PDA

View Full Version : OctoWS2811: Grouping pixel modules to create 'super pixels'



thad.robertson
05-09-2013, 01:34 PM
Hello

I am considering a low-res video wall project that spans a large area with 800 odd pixels in the range of 300mm sq.

Environmental factors dictate that I must have a very bright point source, up to 18 5050's per logical pixel, therefor I would like to group several WS2811 modules together to form one logical pixel.

I can envisage a couple ways of doing this, and I'm wondering which is best:

1. Run strings in parallel from each output of the Teensy3. If I take this approach, how many strings can be safely run from each pin?

or

2. Modify the library to create a logical pixel from n actual pixels. Would this have an adverse effect on timing?

tia

Thad

PaulStoffregen
05-09-2013, 05:22 PM
I am considering a low-res video wall project that spans a large area with 800 odd pixels in the range of 300mm sq.


Why go low-res? Really, why?

Low-res will not save you any significant amount of money. You still need to buy all those LEDs, which are the majority of the cost. You also still need all the power supplies, which are the 2nd most expensive part.

WS2811 LEDs sell for approx $0.25 per LED on Aliexpress these days, at least when you buy a large number. A power supply capable of 1000 LEDs is at least $30 (don't forget extra cooling costs if you get the cheapest types). So for every $270 in LEDs and power, a low-res approach that avoids a $19 Teensy 3.0 for every 1000 LEDs saves you only 7% (even less if you factor in many other costs involved with physically building the system).... and you only get that savings if your alternate solution costs nothing! Alternate approaches very likely involve building stuff that costs nearly as much or even more than just using Teensy 3.0s for high res.

OctoWS2811 makes it easy to just control every single LED, and it costs so very little compared to the LEDs and power supplies. You can use the software as-is, and if your artwork isn't the full resolution (or you intentionally want a pixelated look), just let the software scale it up.



1. Run strings in parallel from each output of the Teensy3. If I take this approach, how many strings can be safely run from each pin?


This can probably work if you use buffer chips, like the 74HCT245. The chips are cheap, so if you build a large enough number, maybe the cost to fab+assemble the boards could be substantially less than $19 each?



2. Modify the library to create a logical pixel from n actual pixels. Would this have an adverse effect on timing?


You could edit the code to receive a small data and expand it to all the LEDs. But that doesn't increase OctoWS2811's capability to drive more LEDs.

thad.robertson
05-09-2013, 06:38 PM
Thanks Paul for the prompt reply...

The low-res aspect is for an art installation; an intentionally pixilated look over a screen size of around 24M x 3M (80pix x 10pix @ 300mm).

The light source has to be bright because, during the day:

the only thing representing black is a dark tint over the diffusion material, and

the installation has to be clearly visible from a reasonable distance.

And the source has to be even enough to rear project the LED's output directly onto the diffusion surface from a fairly close range.

The easiest way to get the pixel density and spread is to group 4 (or 9) 36mm modules together to act as one logical pixel, as shown in the included image (WS2801s in this case), hence the question of the most efficient way control 4 (or 9) pixels as if they were one :)

I like the parallel approach because it provides redundancy in the case of chip failure in a pixel module and also that there is no code modification required. The only advantage of the single string approach is installation becomes slightly easier.

Cheers

Thad

458