RDM repeater

[majid]

Hi

We have a single pixel RGB RDM LED fixture (footprint = 3).
There may be up to 170 fixtures in 1 universe (512/3=170).
As far as I know due to 32 devices limitation of data bus, we have to divide 1 universe into 6 segments.

I am aware of RDM SPLITTER solution to branch 1 data bus into 6 data bus.
But for some projects -due to linear locations of fixtures- it seems better to divide 1 universe in "linear" form rather "branching".

So I decide to design an RDM-REPEATER as a transparent in-line device which can be placed between every 32 fixterures as below :

R: REPEATER
F: FIXTURE

CONTROLLER-----F1-...-F32-----R1-----F33-...-F64-----R2-----F65-...-F96-----R3-----...

I plan to use a MCU and 2 bus drivers and control buses like the attached diagram.

Is that as simple as I thought?
do I overlook something?

I appreciate comments.

[hamish]

Hi Majid

A repeater can be considered an a single port splitter. With any splitter, other responders may be placed on the responder port of the splitter or several splitters all on one bus.

Splitters may also be cascaded, as could your single port splitter or repeater.
Be aware, that the load on the bus (whic is related to the 32 devices) is not the only factor to consider. With each cascaded device, more bit distortion will be introduced, excepting in the case that the downstream signal is regenerated...

If RDM data is received on one uP UART and re-transmitted on another, significant delay will be introduced, making it fairly problematic to start cascadeing devices.

[ericthegeek]

To reiterate what Hamish said: an RDM repeater is basically a single port splitter. It will need to meet all of the requirements for a "Transparent In Line Device" that are defined in the standard.

The timing requirements in the standard (delay, break shortening, etc.) are designed to allow up to 4 cascaded in line devices, but you'd need 5 cascaded repeaters in your topology. You might be able to have 5 if your repeater has timing performance that's much better than what's required in the standard.

I built a 1-in:1-out repeater like what you're describing, it was a fun project.

One nit pick: you'd get 30 fixtures per segment under the conditions you describe, not 32. The controller at the start of the segment and the repeater at the end contribute to the limit. Like this

CONTROLLER-----F1-...-F30-----R1-----F31-...-F60-----R2-----F61-...-F90-----R3-----..

But here's the most important point: There is *no* 32 device limit in RS485. That limit doesn't exist. There's a 32 "unit load" limit, but 1/4, 1/8, even 1/16 unit-load drivers are readily available. With 1/8th unit-load drivers you can hypothetically have 256 devices on one segment.

Make sure you think about fault containment in you design. Wiring them all in an end-to-end daisy chain may make it easier for the installer, but you'll have at least 500 connections along the path. Any single connection going bad can take down a large portion of the chain. Using a single 1:N splitter at the start may give you a more reliable system, and when there is a wiring fault it will only take out a smaller portion of the system.

[majid]

Thanks for comments Hamish and Eric

Originally Posted by hamish

Splitters may also be cascaded, as could your single port splitter or repeater.
Be aware, that the load on the bus (whic is related to the 32 devices) is not the only factor to consider. With each cascaded device, more bit distortion will be introduced, excepting in the case that the downstream signal is regenerated...

Yeah indeed I prefer using 1-1 SPLITTER/REPEATER and not to mess with new design, but our requirements is 1 to 1 compact size outdoor 12-48V DC.
and I could not find a suitable one yet. I considered SWISSON ISP-4R-DC isolated SPLITTER it is compact for 1-4 use and 10-48V but it is wasting unused port in cascade connection.

Originally Posted by ericthegeek

But here's the most important point: There is *no* 32 device limit in RS485. That limit doesn't exist. There's a 32 "unit load" limit, but 1/4, 1/8, even 1/16 unit-load drivers are readily available. With 1/8th unit-load drivers you can hypothetically have 256 devices on one segment.

Well this sweeps away the problem. I am happy to hear this!
I should say 32 limit mentioned in 2.4.1 Line Bias Networks and F.1 Qualification Tests for Command port Transmitter Circuits and F.5 Testing Responder Transmitters oriented me not to cross line of 32 limit in projects and our projects calculations are based on 32 units using 1-4 branch SPLITTER yet.

Our fixture driver IC specs says 1/10 unit load up to 320 devices! I just checked. However it is still ambiguous to me because I am not sure if the controller also is designed with 1/8 or 1/10 unit load driver or not, customers may use any standard controller or in-line devices with 32 unit loads.

Originally Posted by ericthegeek

Make sure you think about fault containment in you design. Wiring them all in an end-to-end daisy chain may make it easier for the installer, but you'll have at least 500 connections along the path. Any single connection going bad can take down a large portion of the chain. Using a single 1:N splitter at the start may give you a more reliable system, and when there is a wiring fault it will only take out a smaller portion of the system.

