Notifications
Clear all

Pixel Power

35 Posts
10 Users
0 Reactions
9,725 Views
Posts: 1076
 Zeph
(@zeph)
Noble Member
Joined: 15 years ago
pmscientist wrote:
Fact is, matching cable impedance does buy you a tremendous amount.

I wasn't clear enough. I'm not saying that there's no value to matching cable impedance, I'm saying that it's hard to CALCULATE how to match a logic chip and PCB with unknown impedance to a known cable impedance. Therefore you may need to do some empirical testing to find the best match rather than calculating it. A simple in-line resistor of the same resistance as the nominal cable impedance may not match the cable with any arbitrary circuits (driving or receiving) on the other side of that resistor.

With RS-485 you do not try to match any impedance between data and ground or data and power; instead you are matching the impedance BETWEEN differential data lines. That along with analog sending and receiving circuitry designed for this purpose allows thousands of feet of transmission (depending on baud rate). This is just a theoretical examination.

Choosing whether to use an RS-485 pixel extender is a whole 'nother and more practical issue. They are supposed to work well, but the cost is not trivial.

It reduces the problem to one of voltage drop for at least 10's of feet. At that distance, degraded edges won't come in to play unless there's a damaged cable or something. Sure, in order to get 100m you may need to go to a balanced signal like rs485, but why bother if you only need 50ft. Plus, you're almost certainly not going to find a resistor that will give you an exact match after accounting for the other impedances.

Not sure that I'd agree that rs485 is less confusing than a resistor value. Is 'develop an RS485 interface' simpler than 'use an X ohm resistor'?

Sure, if you're doing some truly high speed stuff over long distances, a balaced line solves some problems, but it also adds several components. If you just have an edge problem, why not just use a schmit trigger? Doesn't require the supporting infrastructure of a balanced line, and is totally fine for the data rates of most pixels.

The whole point is to keep it simple. Matching the cable impedance with the proper resistor goes a long way to improving things. It's also dirt cheap and simple. This isn't analog stuff, it's not high speed or bandwidth stuff, why complicate things until it's really necessary?


Reply
Posts: 2229
(@plasmadrive)
Famed Member
Joined: 15 years ago
pmscientist wrote:Unless I'm totally missing something, it's not true that a balanced signal is output for any pixels. It is ONLY true for Renard and DMX. See here: http://doityourselfchristmas.com/forums/showthread.php?26154-SanDevices-E68x-Version-3-Firmware-beta

Besides the missing documentation, the pixels need single ended input. Were a differential signal output for pixels, a converter would be necessary to get back to what the pixels expect. Is there such a converter for sale or recommended by sandevices for use with their controllers?

There are a few methods of terminating a single ended line. The method used by sandevices and many other pixel controllers is typically referred to as serial termination. Terminating at the receiving device may make more sense to you, but that's just not how it works with series termination. You can also terminate in parallel, which has DC and AC forms, and there are variants within there. Getting into the details is a bit off topic, but suffice to say there are benefits and drawbacks to each.

In the case of the 6804, a level shifter is probably needed in addition to impedance matching. The 3.3V signals output from the prop chip are towards the bottom of the range for the TTL logic level 'high' of the expected 5V input signal. Adding much cable between the two can tip that balance relatively easily. Add a 270 ohm resistor in there instead of the correct 100 ohm resistor, and you've got a good chance of problems. Add to that a 5V supply that actually operates at 5.1, and an issue becomes more likely.

There's a good chance you were dealing with a termination resistor issue. However, there's also a good chance it wasn't *just* that. You could easily have had a logic level issue too. The OP was about an E681, which has the level conversion, so I tried to keep my comments related to that.

I don't doubt that a differential pixel extender fixes the problem. OTOH, given that they run ~$15 per end, you're looking at $30 per string to go that way. YMMV, but I'm gunna try a different resistor first.

