Ethernet and fixtures

[dave e]

I know that the reality of fixtures (mainly of the moving variety but also pixel stuff)themselves having ethernet ports is a bit in the future but having read so much about the main drawback being the star topology of the network, why couldnt each fixture be given a hub and a network card? This would essentially make each fixture a hub/node in the network, and as far as I know, as long as the hub is active (not passive) the size of the netowrk is in theory huge being that hubs are just daisychained on the uplink terminal. Surely this would also allow daisy chaning of fixtures using cat5 just like DMX?

The main disadvantage of networks as I have found at the moment is the inherent cost of nodes for protocol-DMX translation (around £500 for the latest Strand one) and the number you may need in a large venue. If say you needed eight DMX ares (some with multiple universes) around your venue thats at least 8 nodes, or daft runs of DMX between areas from each node.

Is this just my over simple understanding of the network or have I maybe stumbled across something?

Cheers in advance for any thoughts.

Dave

:)

[bruce]

Speaking as a network engineer....

 

First of all, terminology. Switches, not hubs. Nobody uses hubs any more, although in the past they were commonplace. When using hubs, there were very strict rules about daisy-chaining devices. With switches, these restrictions are less critical, but there are other issues.

 

Why is a star topology so bad? It means you can't daisy-chain, but there are some huge benefits. Specifically, it has very low latency. All devices are connected to a central switch, which will have active management. Monitoring traffic flows is easy. And failure of one device will not take out any "downstream" units.

 

Having spent years building data networks using both topologies, I'd prefer a star any day. Far more reliable and robust.

 

Bruce.

[Ken Coker]

Might I refer the honourable gentlemen to page 25 of:

 

Rock Solid Ethernet

Wayne Howell

Entertainment Technology Press

ISBN 1904031293

 

Ethernet is just not suited to the current control topologies we use in most aspects of the entertainment industry. This does not preclude:

 

I) Specialist applications and installations - I love my eDMX units for long control runs

 

ii) The situation changing radically in the short to medium term as technology advances.

 

Cheers

 

Ken

[Ekij]

Does Ethernet not suffer from deterministic issues (you can't be sure when the packet will get where it is going).

DXM does not suffer from these issues.

Small latency in lighting is going to be noticed.

[martinw]

The most compelling reason for NOT building an ethernet switch into each fixture is that if any one fixture loses power, its switch will also fail. This means that all the other fixtures which depend on that switch for data will stop responding.

 

Martin

[bruce]

if any one fixture loses power, its switch will also fail. 

 

<{POST_SNAPBACK}>

 

Hence my dislike for daisy-chained arrangements. DMX is different - it's a passive "through" connection.

 

Does Ethernet not suffer from deterministic issues (you can't be sure when the packet will get where it is going).

 

Yes, in theory. You leave the error correction to the higher layer protocols (c/f TCP/IP)

 

In practice, in a well-designed small network (as opposed to WAN) it's not really an issue ....

Small latency in lighting is going to be noticed.

<{POST_SNAPBACK}>

 

What do you regard as "small" ?

 

I've just had a look at a network I'm doing some work on at the moment. We've got an end-to-end latency of 1.5 milliseconds. Would that be acceptable?

 

The two ends are 50 miles apart.

 

On a smaller scale (1 or 2 switches, and tens or hundreds of metres of cable) latency is often of the order of a few microseconds.

 

Bruce.

[andystone]

Lets face it there are pros and cons to both. We currently have quite a few users who have purchased Ethernet nodes to go with their Compulite Vector consoles. Partly to get more outputs from the console, but also there are a number of advantages in using Ethernet such as inherent splitting and merging of DMX signals, sending data wirelessly (and quite cheaply) for all kinds of applications such as TV outside broadcast and architectural lighting.

 

I don't see users connecting Ethernet to fixtures for many, many years, there are simply too many DMX fixtures around. However, I do see a growth in the use of Ethernet for distribution, not just for DMX as if you have your Ethernet network, you can easily have DMX inputs/outputs, remote video, remote control, etc, anywhere in the building.

 

I also disagree with something said earlier, DMX sends data down a cable and has no idea whether it has got there, Ethernet does!

[bruce]

No real disagreement with any of the above....

 

I also disagree with something said earlier, DMX sends data down a cable and has no idea whether it has got there, Ethernet does!

