How many F16's can be daisy chained?

[csmith]

Is there a limit to how many F16 pixel controllers can be daisy chained on the same ethernet switch port?

What's the best practice on how many controllers should be daisy chained?

[k6ccc]

I don't really think the limit would be the number of controllers, but rather the amount of data being sent to them.  At least the v3 versions are 100 Mb/s max so I would limit those to about 200 universes if using E1.31 or a bit more if using DDP.  I don't have the ability to run DDP so although I know it is more efficient, I don't know how much better it is.   I also don't know if the newer F16 versions will do Gigabit.

[jnealand]

When you say daisy chained is that what you really mean?  Or is it how many can be connected to a switch?  I suspect those two scenearios will be very different.  Daisy chaining would be F16 > F16 > F16 vs a switch with an F16 connected to each port.

[JonD]

I don't think there are written technical limits.  I have seen others warn not to get too crazy, but I have not ever had any issues personally.  I have had three chained together without any issues, but they were smaller F4v3 at the outer edges of the display.   

You could always ping the last one in the line, and watch the response time during a busy sequence.  If the "Time=" Millisecond starts to greatly increase, or you start dropping packets it might be too many times.

[algerdes]

Don't know if it still applies, but there was a limit of two.  It's primary reason was the chip that provided the onboard "switch" was not really all that robust.  I think Dave Pitts was the one that passed that information, but check with him for sure.

[darylc]

The limiting factor is really how many channels you can fit in 100Mbps.  I know David Pitts has a rule of thumb, it's more than 2.  But in reality with how cheap switches are these days I'd prefer to star out of a main switch.

[JonD]

Best practice would probably be to plug each controller into its own switch port as the Falcon controller CPU is not super powerful.  I only chain from controller to controller when I am at the wiring distance level.  To avoid having cords run across the driveway and some outside stairs, I feed one area from the back of my house.  It is around 250-275 feet (76-80 meters) to the first controller.  I use the first controller as a repeater/booster to get to the other controllers.  I have considered installing a switch next to the first controller, but have not had any issues and if it is not broke I don't fix it.

If you are trying to chain a mega-tree and large matrix it would probably be better to install a switch, but if you only have a few lights you could most likely chain them and get along fine.

[algerdes]

The limiting factor is really how many channels you can fit in 100Mbps.  I know David Pitts has a rule of thumb, it's more than 2.  But in reality with how cheap switches are these days I'd prefer to star out of a main switch.

As an example, we do not ever plan to use the secondary eth on any of our controllers.  All controllers are fed from a full-blown switch.  We have, in an emergency, fed from the Falcon to a prop that was moved close to it and there was no switch port was available.  In all the years, we have done this only twice.  

[csmith]

Per your suggestion I pinged David Pitts.  Here's the response from the Falcon guru himself:


At 40 fps, I would count total universes used by all controllers that are cascaded. Stay under 192 universes total.

At 20 fps double it to 384.

[darylc]

Best practice would probably be to plug each controller into its own switch port as the Falcon controller CPU is not super powerful. 

Fyi the CPU has nothing to do with it.

[JonD]

Best practice would probably be to plug each controller into its own switch port as the Falcon controller CPU is not super powerful. 

Fyi the CPU has nothing to do with it.

Where is the bottle neck then?

[k6ccc]

Best practice would probably be to plug each controller into its own switch port as the Falcon controller CPU is not super powerful. 

Fyi the CPU has nothing to do with it.

Where is the bottle neck then?

Since the F16 (at least the v3 series - don't know about the v4) is limited to 100 Mb/s the bandwidth of the network connections becomes the bottleneck.  For E1.31, the bandwidth is about a quarter Mb/s per universe, and remember that with standard Ethernet, you are limited to about half of the network capacity.  Therefore, about 200 universes.  DDP is more efficient so you can get a bit more (I don't know how much more since I can't run DDP).

[k6ccc]

And for whatever it's worth, the most I have daisy chained on a single network port is my pixel tree (F16v3 with 29 universes) with my P10 matrix (36 universes) downstream of the F16v3.  That was done simply because it was convenient from a physical wiring arrangement standpoint.  Never had a problem doing that, but that is not stressing a 100 Mb/s connection.  Assuming I make planned changes, the P10 matrix will become 72 universes and some wiring changes will make it convenient to have that a dedicated port from the East Attic network switch (which has Gigabit back to the show computer).

[JonD]

For E1.31, the bandwidth is about a quarter Mb/s per universe, and remember that with standard Ethernet, you are limited to about half of the network capacity.

It is a 100mb limit, and has nothing to do with the Falcon controllers?  So if I only have 8 universes on controller, there would be no issues cascading 24 of them together?

[k6ccc]

For E1.31, the bandwidth is about a quarter Mb/s per universe, and remember that with standard Ethernet, you are limited to about half of the network capacity.

It is a 100mb limit, and has nothing to do with the Falcon controllers?  So if I only have 8 universes on controller, there would be no issues cascading 24 of them together?

I don't recommend that from "single point of failure" standpoint or a from a latency standpoint (24 hops is going to get quite a bit of latency, but yes, it should work.