With awareness of disadvantages of daisy chain connection maybe we will prefer it in some projects. There would be up to 170 connections for 1 universe of RGB fixtures not 500

[sblair]

Keep in mind too that the TDR testing ESTA did as part of qualifying CAT5 cabling years ago shows that there is significant discontinuities caused by each connector, so ultimately this will be a practical limit far before the theoretical driver chip limit.

I proved it in our QC labs at HES many times by connecting (20) 5 foot cables together and you'd get tons of reflections compared to just using a single cable 100 foot cable.

[ericthegeek]

Originally Posted by majid

Yeah indeed I prefer using 1-1 SPLITTER/REPEATER and not to mess with new design, but our requirements is 1 to 1 compact size outdoor 12-48V DC.
and I could not find a suitable one yet.

I believe the one I worked on years ago is still available. I don't like to promote products here so please contact me outside the forum .

Originally Posted by majid

With awareness of disadvantages of daisy chain connection maybe we will prefer it in some projects.

Yes, it's still a good idea to use repeaters. Connecting all 170 of them on a single segment would be a terrible idea. Limiting each segment to about 32 devices is still a good rule of thumb for the reasons that Scott mentioned. But you can go higher under certain conditions (good cable, excellent connectors, etc.).

Originally Posted by majid

There would be up to 170 connections for 1 universe of RGB fixtures not 500

I'm counting individual contacts. 170 devices with 3 wires each (Data+, Data-, and Common) = 510 connections.

[majid]

Originally Posted by sblair

I proved it in our QC labs at HES many times by connecting (20) 5 foot cables together and you'd get tons of reflections compared to just using a single cable 100 foot cable.

Thanks for comments Scott.
According to your test results, our 32 fixtures chain is already fragile more due to 32 sockets connection rather than unit loads.

Originally Posted by ericthegeek

Yes, it's still a good idea to use repeaters. Connecting all 170 of them on a single segment would be a terrible idea. Limiting each segment to about 32 devices is still a good rule of thumb for the reasons that Scott mentioned. But you can go higher under certain conditions (good cable, excellent connectors, etc.).

Exactly, I beleive even a blind repeater (without signal regeneration) would be useful to renew signal and reduce signal reflection which Scott mentioned.

Originally Posted by ericthegeek

I'm counting individual contacts. 170 devices with 3 wires each (Data+, Data-, and Common) = 510 connections.

You are right, I got it now!

Originally Posted by ericthegeek

I believe the one I worked on years ago is still available. I don't like to promote products here so please contact me outside the forum.

Sure, I will contact you for more details soon
Thanks.

[sblair]

Originally Posted by majid

Thanks for comments Scott.
According to your test results, our 32 fixtures chain is already fragile more due to 32 sockets connection rather than unit loads.

Yes, this has always been my belief from past experiences coupled with the TDR tests ESTA did many years ago studying the effects of difference cables types and connections.

Those studies can be found here: https://tsp.esta.org/tsp/working_gro...XoverCat5.html

[majid]

Because of purchase issues and for some other reasons we dismissed purchasing RDM-Repeater from market and decided to make our own RDM Repeater.
I succeeded design isolated 10-48V DC 1to1 RDM-Repeater and have made some sample production, executed "OLA RDM Responder Test" as below setup:

T:Tester
R:RDM Repeater
F:Led Fixture

T---R1---R2---R3---R4---R5---F

Originally Posted by ericthegeek

The timing requirements in the standard (delay, break shortening, etc.) are designed to allow up to 4 cascaded in line devices, but you'd need 5 cascaded repeaters in your topology. You might be able to have 5 if your repeater has timing performance that's much better than what's required in the standard.

my tests pass successfully up to 5 cascaded repeaters , it fails at 6th cascades.
here is the snap shot of signal after 5 cascaded Repeaters in both directions there is approx 8 us delay in both directions:


every thing seems normal yet.
is there any other tests I shall do ?

[ericthegeek]

The OLA tests are very good for testing devices at the byte level, i.e. what's in the packets.

But they are of limited use for testing transparent inline devices like repeaters.

To test a repeater, you mostly need to test *timing*. Short breaks, long breaks, responders that reply a bit too quickly, responders that take a bit too long to reply, responders that don't reply, truncated replies, line-driver enable glitches, discovery collisions, responses that have a 2ms pause in the middle, etc.


Also, you'll probably want to figure out why it's failing at 6 repeaters. With only 8us of delay you are well below the 88us that the standard allows, so it's not a delay problem. Are you introducing bit jitter to the signal?