<{POST_SNAPBACK}>

 

except that one. CSMA/CD (ethernet) does not have any handshaking. You throw the packet into the network, and hope it comes out the other end. The ethernet protocol does not guarantee delivery. You need to build this in (if required) at a higher level

(eg tcp/ip)

 

Don't worry - I'm not going to mention the ISO 7-layer model :)

(unless someone really, really wants me to...)

 

Bruce.

[peternewman]

having read so much about the main drawback being the star topology of the network, why couldnt each fixture be given a hub and a network card? This would essentially make each fixture a hub/node in the network, and as far as I know, as long as the hub is active (not passive) the size of the netowrk is in theory huge being that hubs are just daisychained on the uplink terminal. Surely this would also allow daisy chaning of fixtures using cat5 just like DMX?

<{POST_SNAPBACK}>

If everyone hates the star topology so much, then why are there so many manufacturers making DMX splitters? After all these are effectively just the DMX equivalent of an active hub, which is used to create a star topology.

 

The most compelling reason for NOT building an ethernet switch into each fixture is that if any one fixture loses power, its switch will also fail.  This means that all the other fixtures which depend on that switch for data will stop responding.

<{POST_SNAPBACK}>

Surely you could just build an 4-pole (or 8-pole for Gigabit) relay in which is normally passthrough, however when you apply power the switch and fixture would be connected into the line, so when you loose power it just goes back to passing directly through and missing the fixture. I'm sure I remember reading of something that did this in another context but I can't remember what it was.

 

Don't worry - I'm not going to mention the ISO 7-layer model :huh:

(unless someone really, really wants me to...)

<{POST_SNAPBACK}>

Go on please, it will bring back all my wonderful memories of A Level ICT :).

 

Ethernet will also allow all the reporting and so on that I presume will be implemented with RDM, it would also allow a fixture to easily be soft addressed by the desk when the rig is finished, without an address needing to be set on the fixture, although whether this will be implemented and whether it is necessary is another question.

 

PN

[Wilf dLampy]

Something else that ethernet could do, which DMX (currently) can't, is syncronise fixtures.

 

As a very basic example, a rig with 10 scans in it in a nightclub. I want the scans to colour cycle slowly through their wheels. But I want them all to do it together (i.e. dancefloor goes red then yellow then green then.....

 

The Current method involves programming a series of steps, one for each colour (or using a Shape Generator on an Avo Pearl etc). However, it would be nice if the built in scan feature "rotate colour wheel slowly" could be activated. The console sends one command to the scans saying "you 10 are going to do the same colour at the same time, OK" and the fixtures get on with it. The console then has less load to deal with, and if there is a signal failure (desk crash, cable break etc) the scans carry on doing what they were last told- rotating colours.

 

I also like the potential for flood wiring venues at a low cost. Putting outlets anywhere where a fixture may want, at some point, possibly, to be plugged in. This is what happens with offices (computer/phone outlets everywhere) and therefore the parts and skills to do the job are readiily available.

[dave e]

If everyone hates the star topology so much, then why are there so many manufacturers making DMX splitters? After all these are effectively just the DMX equivalent of an active hub, which is used to create a star topology.

 

I would assume that the reason for the splitters is to allow a DMX network to have multiple access points. If you only have one outlet (as in fact my venue does currently but we have 3 ethernet lines!?!) you have to put a splitter or a device with DMX in and through at the outlet, and run the cables to where you need them e.g. an lx bar. Then without a splitter you would need to run off the end of the bar, to the next and the next and the next. This has obvious implications for moving bars in and out (you'd needto keep plugging and unplugging the links between the bars) and you'd very rapidly max out the number of fixtures and length of DMX run possible. The splitter allows say line run for each bar, or two lines for the stage and three for the overheads etc.

My only real annoyance with star topology has come from working with the older Vari Lites. These had to have a smart repeater unit with 6 outlets to drive 6 VL's. The lamps themselves were linked to a smart repeater via thick CPC cable, and the smart repeater (for Arc lamp models) was linked back to a chassis via VL socapex. This has caused us head aches in the past for positioning of the smart repeater, having things like cpc cable coming off the ends of a bar to reach another fixture hung on a boom or ladder, or having to run a cable through dips from one side of the stage to the other, for say a VL5 and doing the same for a VL6.

I just thought that if the option to avoid star was possible it should maybe be seized and developed to avoid problems in the future.

Cheers for all the replys - some good arguments both ways for and against star stuff.

Dave