DMX Lighting?

U

unclebob1

Joined
16 Aug 2012
Messages
642
Reaction score
20
Location
London
Country
United Kingdom
Can anyone recommend me some hardware for DMX to LED lighting?
At the moment i have the following idea:

Raspberrypi sends out the DMX signals - usb-dmx converter
DMX reciever controls that chain of LEDS.

Fairly simple at this point...

The questions are:
If i use ws2801 leds (individually addressable) and use a receiver which translates as needed to the ws2801 string, how many led's can i control individually? I'm seeing a floating number of 170, ie a third of 512 less a few for other parts of the protocol but is that for RGB leds rather then the ws2801 protocols for individually addressable?


2) if the max number of LED's i can have per receiver is 170, i'd need something around 18 controllers for different areas/same areas but extending the string. Am i right in thinking that these will be different universes?

3) Anyone know the distance i can have between the controller and the LED strip itself? Ie does the strip more or less need to be connected to the controller or can i have say 2/3m of decent cable, then soldered to the LED strip?

Lastly is anyone able to provide some names/links to decent hardware that i can use to do this?
 
WS2801 LEDs have an SPI interface. They can be driven directly from the Raspberrypi, or more likely via a simple voltage buffer.
No need for USB, converting to DMX or whatever else.

If you are intending to drive ~3000 LEDs (170x18) then the main problem will be the power supply.
 
Thanks for the response Flameport, i was initially looking at using the pi directly, however i cant seem to find much info on running lots of longer runs from them, a fair few posts about the longer strings not functioning towards the 5m mark onwards, use of spi buffers, etc.

The 18 areas was based on max universe size of 170 leds, realistically there will be about 8 areas - through the longest will be approx 25m.

Power wise, im thinking of a large psu, common ground and power the strips at 3/4m intervals, or a couple of psu's doing the same for the longest run, then a number of smaller psu's hidden under floors, in cupboards etc for the other lengths, again powered every couple of meters.

Again the distance between the pi and the first chip also raised a few issues in my mind as people had problems with flickering leds, part lit-part not lit chains etc.
 
If you really need individual control of RGB on 3000 LEDs (so 9000 control channels) then I'm not sure that a Raspberry Pi is going to be up to the job. If you're really in to your hardware you could maybe make it work by mapping the DMX outputs to screen pixels (seem to remember doing something similar with a BBC Model B back in a different life) and making your control program (whatever that might be) poke the required values into the screen pixels. 50 times a second. Minimum.

DMX is a serial protocol, think max length between repeaters (most high-end commercial devices have repeater circuitry in them) is 100m ish but check in Wikipedia for details. Hardware- Artistic Licence used to make useful chips and things but none of it is cheap, it is all designed to work all the time (cancelled shows due to failed controllers are expensive).
 
I want to be able to blend colours from one side to the other, hence the request for individual control, low frequency refresh. Overall distance between the controller and receiver would be about 10m, with possibly another 5m between the receiver and the first led... or from the pi to the first led maybe 15m...

Flameport have you successfully used long strips of ws2801 on a pi?
 
The description of the WS2801 is misleading. The individual chips are not addressable and cannot be selectively written to unless each has it's own output from the controller.

The data packet required to set the colour of a single pixel ( one WS2801 chip and three LEDs ) is 24 bits of data. To control a string of 100 pixels would require 2400 bits of data sent in a sequence. Even to change just one pixel in a string the entire string has to be re-written.
ws2801_packet.jpg


To control a panel of 100 by 100 pixels would require 240,000 bits of data to update the panel. Best achieved as 100 strings of 100 pixels and not as 1 string of 10,000 pixels if the time to update the panel is to be reasonably short.
 
ok so if i split the strings up, i will have even more issues with distance, as if the pi is 15m away from the 'furtherest' string, will it still work? or alternativly is there a different set of addressable LED's that i can use? As yet, nothing is purchased, but the builder will want to run cables this week!
 
The serial data and clock for each string could be cariied over CAT 5 cable, one pair for data and one pair for the clock with converter chips ( SN75176 ) to convert the differential signals to single ended. ( wasting half the 75176, othe chips are available ).
 
hi bernard, do you have circuit diagram i could follow? i think i know what your going with - ive read a thread on another forum with the same chip, but want to make sure its the same...
 
