X-Dante latency ?

[Heavy90]

hey, my two cents on this. i dont want to offend someone, just share my knowledge and what i have experienced. if you think i'm wrong than lets discuss. Or maybe i missunderstood someone. than im sorry

The overall latency on a Dante network is determined by the slowest device on the network. So, even if you are getting numbers like 250 us (as I am with my X-32 I/O Reaper) on individual devices, because I have a 2-channel Dante audio with a slower Ethernet connection, that limits the entire network to that device.

yes...and no. you set a latency in dante controller for each dante device. this is the time the ethernet pakets should come in. That means, they are faster, but are not allowed to be slower. then, the device waits for the rest of the time to its setted time and forward it. So, every paket comes in at the exact same time that is set up. That means, even if the average latency displayed in dante controller is lower, the latency is what you set up. not quicker.

i think the x-dante card ship with preset 1msec or 0,5msec. every subscription has its own latency setting. so in one network with four devices ("A","B","C","D") you can connect "A" to "B" at 1msec and "C" to "D" with 0,250msec. so, from what i know. the overall latency is not set by the slowest device on the network, it is the slowest device that it has subscription to (really important i think), and the way the latency per device is set up.

for example... you have two mixing consoles with dante cards running 500usec latency and a DVS (Dante Virtual Soundcard). DVS per se is really slow compared to hardware devices. The fastest setting is 4ms. You can subscribe signals between the two mixing consoles at 500usec and between one console and your DVS with 4ms, even on the same network. They arent all dropped down to 4ms. (And for DVS. ASIO adds antoher latency to the 4ms)

I can't remember the exact figures, but it is less latency than aes50...

and definitely not less latency than aes50. the minimum latency in dante controller u can set is 125usec. usually not all devices can do that and mostly it dont work ether. mostly, it is set to 500usec. AES50 in the other hand is about 170usec static. i've measured it. if you are hopping over more aes50 devices it scalles 1:1, so add 170usec for each device in cascade. if you route an aes50 stream in a channel and then to an aes 50 out, u can add 400usec channel processing to it. but thats the same if you do that over dante.

the allen&heath DSNAKE is about the same as AES50 from what i've measured.

[MJKlein]

The network speed is determined by the slowest device. I mean the system latency.

https://www.audinate.com/learning/faqs

Is the latency the same between all devices for the whole system?
Yes.

Think about that for a minute.

Latency can be configured to be different between different devices in the same network, and does not have to be the same for all connections on the network. Dante allows you to configure low latency connections for critical audio paths, while at the same time running higher latency connections for a broadcast or recording feed where latency is less critical.

It is important to note that in order to achieve consistent system latency, the latency imposed by different processors on the network should also be taken into account. Dante has no control over the latency of devices such as consoles or speaker processors, and these devices require a certain amount of latency to fulfill their respective functions.

The network latency is not the same as individual connection latencies. In your example, the 2 consoles are connected to the network at high speeds, but that is not the speed of the network. The network latency is determined by the slowest connection in the network. When you say "they aren't all dropped down", well that's really irrelevant because they have to wait for the network to send the packets at the network speed. Yes, you can have different device latencies to the network but the interconnection latency is all the same.

I'm sure that @GaryH has a better way of saying it than I do.

Typically I have around 380 us but when I turn on my Planet 22X interface with the slower network connection, my RedNet card reported latency goes up. I have all my devices set for 1 ms because that is the minimum setting for the Planet 22X. Setting the RedNet and X-Dante cards as low as they can go accomplishes nothing as the network speed is determined by (I'll say it again) the slowest device on the network. The network has to consider the speeds of all the devices.

[Heavy90]

so, we are talking dante latency, not network latency, right? i dont get the point at "...by the slowest device on the network". If you are talking about two units, i agree. But if we add Dante units every connection has its own setted latency. For the Ethernet Network itself its the same. You can have 10 PC's in a LAN. If only one is slow, maybe because it sits behind a couple of 100mbit switches, all other devices may connected to one 1gbit switch will be fast in time and bandwith, except they are exchanging with the slow one. than only the connection to the slow one is affect imho.

Typically I have around 380 us...

