PDA

View Full Version : Continous Artnet streaming...



OlliR
21-11-2007, 12:44 PM
Situation right now is that Pixelmad only outputs Artnet universes if changes of values occurred on this universes.
It will also update the values of a not changed universe after a period of some seconds.
This method complies with the original Artnet protocol and helps reducing Artnet traffic.

The problem is if the recieving unit drops (whyever) one packet of the traffic it might be in the wrong state for another 5 seconds...
I have seen this at Live Earth Hamburg for the first time - and I thought it was the fault of the ELC node8 hardware used to decode Artnet to DMX.

I would prefer an option to have a continous Artnet stream at 25 updates per second. This will work in 100 Mb/s and !gb/s networks - of course it will be a problem with original 10 Mb/s hardware.

There is also another issue with non-continous Artnet sending: What should the recieving unit output on the DMX side between new packages? I can only repeat the last packet.
As the DMX refresh rate is different to the playback rate of a movie you will see frame holds or even frame drops on the output side - that is not in favour of fluently running playbacks...
The Artnet interface in the SchnickSchnackSystems Power Supply 4E does sychronize the output to the Artnet input. That means one packet in - one packet of DMX out. This way the output will be synced to the playback rate.
It it will not work correctly with non-continous streams.

What are others´oppinions on this?

samsc
03-01-2008, 08:00 AM
Oliver.

Your dmx nodes arent working correctly-

i have had nodes not working correctly reported on multiple occasions.

From multiple different manufacturers-

Martin nodes on rush.
etc.

---

The nodes need to be fixed- continuous mode with the devices will just make things worse - and probably prevent them from working at all.

I cant fix a problem in the nodes by sending more data to them.
This will just make things worse-
these devices overload really easily.

Suggest you contact elc or buy something from luminex?

---

There is no way to guarantee ANY network timing- with ethernet-
its all buffered multiple times.


The Artnet interface in the SchnickSchnackSystems Power Supply 4E does sychronize the output to the Artnet input. That means one packet in - one packet of DMX out. This way the output will be synced to the playback rate.
It it will not work correctly with non-continous streams

what do you mean "It it will not work correctly with non-continous streams"?

I suspect if they are saying this - they have not got their algorithm correct.

samsc
03-01-2008, 08:14 AM
The problem is if the recieving unit drops (whyever) one packet of the traffic it might be in the wrong state for another 5 seconds...
I have seen this at Live Earth Hamburg for the first time - and I thought it was the fault of the ELC node8 hardware used to decode Artnet to DMX.



were you displaying still images?

jasonrudolph
03-01-2008, 04:01 PM
this was the same problem I was seeing using the Martin Ether2DMX boxes for the special olympics, and it was the same problem on Rush. When you would do a fast bump.change, the previous state would stay for about 5 seconds.

We changed to Virtuoso Node 2s (artnet) and the problem went away. On Rush they changed to the Artistic Licence boxes (I think) to solve the problem.

Richard, maybe it would be good if you put up a page somewhere with known problem boxes and successfully used ones, to help people out in speccing the right gear..

OlliR
06-01-2008, 09:01 PM
were you displaying still images?

I had the biggest problems when flashing layers with still images. So there was really only one Artnet package sent and that might have been lost...

Maybe it is not necessary to have a continous stream but at least a defined number of repeats before stopping...

Could make life easy as you not always in a position to specify what make of nodes will be provided for your system...

By the way Catalyst is the only media server that does reliably play 25 fps on Artnet installation +20 Universes. I tested some other systems - and was not satisfied at all...

Cheers,

Olli

OlliR
31-01-2008, 12:18 PM
I start to agree that a continous stream when not necessary won´t help anyone...
But what about a selectable number of repeats of the last change before stopping to transmit a universe?

I worked with Enttec Datagate and had to see that these units didn´t work good with Pixelmad as well...

The problem is that I am in charge of specifying Catalyst equipment but when it comes to ArtNet node the lighting company will supply whatever they have in stock - and they will say Catalyst is the problem because our nodes work fine with Hippo and grandMAvideo...

It is then a really hard point to tell someone that the problem is in their nodes and not in my server...

samsc
31-01-2008, 12:39 PM
I start to agree that a continous stream when not necessary won´t help anyone...
But what about a selectable number of repeats of the last change before stopping to transmit a universe?

