When I did the Plasma Icicles I ran into all kinds of noise issues, most having to do with bundles and distance.. I tried null pixels several times over.. when I finally went with Pixel Extenders the problems went away.. like right now.. One run is around 60' and I have no issues with it. I am also using the original resistors in the 6804. I am sold completely on Pixel Extenders for anything but short runs. Some have been doing fine without them.. but I was not so lucky
Yea, null pixels are basically an effective hack to get around what's really a transmission line termination issue. The right thing to do is terminate the signal, which sandevices does. The thing is, unless you know the cable someone is going to choose, any particular resistor value is just as likely to be a potential problem as be the correct value. That's not jstjohns' fault, it's just that different cable types have different requirements and he doesn't have control over what cable happened to be laying around the house or what was on special (not that anyone on these forums would ever make a cable choice that way).
An additional consideration on the 6804s is that they don't have a buffer/level shifter to get their signals up to 5V, though they do use a neat diode trick to get to 4V. The lower voltage means a shorter distance before voltage drop becomes an issue. The other sandevices controllers do shift up to 5V, though they also cost more.
Using RS485 is one way to address these issues (I'm assuming that's what the pixel extenders you're using are doing). Resistors are certainly cheaper, though they can still leave you with potential voltage drop issues. Thankfully, voltage drop requires a longer distance to become an issue.
WIth all the posts about pixel problems, seems like we need a good tutorial on transmission line termination to go along with the other pixel info. Fact is, with a 4ns rise time on the propeller chip, there's a 250MHz signal to terminate. Since square waves are really just a summed series of sine waves, the rise/fall time of the transitions determine the need for termination, not the square wave frequency. At a practical level, this means anything over a couple feet can start having issues. Chips with slower rise times will lengthen that initial distance. Even a relatively moderate rise time of 10ns means a 100MHz signal to terminate, which only gets you to ~5ft before the potential for problems increases.
Getting good pixel data transmission isn't about luck. It's about transmitting data with low noise levels. This is mostly a termination issue until you approach 100m w/CAT5. Well, there is the waterproofing, cable strain, getting enough power to the right places, avoiding ground loops, and controlling all those channels too. Maybe there's some luck involved, but the data lines don't need to be part of that.
Perhaps it would be helpful if someone (who really knows the science behind it) were to create some sort of chart in the wiki that shows suggested temination values for different types of wire. I use a 2 pair composite control cable for mine with an 18awg power pair and a 22awg data pair. I know that the impedance is 95 ohms. But a 100 ohm resistor is close enough. I also have one of the data wires bonded to ground at the controller end since I'm using single ended unbalanced data. This solution works very well for me. But it took a bit of research and trial and error to get there. A chart really would have helped a lot.
A few other common wire types I think should be included would be cat5, SPT (as in the 3 or 4 wire type the pixel strings themselves are usually made from), thermostat wire, sprinkler wire, alarm wire (last 3 are very similar, but may have some differences).
--Jon Chuchla--
Sent from my iPhone using Tapatalk
A chart is an excellent idea. Anyone with a scope and a frequency generator can do a good enough job measuring impedance for these purposes. If you're up for some light reading (relative to an academic paper), here's a good page: http://www.vagrearg.org/content/transline
If I could make a reccomendation, I'd stick to cables that have some interference mitigating properties. Twisted pair being the starting point. Thankfully, CAT5 UTP (and really any CATx classified UTP) is required to have a 100ohm impedance if sold with that spec. Coax is often sold with it's impedance as part of the product name (75ohm and 50ohm being common). Most vendors of cable meant to transmit data provide a spec/datasheet w/the impedance documented. IIRC, you actually found a nice cable that's meant to transfer both data and power.
The thing with measuring power cable impedance is that the dielectric used to insulate, the thickness of the insulation, the conductor material and thickness, and more influence the impedance. These factors all vary by manufacturer and product line/quality. It's not really possible to measure SPT once and apply that impedance everywhere. Thermostat/alarm/sprinkler wire might be worth a shot.
Maybe a poll asking what cable types people are using for pixel data is in order.
When I said spt I wasn't thinking the spt1 or spt2 we usually use for power. I'm thinking the 18 or 20 gauge 3 or 4 conductor stuff that they make pixel strings with. Often it's multicolor. But it's now easy to get in white green or black from ray Wu and others. Do you think that stuff is predictable and consistent enough to include in a chart?
--Jon Chuchla--
Sent from my iPhone using Tapatalk
