PDA

View Full Version : WS2801 adaptor Card



Matt_Edwards
06-26-2011, 01:43 AM
What I like about the Ethernet PropController is the versatility.
How about a daughter board to drive WS2801 pixel strings?
With out any mods, J1 to J4 can run 3 separate string. Considering most strings came in lots of 50, 3 strings would represent 450 DMX channels. Is this sort of limitation unacceptable?


What do other PropController users think? worth while? Or don't bother?


BTW, With a simple mod, the output config could be changed to 4 each strings per port.

pmscientist
06-26-2011, 12:13 PM
A daughter board for pixels would be handy. I think jstjohnz put forward a prototype design in the main PropController thread a while back, which might be worth a look.

A relatively simple board that breaks out the pins in a layout that makes wiring pixels easier and incorporates connections for an external high amperage 5v or 12v source would be good. Since there are already drivers for the 2801, 6803, and 1804 chips. This would be a mostly hardware project.

Another thought might be to add a board with some additional circuitry that allows more control channels. Although with the 4 universe limit, the benefit of this would probably be minimal relative to the time required.

Assuming use of the current hardware as is, if three string sets are to be controlled by each port, they will all have to be of the same type since that requires a shared clock signal. The 180x chips are an exception to this. Since they don't use a clock line, you can get 4 channels of these pixels per port.

The shared clock isn't really a limitation since strings can be daisy chained to form strings arbitrarily long, assuming power is supplied at the correct intervals along the chain. It's really nice to chain the 1804s, since they often come in 20 module strings. Since the PropController can control 4 universes, the theoretical limit is 680 pixels or 2040 DMX channels per string when chaining is utilized. In order to ensure you're able to keep up with the DMX data, one universe, or 170 pixels, per string (per clock signal with 2801s and 6803s) is probably a more practical limit. You could get 4 universes out of a single port if 1804s are used. Of course you'll probably want to save some channels for traditional light strings, maybe some animation control, and other fun things.

Given the three currently supported pixel types, the possible combinations for mixing and matching pixels with the existing ports are:

3x2801
3x6803
1x2801 + 1x6803
1x2801 + 2x1804
1x6803 + 2x1804
4x1804

I can see two options for static boards supporting these combinations. The first is to develop totally fixed configuration boards that address each combination. I think you can reduce that down to 4 boards since the wiring for the 2801 and 6803 is the same, just the signals are different. The second is to incorporate all the necessary port combinations, and enable or disable the desired ports with jumpers/dip switches/etc.

There are some ways to simplify the second option. You know that at least two lines will always be data. One of the lines will be a clock signal in all but the all 180x configuration. So long as you are careful about which lines the drivers are configured to use, you could take advantage of those conditions.

Matt_Edwards
07-01-2011, 08:20 AM
Knocked together a SPI IF daughter board. It has HC541 Buffer and 220R series Resisotrs.
http://baulkolites.com/wp/wp-content/uploads/2011/07/SPI-IF-DaughteRBoard.jpg

are there preferred connectors? I have used MTA 100s.
20x5 mm fuse still seems cheapest option.
has onboard 5 regulator in case light string is 12v type.
If you don't use the reg, wire link is fitted in V1 and V2.

Cheers
Matt

pmscientist
07-14-2011, 11:03 PM
Looks good to me. Probably the best balance between maximizing string types and getting too crazy with configuration options.

What are your thoughts on the power supply? Wouldn't it be easiest to pull the 5V off the pin 7 on the connector?

I went with Tyco headers and Molex sockets for the connections. That was the least expensive option discovered in the thread for jstjohnz' pixel controller. Fewer
connectors here, so cost is less of a consideration, but I went with the same model so all the connectors would use the same process.

Here are the part #s for the ones I used

4 pin:
Tyco - 640456-4
Molex - 22-01-2047

8 pin:
Tyco - 640456-4
Molex - 22-01-2087

Matt_Edwards
07-15-2011, 07:28 AM
I have been toying around with the layout again, this one is a Dual SPI Daughter board.

I don't like the idea of drawing +5v from Pin7 as the tracks are a tad narrow.
10727

n1ist
07-15-2011, 08:19 AM
Would it make sense to rotate the output connectors 180 degrees? Right now, the path from the ground pins back to the input connector snakes back through the pins of the resistor networks. Also, I would add a ceramic cap in parallel with C2. It may not be needed, but I have had regulators go into oscillation without it before. It can always be unpopulated if not needed.

/mike

Matt_Edwards
07-15-2011, 07:31 PM
100nF added as suggested.
The 0V retrun paths quite short on the Top layer. Also Vias either side of X5 shorten length.

budude
07-15-2011, 07:58 PM
How about Y-ing out the trace under D1 to C4 so that it gives you a better ground plane under D1?

pmscientist
07-17-2011, 12:50 PM
I don't like the idea of drawing +5v from Pin7 as the tracks are a tad narrow.10727

I can see that. Personally, it'd be nice to have a jumper so one could choose to use the PropController source for the buffer chips, if desired. I'm not suggesting powering pixels via that, strictly the buffer chips. It would be an easy way to save the cost of the components for those so inclined. The high current connection will still be necessary to power the pixels.

I really like the dual interface design. What do you think about splitting the pixel power bus such that each side could use a different supply? I'm thinking that rotating the pixel headers 90 degrees, and then relocating the inputs to the side would allow for this without enlarging the board any. This would be helpful both for using pixels with different voltages, and for using multiple supplies with pixels of the same voltage.

Matt_Edwards
07-17-2011, 06:18 PM
I can see that. Personally, it'd be nice to have a jumper so one could choose to use the PropController source for the buffer chips, if desired. I'm not suggesting powering pixels via that, strictly the buffer chips. It would be an easy way to save the cost of the components for those so inclined.
Yes, I think you have a valid point here. The PropController could easily supply the buffered.


The high current connection will still be necessary to power the pixels.

I really like the dual interface design. What do you think about splitting the pixel power bus such that each side could use a different supply? I'm thinking that rotating the pixel headers 90 degrees, and then relocating the inputs to the side would allow for this without enlarging the board any. This would be helpful both for using pixels with different voltages, and for using multiple supplies with pixels of the same voltage.
I reckon the 7805 needs to be lost. As long as good 0v connection is maintained between the Pixel Supply and the PropController, things should remain sweet.
I will have a replay with the layout.
Matt

pmscientist
09-14-2011, 01:46 AM
So, I've been avoiding level shifting issues in my own tinkering. This actually got me looking at various buffer/level shifter chips, purely in the interest of not 'wasting' two unused channels.

After reading the datasheets, it looks to me like, technically, the 74HC series reliably sees 3.5v (0.7 * Vcc) as high. I'm wondering, is the 74HC series really ok to use in this application? Would the 74HCT series be considered necessary, or just more reliable for a 3.3 -> 5v application?

In regards to chip options, any thoughts on using the 744050, 74244, 74365 or 4504 (4000 series) chips? The 4504 seems really nice, as it will do both up and down level conversion. The 4050 seems nice, in that it doesn't use OE pins and you don't need to tie unused channels high or low, though I haven't seen an HCT version of that. The 365 looks similar to the 541, just with 6 channels, and the 244 is similar as well, but uses the OE pins differently.

DynamoBen
09-15-2011, 12:12 AM
For what its worth I used a 74HCT on the dimmer daughterboard based on feedback from others.