channel count that can be achieved by using a PCIe card.
|
I don't see any technical reason why using something else between the computer hardware (which internally always is PCI; I understand that Thunderbold in fact is a kind of PCI-wire, as well) and the Dante network but a (decent) Ethernet card should get you better latency. If you encounter such a problem something must be going wrong in the setup you are testing (or the DVS driver is not designed well, which is hard to believe).
-Michael
Last edited by mschnell; 12-08-2018 at 01:37 AM.
|
|
|
12-08-2018, 12:48 PM
|
#8
|
Human being with feelings
Join Date: Aug 2018
Location: Maastricht
Posts: 92
|
Quote:
Originally Posted by mschnell
I don't see any technical reason why using something else between the computer hardware (which internally always is PCI; I understand that Thunderbold in fact is a kind of PCI-wire, as well) and the Dante network but a (decent) Ethernet card should get you better latency. If you encounter such a problem something must be going wrong in the setup you are testing (or the DVS driver is not designed well, which is hard to believe).
-Michael
|
There's nothing additional between the computer hardware and the network. The Dante PCIe card is actually a highly specialised network card (using four PCIe lanes) which also handles part of the audio processing, hence the low latency. All current models have two network ports for glitch-free redundancy. See, for example, the Focusrite RedNet PCIe. The 1.6ms round trip latency (including network transport) they mention is actually achievable, my mid 2010 Mac Pro runs out of steam when handling lots of tracks but if I set the buffer to 64 samples @ 96kHz I get these numbers. My network has just one switch in it and runs stable at the lowest network latency setting of 0.15ms. With a 96kHz sample rate and the buffer in Reaper set at 128 samples I get 3.6ms round trip latency. When I find the time I'll measure the real analog to analog round trip latency using an oscilloscope. The important thing is that it feels instantaneous to me, even with multiple UAD plugins in the signal path.
|
|
|
12-08-2018, 02:18 PM
|
#9
|
Human being with feelings
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 14,688
|
Quote:
Originally Posted by Gerrit
which also handles part of the audio processing
|
Which "part of the audio processing" exactly ?
I decently doubt that there is any audio specific signal handling in this card. I suppose it does stuff like "Quality of Service" and "Block Size Optimization" which will benefit any high priority streaming application. "Audio" is just the content of the TCP/IP packets on a much higher level in the protocol stack as what can be handled by the Ethernet card.
-Michael (just interested in the theory behind this stuff)
|
|
|
12-08-2018, 09:11 PM
|
#10
|
Mortal
Join Date: Jan 2006
Location: Wickenburg, Arizona
Posts: 14,047
|
Quote:
Originally Posted by mschnell
I understand that Dante <-> [other digital standard] conversion only makes sense if you already have a lot of precious "legacy" equipment you don't want to dump, and Digiface is provided fur such purpose. On the PC site I understand that Dante best just uses a standard GBit Ethernet interface.
-Michael
|
In this case though you could have real dante on a laptop
|
|
|
12-08-2018, 09:13 PM
|
#11
|
Mortal
Join Date: Jan 2006
Location: Wickenburg, Arizona
Posts: 14,047
|
Quote:
Originally Posted by mschnell
I don't see any technical reason why using something else between the computer hardware (which internally always is PCI; I understand that Thunderbold in fact is a kind of PCI-wire, as well) and the Dante network but a (decent) Ethernet card should get you better latency. If you encounter such a problem something must be going wrong in the setup you are testing (or the DVS driver is not designed well, which is hard to believe).
-Michael
|
DVS is 8 seconds latency on top of the dante latency, BEST case. I documented it pretty well in my latest RTL chart.
http://kailuamusicschool.com/tech/ro...tency-roundup/
I dont have an answer as to whether it is purposey crippled on the DVS side, I dont doubt someone could hack it and make it as fast as "real" dante, but for now, it is MINIMUM 8 miliseconds longer than "real" dante
|
|
|
12-08-2018, 09:14 PM
|
#12
|
Mortal
Join Date: Jan 2006
Location: Wickenburg, Arizona
Posts: 14,047
|
Quote:
Originally Posted by mschnell
Which "part of the audio processing" exactly ?
