Lessons learnt DMX over IP.. Part(1) there is standards / there is no standards

[Barnabybear]

Am I correct in thinking that to find out at what point the 681 gets ‘swamped’ we would not need pixels, just a bunch of boards and a sequence to test with? If so how would we know at what level the ‘swamping’ starts?
The other thing is there is the link on the Sandevices web site to a show that was in the 100 universes area. It would be interesting to know how they did that?

[budude]

lol - cisco haters... always *one* in the group... ex-Juniper'er are we? hehe

You surely don't need the fastest switch in either case - you only need fast ethernet on this stuff anyway - a decent switch with VLAN, IGMP/snooping support will suffice for a long time. You can probably pick up some of the older Linksys business model switches pretty cheap these days.

[mrpackethead]

No, the data rates required for even 500 universes are not anywhere near what will cause problems for any reasonable switch.

[jstjohnz]

Yes, using multicast is significantly easier than unicast thats fore sure.



in 2008 we hadn't heard of the RGB pixel mega tree. In 2009 just another mega tree, in 2010, we installed some pretty large scale commercial shows, in 2011 we saw widespread deployment of pixels domestically.. THeres a trend, and in 2012 i think you'll start to see some pretty huge domestic installations who start to get scaling issues.




Eight universes in total on the wire?

8 Universes processed by the controller to drive pixels. Since the packet selection is done by the hardware sockets on the wiznet chip I don't think there's a practical limit as to how many universes can be on the wire. I've run more than 100 universes with no issues. Theoretical limit for 100base T is about 400 universes I think.

[mrpackethead]

nice work.... that wiznet processor is pretty useful from that respect then. So thats good news for 680/681 users..

[jstjohnz]

Am I correct in thinking that to find out at what point the 681 gets ‘swamped’ we would not need pixels, just a bunch of boards and a sequence to test with? If so how would we know at what level the ‘swamping’ starts?
The other thing is there is the link on the Sandevices web site to a show that was in the 100 universes area. It would be interesting to know how they did that?

This is somewhat of a dupe to my previous post, but the difference is that the wiznet chip on the E681 filters the incoming packets on the chip, and only the packets that belong to that controller get passed on to the CPU, rather than the CPU having to examine every packet that comes down the wire.

I don't think there's a hard limit with this type of system, rather the limitation would be when the ethernet network itself starts to get loaded. 100 universes would be about 25% of the theoretical capacity of a 100mb ethernet channel. Each packet is roughly 680 bytes, 40 times/sec = 28kB/sec/universe. 100 universes would be about 24 megabits/second.

To test this limit you don't need anything other than a single controller and some pixels. Just crank up the sender to send 100 universes of data, say a 170x100 matrix in Madrix, or opern 100 sendsers in Sacnview, and try addressing the controller for different universe ranges and watch for dropped packets.

[jstjohnz]

nice work.... that wiznet processor is pretty useful from that respect then. So thats good news for 680/681 users..

The downside is that you're limited by the number of hardware sockets on the chip. For example, in the current design, the 4th E1.31 universe has to be shared with the web server. Not an ideal situation, but it's workable. The 5200 chip should allow 8 universes of pixels, or 7 universes of pixels and 1 dedicated to the web server.

The choice of CPU is a tradeoff also. The prop is great for being able to drive multiple pixel types simultaneously because everything is bit-banged with the multiple cores, OTOH the memory limitations make it hard to produce an elegant looking web interface. But I think I'm veering off topic...

[mrpackethead]

Yeah, i see the data sheet says up to 25Mbps processing, so thats around 100 universes... have you tried at 200?

[jstjohnz]

I have not. But you've got me curious now, and it shouldn't be hard to test.

[mrpackethead]

Its got me curious as well. Whats apparent ( and this thread in a strange way ) is a couple of different way of solving the problem..

(a) use a mechanism like IGMP to stop packets arriving that are not supposed to be arriving or
(b) use enough horse power so your device can cope with all the traffic.