I worked with Enttec Datagate and had to see that these units didn´t work good with Pixelmad as well...

The problem is that I am in charge of specifying Catalyst equipment but when it comes to ArtNet node the lighting company will supply whatever they have in stock - and they will say Catalyst is the problem because our nodes work fine with Hippo and grandMAvideo...

It is then a really hard point to tell someone that the problem is in their nodes and not in my server...

Did you use routable artnet direct to the ip address of the boxes?

--

Its really really not me- if the units dont work.

They are probably overloaded and dropping data - then they blame me.
Im trying not to overload them even more by only transmitting when it doesnt change.

enttec
07-02-2008, 10:34 PM
Olli

Sorry to hear you had a bad experience with our Datagate, there are a couple of things I'd like to clear up, so hopefully people using our gear will have everything working as they want it to.

There were issues with old versions of the firmware and multiple (> 12) arnet universe broadcast. The fact is that when we designed it, it never came to mind people would want to broadcast artnet in such large quantity, we thought unicast would be the norm.

Since V2.0 we greatly improved this and even perfomed a packet loss/latency shoot out at one of the large production houses here in Australia where we compared the Datagate against other devices, some of which where mentioned above.
In the end the Datagate outperformed all of them, packet loss was minimal and timing on all the 8 ports was identical, all other devices had delays between the various ports.
This is up to around 35 to 40 universes, which is the limit for a 10B T network.

In a broader sense though broadcasting that much data just isn't good practice you load up all the devices on your network and even if you have 100bT it just pushes the limit a bit further but the limit is still there. Unicast and a properly design network should be used for large install as you have NO limit to the number of nodes universes.

We have helped users with large video displays installations using datagates, every time it worked perfectly when they used unicast and syncronised the datagate dmx port refresh rates with the refresh of the sending software.

As these projects/gigs get more and more complex it becomes less plug and play and requires some network engineering. We are always happy to help our users and share our expertise to make things work for them

Don't hesitate to contact me if you have any questions

Thanks

Nicolas Moreau
www.enttec.com

OlliR
08-02-2008, 09:10 AM
Don't hesitate to contact me if you have any questions

Thanks

Nicolas Moreau
www.enttec.com

Thanks Nicolas,

I will get back to you if I need assistance for future productions.

All in all the Datagates worked fine - It is just sometimes when going to a blackout without fadetime that some outputs keep sending out the last values - so my best guess is that the OFF packet got lost - that is why I suggested repeating packets before stopping to transmit...

Compared to some other products the Datagates performed fine. Good to now they do work with Unicast as well.

enttec
08-02-2008, 08:20 PM
That's one of the main problems with DMX over Ethernet systems such as Artnet. It's UDP, so packets can get lost and if they do it can have devastating visual effects as we all know. Murphy's law does say it will lose the off packet patched right in the middle of your screen :eek:

I think that continuous unicast would be the best solution, I can tell you that the datagate can definitely handle 25fps unicast for all 8 ports.

Thanks

Nicolas
www.enttec.com

samsc
08-02-2008, 10:30 PM
That's one of the main problems with DMX over Ethernet systems such as Artnet. It's UDP, so packets can get lost and if they do it can have devastating visual effects as we all know. Murphy's law does say it will lose the off packet patched right in the middle of your screen :eek:

I think that continuous unicast would be the best solution, I can tell you that the datagate can definitely handle 25fps unicast for all 8 ports.

Thanks

Nicolas
www.enttec.com

thanks nicolas.

please try to test with continuous data up to 30fps across all 8 - thats what the americans use.
people do use this to playback video across huge backdrops on tv shows - and its vital that nothing ever drops a frame - really vital-
thats why im baffled by the frame drop thing- this stuff almost always gets used at peak load. hammered.

the test you need to make with your boxes is this - strobe flash black and white at maximum speed with no glitching - ever.

