Project status update Sep 25, 2011:
PC boards are on hand and a group buy is underway.
Rev 1 of operating manual attached
I have just assembled the first prototype of a new version of the E680 pixel controller, somewhat un-cleverly named the E681.
Same 16-string capacity as the E680.
Single PC Board 6.8" x 4.1". Quite a bit larger than the E680.
Designed to mount one (and power supply) or two (external supply) in a standard CG-1500 enclosure. Makes it easy to use the very inexpensive CG-1500 enclosure as a housing for your controllers. There's room for an E681, a power supply, and a small ethernet switch. Or 2 E681s if your power supply is external. Mounting holes mate directly with CG-1500 standoffs.
DC-DC converter for controller power means the controller can operate off of any pixel voltage. No more heat sink! Now you can power the controller from the pixel power supply regardless of the voltage of your pixels. Eliminates the need for a separate supply, although that option is still available.
Regulated +5VDC output for powering a small ethernet switch There is an on-board power connector that can supply 5VDC to a small ethernet switch such as the Trendnet TE100. The switch ports allow easy daisy-chaining of multiple pixel controllers (without having to worry about supplying AC power to the switch's wall transformer).
3.5mm "Euro Style" pluggable screw terminal blocks for pixel connections. Eliminates the crimp pins! Pixel connections attach to pluggable screw terminal blocks.
Upgraded pixel power input connectors, will handle up to 32 amps on each side, or 64 amps total.
Individual mini plug-in automotive-style blade fuses on each pixel string output.
Optional buffer ICs for driving pixel clock and data lines. If not needed, a set of dip jumpers allows for driving pixels directly from the CPU pins. The optional buffer chips allow for a full +5V drive capability when needed. Selectable +3.3V or +5V on *each* buffer chip for compatibility with both 3.3V and 5V pixels.
Pluggable resistor networks in all clock and data lines for enhanced line driving capability. Some pixel types can be driven over longer distances if a series resistor is used in the clock and data lines. By using pluggable resistor networks, it's easy to change these when needed for a particular application.
The following text is from a post on 6/18 where I discuss some of the thought process behind this new design. I'm including it here also for those just joining the thread:
Some miscellaneous rambling. This board isn't intended as a replacement for the E680, rather as an alternative feature set/form factor. There are really no new capabilities here, unless you happen to get into a situation where you need the features of the line driver ICs and/or the series resistors in the clock and data lines.
Basically what I did is to look at other similar boards, and AFAIK there is really just the E16 from MPH and the upcoming PIXAD8 from J1SYS. I'm not ignoring RJs smart strings, it's just that I'm trying to compare apples to apples and that's a whole different beast.
The physical size of the board is pretty much determined by the desire to make it compatible with the CG-1500 enclosure. These are very inexpensive and make a perfect housing for this type of application. The goal was to be able to put a fully self-contained pixel system in a CG-1500, including the controller, a power supply, and a small ethernet switch. The ethernet switch allows what amounts essentially to daisy-chaining, where you run ethernet into one controller, and out to the next controller in line. This simplifies the network wiring and eliminates the need for a centralized ethernet switch with cables branching to every controller.
Another goal was to be able to provide DC power for the ethernet switch from the controller, eliminating the need for the switch's wall transformer, and the need to wire 120VAC to it. Now for design reasons, the switch power output is 5VDC only, and not all small switches will work off of 5V, but some will. Personally I like the Trendnet TE100-S5, 5 ports, and as cheap as $10-$15. So you can have ethernet in, ethernet out, one port for the E681, and still have 2 ports left over if needed to drive other nearby E1.31 controllers.
Back to the CG-1500. In addition to being inexpensive, it's waterproof and has a gasketed door. It also has gasketed cable access ports along the bottom edge. Enough to accommodate a full configuration of 16 pixel string pigtails, ethernet in and out, and power.
One caveat, you may need to provide some type of ventilation for the CG-1500 if the power supply is mounted internally. But, it will make a very nice clean fully self-contained system.
So, after going to the larger board size, there was enough real estate to add some other features. Hence the new connector style, the fuses, the line drivers, and the resistor networks. The line drivers are designed to be as universal as possible. If you don't want/need them, they can be bypassed and you are driving the strings directly from the prop pins, as the e680 does now. If used, each of them can be jumpered for 3.3V or 5V operation. Yes there are some 3.3v only signal-level pixel chips.
Here's the deal with the line drivers. The E680 drives the pixel strings directly from the CPUs I/O pins. The prop I/O pins are very good at this, able to source and sink much more current than any standard line driver chip. The only limitation is that the prop operates at 3.3 volts, so it's outputs are 3.3 volts. Although I have never encountered pixels that couldn't be driven from 3.3 volts, with the E681 you will have the option of adding the external line driver ICs to get full 5 volt outputs if you need them. And this can be done on a cluster by cluster basis.
Likewise the resistor networks. These series resistors are a feature of both the MPH and J1SYS boards. In my testing with the E680 I obtained the best results with no series resistors. But again I wanted to provide the option, and I wanted to go one step further by making them pluggable to allow easy changing of values.
So with the E681 you really do have a lot of flexibility. You can drive with 3.3 volts directly from the CPU, you can use external line drivers from various logic families (LS, HCT for example), and you can select between 3.3V outputs and 5V outputs. You can have any value of series resistor you want/need.
The fuses are present on both the E16 and the J1SYS boards, and are obviously a desirable feature (I won't say how I know that!), but one that there just wasn't room for on the original board. Yes there is some additional cost, but it's not bad. The layout will accommodate either 1-piece fuseholders, or individual fuse clips if you want to shave the cost a bit.
Bottom line, it's another choice if you want/need the feature set and are willing to tolerate the cost. If you don't need these bells and whistles, then the E680 is the obvious choice, less expensive and much more compact. As far as lighting up the pixels, the 2 boards are identical in every way.
I've been asked about the time frame on this project. My first order was for 6 boards. It's a gamble based on how much faith you have in your board design. Order too many, and you will kick yourself if there's a layout error and you have to pitch the lot of them. Order too few and you will kick yourself if the design is good because once you've paid the per-order setup charge, the per-board charge isn't bad.
Well I have found one board layout error. I re-defined my custom component for the WIZnet ethernet module because I wanted to make sure the dimensions were exactly right, since there is some pretty tight spacing on this board due to the weird screw-hole pattern of the CG-1500. The bottom line is that when I did this I used the wrong pads, and the holes aren't big enough for the pins on the WIZNET module. With considerable force I can mount sockets, and plug the wiz module into the sockets. So I'm still able to test the functionality of the board, and so far I've seen no issues in that regard. Other than this issue, one capacitor needs to be moved slightly, and there have been some suggestions for additional silk-screen legends.
Bottom line, unless I discover additional layout-related issues in testing, the needed changes are minor enough that I would feel comfortable ordering a large run of boards next time, skipping an additional prototype stage.
So, if all goes well, I could have boards available in about 4 weeks.