Given the price of the Pi's themselves you might want to use multiple Pis (not sure how to write the plural of Pi)- and link them with Cat 5. Quick look at the spec on the suppliers controller shows max length from controller to first LED should be 8 metres.

You're going to need some fairly hefty power cables as well- each strip of 32 wants 2 amps at 5 volts
 
Using multiple pi's is not a problem, just the ability to position them!
Power - most likely use 2*2.5mm twin and earth (normal house cable) with flex on the end to connect to the strip, powered from a few psu's.

so:
Adaptor
2.5mm 2.5mm
flex flex
LED/LED/LED/LED

ie 2.5 to each end.
 
Last edited:
sn75176 data sheet http://www.ti.com/lit/ds/symlink/sn75176a.pdf

I would suggest you start with Pi or similar with a single strip of a few pixels ( WS2801 and 3 LEDs ) and experiment to find out all that is involved in changing the colours of the pixels. For one thing you will have to compensate for the differnt brightness levels of the red green and blue LED elements. In theory one would think putting the same value to each of the three LEDs would give white light ( all three same brightness ). In practise doesn't work like that.

Only by experimenting can you learn enough to be able to create a design for your own hardware for the system you want to create.
 
This is going to be one of those schemes where the visible bit (the LED strips) cost nowt but by the time you've got the cabling and control systems and power supplies in you've suddenly spent a fortune. Looks as if you can daisychain max 2 strips (for power) so you may end up with something like Big Fat 20A PSU down 2.5 T & E and 1mm flex drops off that 2.5 to each pair of strips. Which gets you into fusing territory- you'd really need to fuse each 1mm flex at 5A or so, otherwise a short on one drop could smoulder away for ever.

Sounds as if the most complex bit of the scheme will be arranging things so you can site controllers etc. inconspicuously and accessibly near the strips.
 
oldbutnotdead... thats exactly what this has turned into!!!

If i use nonaddressable - so the led's are the same, can anyone recommend any solutions which wont be as complicated?

So the area's to light are:

(non addressable) Perimeter of the room at ceiling level (approx 27m)

(addressable/non addressable) two skylights - 2mx1m each

(addressable) 1 kitchen island - 2.25m*0.9m - either at plinth or under worktop level

(addressable) Length of kitchen cupboards - approx 4m either at plinth or under worktop level.

(addressable) Under glass splashback - approx 3m

The island, kitchen and splashback i can hide the PSU and PI easily under plinth/in cupboards, its teh perimeter and skylights where i have the difficulty!
 
Last edited:
Addressable or not you've still got the same problems- you need chunky power cables and data cables to each strip. It appears from the spec on those units that you can daisychain the data for ever (since the signal is regenerated within each LED- cost will be relatively slow update because of the word length) so I suspect whatever you do you'll be looking at access panels around the room at high level and near each skylight.

So to go with your original plan you'll need something like;
CEILING- a Pi in a corner (any plans to hang pictures permanently- I've got a full-length mirror in the downstairs khazi which is a brilliant access panel for the pipework to the shower in the next room) with CAT5 back to central location (think you can run the Pi POE, if you can it'll save a cable but CHECK).

POWER- this is the nasty one. There's a low voltage cable calculator on TLC's site- it only does 12v but it tells me that to get 250w (in that case that only comes to 20-odd amps) 5m from transformer to lamp needs 10mm cable. I don't see the situation being any better for 5v (in fact it'll be worse- in your scenario at the same current you'd only get 100w). 10mm cable is going to be stupid expensive,to do the job within any sort of reasonable budget you're going to need mains every 4 metres (in access panels) and then 2 x 5A PSUs (or 1 x 10A) to drive 4 sections (remember you can link power backwards round the loop as well as forwards).

Going to be expensive and messy. If you have stud walls or drylined or whatever then you're in with a chance of sinking the control gear in the wall, if you're on solid walls then you'll be a bit stuffed. How is access from above the ceiling- you'd need access traps in the floor at every point where you have a power supply, easiest things would be the metal boxes usually used in office raised floor installs but they ain't cheap either. As long as your PSU is within a metre of each strip end you're feeding then 1mm flex will do the job (delivery 8 amps to 4 strips). At 2 metres you need 2.5 flex

You sure you still want to do this........
 
You must log in or register to reply here.
Back
Top