This is how i benchmark these systems - if they cant handle this -they simply arent good enough.
Have to be able to handle black white strobing with no glitching and no latency.

its comes across as a huge glitch on shows - you dont want section of a backdrop ever go missing.
I did some tv shows last year with 100m of backdrop- and it all needs to work.

on the size and scale that this stuff is visible - nothing should ever drop anything.

i have never seen udp packet loss as an issue

richard

enttec
09-02-2008, 01:24 AM
Richard

This is how we test, and we haven't seen any problems with unicast.
If someone, has experienced differently please contact me

Nicolas
www.enttec.com

enttec
09-02-2008, 02:32 AM
i have never seen udp packet loss as an issue

richard

Packet loss can happen anywhere, it's impossible to make sure it never happens.
A switch will drop packets, the nodes can drop it if 2 packets arrive too close to each other or it get's busy ....
The tcp/ip stacks that all of us use can drop UDP without warning for a number of reasons.

What I'm saying is because there is no ack, there is no way of knowing that the off frame was ever received. UDP was designed for cases where you don't care about losing packets, in these applications we do.
So there are 2 options either switch to TCP which was designed specifically for making sure every packet got to destination (unlikely) or stream Unicast UDP packets continuously. That way if a packet does get lost it will only be until the next update which would hopefully very shortly after.

Thanks

Nicolas
ww.enttec.com

samsc
09-02-2008, 08:54 AM
im my experience doing this for 4 years now - udp/ethernet packet loss has never been an issue.
its been dodgy software in various artnet boxes that broke things.
Artnet box designers never expected the continuous volume of traffic.

TCPIP is has not been necessary.
TCPIP has poor response to failure in realtime systems.
I would say if boxes cant handle udp in realtime - handling tcpip is even harder - and much more complex.

Its much better to work in open loop.
UDP Failure recovery times are immediate - and require no re-negotiation.

Handling TCPIP re-negotiation is often not handled correctly either. ( i remember pulling the network cable out of a hog3 years ago, and it never came back online... )

samsc
09-02-2008, 09:01 AM
I count dmx input network packet loss- i read all output artnet packets ( and artnet status ) from the network input.

I check the sequence id in the artnet packet - and count the errors.

I have never seen a significant number of packets lost over the nextwork.

enttec
09-02-2008, 12:09 PM
Artnet box designers never expected the continuous volume of traffic.

Yes, I know and we fell in that trap too.
First versions of the datagate firmware would die if you send more than 12 universes, it never came to us people would be broadcasting 40 universes of artnet.
New versions can handle this a LOT better but it's still prone to missing packets when you get to the upper limit.
With Unicast it's not an issue and never was.
Unfortunately, situations where box manufactures and software designers can talk aren't very common.
So this is good Richard if we know what the other is expecting it allows us to make the whole system work better.
that's my 2 cents any way

Cheers

Nicolas

www.enttec.com

samsc
09-02-2008, 09:25 PM
So this is good Richard if we know what the other is expecting it allows us to make the whole system work better.
that's my 2 cents any way


www.enttec.com

thanks nicolas.

i appreciate your input.

this stuff gets hammered - and the slightest flaws show up....

most of the shows i do with this stuff are both huge and high profile.
i did a 30 universe show very early on at the wynn hotel in las vegas - and it really really needed to work perfectly the entire time - and it does as far as i know.
i worked with unicast artnet from the very beginning. artnet always supported this.
they use fibre switches, and that kindof thing.

richard

samsc
15-02-2008, 09:41 AM
thanks nicolas.

i appreciate your input.

this stuff gets hammered - and the slightest flaws show up....

most of the shows i do with this stuff are both huge and high profile.
i did a 30 universe show very early on at the wynn hotel in las vegas - and it really really needed to work perfectly the entire time - and it does as far as i know.
i worked with unicast artnet from the very beginning. artnet always supported this.
they use fibre switches, and that kindof thing.

richard

Simon at stage electrics just did a 48 universe show with luminex boxes, and unicast artnet, and reports no problems at all.

enttec
16-02-2008, 03:01 AM
We had a client in the US that did a 200 universe !!!! display with datagates, and it worked fine too, there was no visible lag between any of the outputs.

Unicast isn't a problem and scales very well. I guess we just need to get the word out there !

Nicolas
www.enttec.com

samsc
16-02-2008, 07:22 AM
We had a client in the US that did a 200 universe !!!! display with datagates, and it worked fine too, there was no visible lag between any of the outputs.

Unicast isn't a problem and scales very well. I guess we just need to get the word out there !

Nicolas
www.enttec.com

thanks nicolas.