Impedance matching 4ft of 20awg wire

They are there to match the characteristic impedance of the cable to the load.

Without them reflections occur.
This distorts the signal.

With a square wave that appears as ringing which can cause clocking errors.

Eg a video cable with an impedence of 50 ohms should be terminated with 50 ohms.

Likewise an audio cable is normally terminated with 600 ohms.

Pixel to pixel distance is so short to make them unnecessary between pixels.

I should add that it is essential to avoid ground loops for this to work properly.

Qdeathstar:
I think disconnecting the data and clock lines will introduce noise that will change things from the get-go.

I also have breaks in the line about 1/2m in length every 10m or so... Maybe I should put a resistor there (though it would be difficult)

Unplugging the arduino psu would be just as good i think to avoid the soldering iron.

Your earlier offer of a video may be helpful, if only to see the physical layout.
At 50M length a circuit diagram does not always tell the whole story.

Ok, I will take a video tonight and YouTube it

As promised, video. This is a ten second clip of one of the blinks. I recorded video for about 2.5hrs, and got three blinks, but this is one. (they all look roughly the same). I have it play twice, once it real time, and one slowed down 20x. You can see the blink "start" at the end of the strip, and work its way back toward the middle.

I also got a rough schematic. The red arrows represent data and clock flow, i've labeled the MCU's location, and the blue dots represent power injection points. I have those protected with ptc fuses and filter noise out with caps.

http://s22.postimg.org/7av1gye4x/Image2.png

hopefully a video is worth infinate words.

Qdeathstar:
You can see the blink "start" at the end of the strip, and work its way back toward the middle.

I think that means that you can rule out data and clock line problems.
Data/clock signals can only move forwards.

I would think it's a power problem. Spikes from long supply lines. Too fast for normal electrolytic caps.
I would start with a 5volt TVS diode across the supply at the very end of the strip.
And if problems persist, some more at various other easily accessible points on the strip.
Leo..

thanks for your response, the supply voltage is 12v, so should I get a 12v one?

I also see they can be bidirectional or unidirectional, does it matter which I get?

Yes. 12volt TVS for a 12volt strip.
Uni-directional. Cathode (ring) to +12volt.

e.g this one.
http://nz.rs-online.com/web/p/tvs-diodes/6295254/
Leo..

When you said blink i assumed you meant off.

Were the leds at full brightness when that happened ?

The fact that all the leds kept the same colour makes me suspect that was caused by a sudden burst of power.

EDIT

I suspect the percieved traveling of the lights is an artefact of the video recording.

the leds were not at full brightness. they were at 10 percent. I think they probably get to full brightness during the flash. what do you mean by “artifact”. what do you think it actually looks like? even at full speed you can kinda see the last five or so pixes go wonkers, followed my more than half…

thanks

Video is scanned from the top down so if at the time of the flash the scan is half way down that frame does not 'see, it, the next frame picks it up however.

This gives the perceived effect of movement where there is none.

A similar faster effect occurs in the horizontal scan direction.

Stop motion, (frame by frame) is needed to be sure, that event is just too fast.

Does your power supply allow the output to be floating, that is not connected or referenced to mains earth.

Lab supplies normally do this but i am not sure about server PSU's

EDIT

Trying to video with the camera on its side may give more insight.

Boardburner2:
Video is scanned from the top down so if at the time of the flash the scan is half way down that frame does not 'see, it, the next frame picks it up however.

This gives the perceived effect of movement where there is none.

A similar faster effect occurs in the horizontal scan direction.

Stop motion, (frame by frame) is needed to be sure, that event is just too fast.

Does your power supply allow the output to be floating, that is not connected or referenced to mains earth.

Lab supplies normally do this but i am not sure about server PSU's

EDIT

Trying to video with the camera on its side may give more insight.

hmm. interesting. I will try to record again later tonight. I recorded at 60fps, and the blink seems to take about half a second to propogate. Maybe I'll try recording at 120fps. And sideways.

I am not sure about the power supply.

Here is the link:

Qdeathstar:
hmm. interesting. I will try to record again later tonight. I recorded at 60fps, and the blink seems to take about half a second to propogate.

My suspicion is that whole blink lasts only a few milliseconds.

I suspect that psu is mains grounded.

Do you have a mains isolating transformer you can borrow.
Big yellow thing often found on building sites.

as an afterthought , do you know what your electric utility supply type is ?

TN-S ? TN-C ? , something else

Looks like your setup is very much an antenna, you will need to break it up into shorter sections separately
driven to avoid susceptibility to airborne transients. Possible some high-speed logic opto couplers (not
ordinary ones, way too slow) inbetween separate sections to pass on the signals?

Am I right in assuming the strips are not near any other metal parts of the building?

Basically you are hoping to run effectively a single unshielded 5V logic circuit spread over 50m up on
a roof, which is pretty crazy - no EE would ever imagine that could work reliably outside of a faraday cage
chamber. Its going to behave like an ad-hoc short wave antenna and pick up who knows what.

Boardburner2:
as an afterthought , do you know what your electric utility supply type is ?

TN-S ? TN-C ? , something else

Ehhh.. It's 120/240 single phase. Dominion Power is the power company. How would I find that information, and how would I go about getting it.

Mark the strips are ran inside of aluminum Chanel that is glued with double sided 3m tape to metal flashing.

Boardburner2:
My suspicion is that whole blink lasts only a few milliseconds.

I suspect that psu is mains grounded.

Do you have a mains isolating transformer you can borrow.
Big yellow thing often found on building sites.

I know it's not recommended but if I cut the ground lead off the transformer would that do anything to isolate the ground? I have the transformer plugged into a power conditioner.

MarkT:
chamber. Its going to behave like an ad-hoc short wave antenna and pick up who knows what.

The flash on the video implies high power.

Unless someone is playing with nukes and i would think someone would have noticed, i suspect some sort of earth fault.

Not sure what though.

What can I try? I'm going to give the tvs diodes a try first, but always like more options

it seems you were right about the direction issue, Boardburner. I am video taping horizontal, and the blink happend simultaneously.

I had I zoomed in the corner expecting a certain result but did not get it. I am now zooming out to see how far the blink goes. I'll post an updated video.

the thing I shiuld mention again, is that I have it on only 10 percent of full brightness during the video, and when the blink happens, it seems like it goes to full brightness, but not brighter than that. Ofc, it happens so fast it's hard to be certain.

MarkT:
Basically you are hoping to run effectively a single unshielded 5V logic circuit spread over 50m up on

I am starting to think this may be down to ground bounce type effect somewhere.

Either in the strip or psu.