My last batch of boards from Itead were panelised and arrived scored so I could separate them by snapping them.

This would be perfect for me, but how did you manage to get them to do it?  Their instructions show the exact same rules and and image as seeedstudio that you can only have silk screen dividing the panels.  http://imall.iteadstudio.com/open-pcb/pcb-prototyping/im120418003.html

OSHpark and batchpcb are both extremely overpriced.  If I just ignored the issue I posted about and wasted 75% of the board area on seeedstudio/itead, I'd still get 10 boards from them for less than 3 boards from OSH or batch.

I wonder if anyone can offer a suggestion where I can get some non-square PCBs fabricated.  I need some boards that around 100mm x 20-25mm.  I've used seeedstudio for boards many times and I very much like their service but they seem to be most suited for square shaped boards.  It seems wasteful to pay for 100mm x 100mm board (100x50 is strangely the same price) and then  give them a board that's under a quarter the size.

I've considered panelizing but I don't have a good way to separate the boards.

Thanks for any suggestions.

I used 27K for the ISET resistor. 1K may be too low.

My guess is this is the problem.  The datasheet says it should be at least 9.53k which sets the current at 40mA.  The current drivers are set at 100 times the current through this resistor and 40mA is the limit for the device.   

1 and 7 only use 2-3 segments, any other digit uses more, so it does explain your symtoms too.

Thanks for the answers. I'm using only one led (RGB with for pins) and it have one cathode, thats why there is only one resistor.

Current is constant in a series circuit, so it doesn't matter where the resistor is as long as it's within the same series loop.   Put the 3 resistors between the LED anodes and the arduino pins.

Good luck with your project.

What makes you think your monitor is displaying the "exact" color you think it is?  For one thing, they need to be calibrated http://spyder.datacolor.com/

Besides that monitors have a certain colorspace with better monitors able to show more color gamut.  Viewing angle effects the color (much more variation on TN panels than VA or IPS panels). 

If you show the same picture on 2 different uncalibrated monitors, you'll get very different results.

As others have said background color makes a huge difference in perception.  So do viewing conditions.  When calibrating a monitor it factors in ambient light.  When printing pictures for exhibition, my print software compensates for the ambient light the picture is intended to be viewed in.

How do you intend to generate the colors on the LED, run it at full power and PWM it?  That's relying on the performance of the viewers eye to render the color and you're really displaying nothing remotely like the color on the screen even if you manage to get it to look similar.

Even calibrating something as simple as your white balance, getting pure white and pure grey is very hard to do.

If you only need 1 LED on at a time and no other IO, you can control 380 LEDs with no additional support hardware using charlieplexing.

You may also consider an arduino mega, 25 pins will let you control 600 LEDs.  Keep in mind if you need multiple LEDs on, duty cycle will be a major issue.

For example, use a wallwart 5V power supply, delete the 7805 regulator, add IO connectors that are more secure than plug-in headers, etc.

I would be careful with that advice.  Modern switching wall warts should be fine, but if he gets a cheap, old linear based one, it could easily give 8V no load and fry his chips (yes, I made that mistake in my pic days).

Why not a solder pad on board and solder the bolt-head to the board.

I should test this before rejecting the idea, but my understanding is solder won't stick to galvanized steel or whatever they make the bolts out of.  Any idea if that's true?  Maybe if I find some brass bolts?

Thanks for the suggestions anyway.  At least you won't have my issue as badly because those dips are easy to drop into a working project with a socket or breadboard.  This is my first scratch-built SMD project, I never realized what a pain it would be to debug.

I ordered a 5 pack of soic->dip adapter boards from ebay, so I can drop one of the 7219's into my breadboard.  The dip device I have is from maxim's sample program; can't get more reliable than that. 

It's going to take 3 weeks to get the adapters, so now I'm debating also ordering some more 7219's from a different supplier so at least I'll have them sooner.  I love what the internet has done for this hobby in terms of price and selection, but hate what it's done in terms of slowing down the process.

The code is working fine with a dip-based setup on breadboard.  Both the sample that came with the ledcontrol library and my own code.  Even using the same arduino without reprogramming, just unplugging the boradboard and connecting the pcb.

Okay, so given that things appear wired correctly:
What value is Rset?
Is Vcc at 5V?
Is the matrix installed upside down?

Rset is 18k.   The device package says "1802", and on the two boards I assembled, I measure 18.1 and 18.8k.  The 18.8 surprises me because they are supposed to be 1% precision resistors, but should still work fine.

I do have 5V on Vcc.

The first time I mounted the matrix to the wrong side of the board (actually the side I intended but I did get the header rows backwards on the PCB), the second time I installed on the opposite (electrically correct) side and I just double checked it's in the right way around.

I'm seeing pins connected together here.
Can you confirm with the art work if these are connected?

Code:

MATRIX1  16       8        8          1
         IC1      2        DIG0       1
MATRIX13 1        8        8          1
         IC1      21       SEGE       1

I am not totally clear what this means.  I can confirm (on the printed board with a meter as well as looking at the board file in eagle) that IC1 pin 2 connects to pin 8 on the correct jumper to it's in the right spot for my matrix.  IC1 pin 21 connects to pin 8 on the other jumper, directly across the board from the first one, again correct for my matrix.   And pin's 2 and 21 are not connected to each other.

Edit:  I understand now what you were saying that it's the same pin number on the jumper/header row so they'd be connected together.  It's a different header on the board and the right pin on the IC does go to the right header and they're not connected together.  I don't know why the netlist looks like that, is it the first column (1 vs 16) that defines them as separate devices?

Sorry for the delay, my schematic from eagle is:



and my netlist is http://www.profquotes.com/misc_images/j64.net

The schematic is a very basic implementation right off the datasheet, the jumpers on the left are for signal in from the MCU and out to the next identical board to have several matrices.   The jumpers on the right go out to the matrix. 

I've gone over my eagle schematic very carefully and checked every connection on my pcb with a multimeter and couldn't find any problems.  I also assembled a second board from scratch just in case it was an assembly defect and I have the exact same result.  This is a very frustrating problem.

At this point, I want to say the only possible problem is the max7219's are defective.  I bought them on eBay and paid $7 for a 10 pack which is I realize is very cheap.  So I'm wondering if I should try and buy one from a more reputable source, but I'm not sure where to try.  I also am considering buying some SOIC to DIP pcb's to try dropping one of the chips into a working circuit on a breadboard.  Is there no better way to see if the SOIC IC is any good?

Code:

if ((millis() & 0x1ff) == 0) {
  Serial.println(data);
}

If your loop takes more than half a millisecond or so, won't you miss some of the times the last 9 bits of millis() are 0?  It seems you'll have your println at some pseudo-random multiple of half a second.