Splitting Data Signal of WS2811 into Many Outputs

Status
Not open for further replies.

EvilGenius

Member
Hello
I am working on a new LED lighting project that will utilize multiple 5V Intelligent (WS2811 Based) RGB Pixel Strings.
I would like to build an animated tree where one SPI signal comes in and then splits out to many synchronized signal output each attached to a smart string of pixels. I know from experimenting I can branch out 2 from Data Output of the WS2811 into 2 WS2811 devices and continue with two strings. But that is not enough for what I have in mind. I need the capability of outputting 16, WS2811 with their own attached string.
Imagine a tree with a big trunk that splits into 16 main branches at the base.
Datasheet for WS2811 is attached for your references.
General concept for the device: A) Receive the data in the SPI format from the previous WS2811 pixel. B) Amplify the signal C) Split the signal into 16 Outputs D) Amplify the individual outputs
From what I could tell in the datasheet, I need the internal devices to be fast speed (at least 1Mhz) to keep up with the data transitions. Input current to the device (from previous pixel) at max is about 10ma, and the voltage level input to the device is low at 0.3Vdd (range between 1.8v to 2.1v) and the high at 0.7Vdd (range between 4.2v to 4.9V)
I was thinking of using bunch of 5V Compactor amplifiers (op-amps). One as master amplifier to boost the signal, and 16 more to isolate, amplify and output the data for the pixel strings. All op-amps to operate as non-inverting (data connected to +V). The negative leg of op-amps to have a reference voltage of 2.5V accomplished by a simple 10K voltage divider.
Here are the questions:
1- Will this work?
2- What op-amp do you suggest since I need 1Mhz switching and single source 5V? Can the master op-amp handle 16 outputs?
3- Since op-amps will share the supply voltage with 5V pixels, should I worry about the switching of pixels effecting the op-amp operation? Decoupling capacitor?
4- I am open to suggestion for any ideas specially the ones that are low cost!?
5- I have searched the net and could not find anything useful. Any source for additional info and ideas are very much welcomed.

Regards,
Rom

Attachments

• 315.7 KB Views: 111

EvilGenius

Member
I am still going to work on the op-amp solution but meanwhile I came up with a simple solution using WS2811 itself to split 2, 2, 2.
Signal arrives at the first WS, splits into 2 WS, then splits into 4 WS, then finally splits into 8 WS which provides 16 outputs. I used 15 WS2811 chips with supporting o.1uf caps, and 100 ohm resistors. This allows one data in and 16 parallel output.
PCB footprint: 1.6" x 3.7", with estimated cost of parts at $6.90 plus$2 for pcb for total of \$8.90 for a professional board!

Attachments

• 342.1 KB Views: 196
Last edited:

AnalogKid

Well-Known Member
This is a CMOS part with an input current of only 1 uA. There is nothing in the datasheet that makes me think you can't just tie 16 chip inputs in parallel and drive them with a uC output pin. If you really want isolated data paths, runt he uC output to all of th inputs of two 8-bit buffers or transparent latches, and have each buffer output drive one LED chip. Better than opamps at these data speeds.

ak

EvilGenius

Member
This is a CMOS part with an input current of only 1 uA. There is nothing in the datasheet that makes me think you can't just tie 16 chip inputs in parallel and drive them with a uC output pin. If you really want isolated data paths, runt he uC output to all of th inputs of two 8-bit buffers or transparent latches, and have each buffer output drive one LED chip. Better than opamps at these data speeds.

ak
Thank you for your response. In the past I tried wiring several inputs of pixels to one output of WS and it failed. I am not sure if 1 foot wire per each input was too much or data's integrity suffers as you split them more than twice. It could also be the distance of inputs to output as on a pcb with only an inch or so, the chip's output might be able to accommodate a couple of more inputs.
As for your buffer idea: I like that a lot for couple of reasons. For one the inputs are isolated in case of a short and secondly it compacts the design into a centralized point reducing pcb footprint. I most likely need to consider CMOS TTL buffers which can distinguish between low data and high levels at high speed. I wonder if ULN family will work?

EvilGenius

Member
I looked at the old friend ULN and it is inverting logic and wont work. But looks like SN54AHCT541 buffer will do the trick. It is TTL compatible, has high speed of CMOS (ns), 5v supply in a 20 pin package. I work on the pcb later today.

AnalogKid

Well-Known Member
If you mean the ULN2003 or 2803, those are high current open-collector relay/light bulb drivers, not data buffers. Pick you favorite device family (TTL, LSTTL, HC, AC, ACT, whatever) and look for octal buffer or octal transparent latch.

ak

EvilGenius

Member
Here is the board layout for this device. It consists of a WS2811 to condition the input signal, and two Octal Buffers to create isolation and to split the signal to 16 channels. The board is designed for 12V power supply system with an on-board voltage regulator to convert it to 5V to comform to TTL output. This device can be connected to 16 strings of smart pixels. For example: a large tree with 16 branches expanding radially outward 15 feet in all directions. Happy lighting!

Attachments

• 215.8 KB Views: 109
Last edited:

AnalogKid

Well-Known Member
1. You need a decoupling cap near pin 20 of the left 241.

2. You need a decoupling cap on the 2811 pin 8.

2. If you change to a 541, all of the inputs are on one side and all of the outputs are on the other side. This makes it *much* easier to add output connectors rather than direct-wire to the board.

ak

EvilGenius

Member
The capacitors are there, they just needed to be rearranged. Here you go.

Attachments

• 221.8 KB Views: 99

EvilGenius

Member
Here is another version with all outputs on my side with double sided copper.

Attachments

• 185.7 KB Views: 131

Pommie

Well-Known Member
Can I just point out that the WS2811 is not an SPI device. It uses pulse proportional modulation as the encoding method. Just for anyone who reads this in the future.

Mike.

EvilGenius

Member
Can I just point out that the WS2811 is not an SPI device. It uses pulse proportional modulation as the encoding method. Just for anyone who reads this in the future.

Mike.
Valid point Pommie. With all due respect however, I would like you to google proportional modulation controllers (to purchase a low cost Chinese controller that runs on WS2811) and see how many hits you get and how many of them you can easily buy on Ebay or Amazon! Right or wrong in the LED industry right now, there are two major players for cheap smart controllers. There is DMX and there is another protocol that WS2811, 2812, 2812B use, which goes by the name SPI. It is simply a distinguishing label to separate it from DMX type. We can go against the system by going around correcting all International manufacturers until everyone uses the same uniformity. Until another product comes out that calls its communication protocol an SPI again.

Pommie

Well-Known Member
I just pointed it out in case someone assumed they could use the SPI interface of a µC. Just out of curiosity, how tolerant are they on frequency?

Mike.

EvilGenius

Member
Not very. Their timing is very tight. Please See the datasheet.
That is why I needed a CMOS buffer to accommodate at least 1Mhz switching.

Attachments

• 313.1 KB Views: 64