I decently doubt that there is any audio specific signal handling in this card. I suppose it does stuff like "Quality of Service" and "Block Size Optimization" which will benefit any high priority streaming application. "Audio" is just the content of the TCP/IP packets on a much higher level in the protocol stack as what can be handled by the Ethernet card.
-Michael (just interested in the theory behind this stuff)
|
Very likely it is just like UAD, and the "reaL' dante cards are just dongles
|
|
|
12-09-2018, 12:47 AM
|
#13
|
Human being with feelings
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 14,688
|
Quote:
Originally Posted by pipelineaudio
In this case though you could have real dante on a laptop
|
A leptop needs to be optimized for lowest power consumption, hence, IMHO, using a laptop for critical application is not the way to go, anyway. But if DVS adds a lot of latency, this might indeed be "on purpose" and not for technical reasons.
-Michael
Last edited by mschnell; 12-09-2018 at 01:10 AM.
|
|
|
12-09-2018, 12:55 AM
|
#14
|
Human being with feelings
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 14,688
|
Quote:
Originally Posted by pipelineaudio
Very likely it is just like UAD, and the "reaL' dante cards are just dongles
|
That might be. But if DSV implements any crippling on purpose, they need to publicly state this. Otherwise they would be liable for fraud against their customers.
-Michael
|
|
|
12-09-2018, 03:10 PM
|
#15
|
Human being with feelings
Join Date: Aug 2018
Location: Maastricht
Posts: 92
|
Quote:
Originally Posted by mschnell
Which "part of the audio processing" exactly ?
I decently doubt that there is any audio specific signal handling in this card. I suppose it does stuff like "Quality of Service" and "Block Size Optimization" which will benefit any high priority streaming application. "Audio" is just the content of the TCP/IP packets on a much higher level in the protocol stack as what can be handled by the Ethernet card.
-Michael (just interested in the theory behind this stuff)
|
On the Dante PCIe product page it says:
Quote:
onboard hardware processing of packet procession
|
The data packets do not arrive in a neat order that can be handed over directly to the DAW trough ASIO or Core Audio. The incoming packets have to be sorted and transformed into 128 perfectly synchronised audio streams first. This task is handled by FPGAs on the card and there's no way the speed of something implemented in a FPGA can be achieved through software on a personal computer.
Do not compare Dante with AVB in this respect, the way timing is handled on AVB is completely different. AVB requires special switches that handle part of this task, from the article IEEE 802.1 AVB--Audio and Video Sync'd Over Ethernet:
Quote:
The network which transports these audio and video streams must use Ethernet switches that are compliant with the standard. These switches are capable of measuring the transport and throughput delay of each bit of data that passes through them, and send that information along with the information. The receiving device can then properly reconstruct the time reference, synchronize them, and play them back together.
|
In contrast Dante can use any halfway decent network switch but relies on specialised hardware to transform the packet stream into an audio stream really fast, i.e. with low latency.
|
|
|
12-09-2018, 04:05 PM
|
#16
|
Human being with feelings
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 14,688
|
It is true that the TCP layer (other than e.g. UDP) allows for the IP transport medium (e.g. multiple Ethernet switches) to mingle the order of packets and the receiver needs to recreate the correct order before handing over the stream snippets to the next layer towards the application.
But as this is per-packet and not e.g. per-byte, and not complicated at all (you just need to wait for a missing packet and store those coming in until then) this is not something that would benefit a lot from being handled outside of the realm of CPU and RAM. Moreover it in fact is not specific to audio but a normal TCP task.
In fact the multiple audio streams embedded in a TCP stream are Dante specific, But those need to be transferred somehow via the the PCI bus. Providing a per-stream buffer might be beneficial for a fast multi-channel audio driver, as the CPU would not need to handle stripping off the TCP/IP headers and Audio channel division. This results in lower CPU overhead but not necessarily in lower latency - unless hitting the limit of the CPU.
-Michael
|
|
|
Thread Tools |
|
Display Modes |
Linear Mode
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT -7. The time now is 04:29 PM.
|