Dante is waiting for the configured time, even if the paket came erlier. I think otherwise you would have jitter. If it is to late, you will have dropouts.

even if you say dante controller is saying you have 380usec, your latency is 1ms if you configured it to that. I mean, i've tested it as i wrote above. I had one DVS running 6ms. But the two consoles are exchanging audio in under 2msec Analog In to Analog Out over Dante (Console A to B). Only the DVS Audio came later to Console B. If the slowest device would slow down other devices i wouldnt be able to reach latency under 6ms (just Dante plus AD/DA conversion, channel processing etc.).

[GaryH]

From Audinate on latency:

“Dante Virtual Soundcard is configured with custom latency values for reliable operation. Dante devices with Rx channels that are subscribed to Tx channels from a Dante Virtual Soundcard transmitter will automatically configure themselves to use these higher latency values for those channels only. The latency on all other subscriptions on the receiver is unaffected”

I take that as any other device to device subscriptions outside of DVS the latency is as fast as they can be reliably set for. I may misunderstand. The latency figures within Dante controller pretty much appear to show that. Anyway, I don’t see latency being reported as too much of an issue in most Dante systems. I’m sure there are cases though.

[Heavy90]

...
I'm evaluating using my X32-Core(s) with X-Dante card(s)...
on other stageboxes that have Dante cards / connections, i.e. A&H iDR48 in a club I work

How much will the latency 'X-Dante IN > X-Dante OUT be ?

And maybe someone knows the same answer for the A&H iDR48 too ?
...

Hey @Klaus,

just to come back to your orignal question. sorry for wrapping this up so much.

if i understand you correct your configuration would be as followed:

Input on A&H iDR48 ~200usec
Dante from A&H iDR48 to X-Dante X32 500usec
processing on the X32 Channel, routing to bus or main out 400usec
(using FX Rack Insert on X32 650usec)
Dante from X-Dante to A&H iDR48 500usec
Output on A&H iDR48 ~200usec

=1.8ms total system Latency Analog In to Analog Out using Dante. (2.45ms using FX Rack on X32)

using AES50(e.g. with S16) it would be 1.14ms total system latency (1.79ms using FX Rack on X32)

i estimate the A&H iDR48 converter latency, but it should be around that. the other numbers i would call a fact.

[MJKlein]

I don't see the point that's being made, sorry.

[Heavy90]

I don't see the point that's being made, sorry.

what do you mean?

Klaus has asked about the X-Dante latency. I think GaryH and you were not right about that. GaryH's Post i understand as approval. And because of me spreading out the discussion off from the original question i thought i give him an exact and short answer. (see above)

once again... if you set your dante connection e.g. to 500usec. your latency is 500usec. If a paket arrives later, it will be discarded by the dante layer. no other network partners (faster or slower) will add latency to others, unless they using all the needed Bandwith.

And AES50 has lower latency than Dante!

thats the points i wanted to make.

[Heavy90]

i might was getting you wrong, you said "...that limits the entire network to that device." i was reading 'limits the entire network'. so on that point youre right of course. I'm sorry.

"The overall latency on a Dante network is determined by the slowest device on the network." thats not right imho. if you mean two devices back to back , yes, otherwise, with more devices, no.

[MJKlein]

what do you mean?

I mean I don't care, lol.

"The overall latency on a Dante network is determined by the slowest device on the network." thats not right imho. if you mean to devices back to back , yes, otherwise, with more devices, no.

"imho"? So, you don't really know? You're guessing?

I said what I said because I read that on the Audinate website and also heard it in their video tutorials. Now, if I'm wrong, I really don't mind and I'll take it all back if you want. This is a case of "getting technical" on a point that does not matter really. I use Dante every day in my studio and it just works. What is concerning to me is the why. You've been polite and no offense has been given and none taken, but this is beginning to sound like the need to be right rather than a useful discourse. It's ok with me and this is an open forum but I'm done with this topic.

Everything is cool. I am just busier now than I was 30 years ago.

[GaryH]

@MJKlein @Heavy90 I probably made this more confusing than it needed to be and I am sorry. I think I was (half) remembering the X32 published spec for Network I/O latency (stagebox in > console processing* > stagebox out) (via aes50) of 1.1ms and then being aware of Dante device to device latency figures which can be sub ms. The whole subject of latency can be very confusing. I am no latency expert. I only know it when it gets in my way I am pretty sure the latency issues for @Klaus won't be a big problem and this seems to be born out by @Heavy90 latency findings. The more Dante users/expertise in this forum the better imo.