The balanced output is not for the pixels.. what I said was if you are using WS2811 or WS28512 pixels the "Clock" output, (which is not used for those two types), is the D- for a differential output. Jim has programmed it that way for this very reason. (using Pixel Extenders or some such devices). BTW, the Pixel Extenders to convert back to single ended.

Keeping in mind that the San Devices boards do support DMX output which is differential. So the differential ability is already there. I think that is where your confusion has come in regarding those boards.. and the J1S Pixel extender at $15 each is worth the hassle time for trouble free operation if you have a distance or noise issue like I did and like so many others are experiencing. The resistor change my work for many, but not for everyone.. so I guess it just depends on the time people want to spend chasing noise.

One more thing.. You are right.. I think this has drifted off topic.. Sorry all..


Reply
Posts: 352
(@pmscientist)
Reputable Member
Joined: 16 years ago
pythan wrote:The balanced output is not for the pixels.. what I said was if you are using WS2811 or WS28512 pixels the "Clock" output, (which is not used for those two types), is the D- for a differential output. Jim has programmed it that way for this very reason. (using Pixel Extenders or some such devices). BTW, the Pixel Extenders to convert back to single ended.

Keeping in mind that the San Devices boards do support DMX output which is differential. So the differential ability is already there. I think that is where your confusion has come in regarding those boards.. and the J1S Pixel extender at $15 each is worth the hassle time for trouble free operation if you have a distance or noise issue like I did and like so many others are experiencing. The resistor change my work for many, but not for everyone.. so I guess it just depends on the time people want to spend chasing noise.

One more thing.. You are right.. I think this has drifted off topic.. Sorry all..

Ok, I found the line in the E682 addendum that talks about this.

"This is only used for DMX type outputs and possibly for 3 wire pixels, where a 2 wire receiver is placed at the pixel string."

This reduces the expense to $15, but the right resistor is still cheaper. Differential ability for one type of signal is independant of that support on another type. Just because it's there for DMX doesn't mean it can be assumed to be there for others.


Reply
Posts: 352
(@pmscientist)
Reputable Member
Joined: 16 years ago
Zeph wrote:I wasn't clear enough. I'm not saying that there's no value to matching cable impedance, I'm saying that it's hard to CALCULATE how to match a logic chip and PCB with unknown impedance to a known cable impedance. Therefore you may need to do some empirical testing to find the best match rather than calculating it. A simple in-line resistor of the same resistance as the nominal cable impedance may not match the cable with any arbitrary circuits (driving or receiving) on the other side of that resistor.

With RS-485 you do not try to match any impedance between data and ground or data and power; instead you are matching the impedance BETWEEN differential data lines. That along with analog sending and receiving circuitry designed for this purpose allows thousands of feet of transmission (depending on baud rate). This is just a theoretical examination.

Choosing whether to use an RS-485 pixel extender is a whole 'nother and more practical issue. They are supposed to work well, but the cost is not trivial.

Sure, but you're over complicating the issue to a point way beyond anything relevant to the actual application here. This isn't an analog signal.

What I'm saying is that the value doesn't need to be calculated at all for this application, and in many cases you don't even need to do any trial by error.

You don't need to know a thing about the board, the chips, anything except the cable. Even then, you don't need a resistor that's an exact match. If you don't have cable with a documented or well known impedance, you may need to do your own testing. Even that is something that a chart can go a long way to limiting the need for.

The fact is, for this application, someone without much interest in the details can take a resistor with an approximate match to the cable impedance and improve their signal tranmission substantially. Let's not let the transmission line theory get in the way of people being able to improve their system in as simple and cost effective manner as possible.


Reply
jchuchla
Posts: 8698
(@jchuchla)
Illustrious Member
Joined: 15 years ago

the other thing to keep in mind if using an e682 to drive a balanced signal for pixels, is that it only works for one output per cluster. I looked into this once when I was considering trading in my e681s for e682s. If using balanced to drive pixels, you can only use the first output per cluster. So for my needs it wouldn't have worked and thus I kept my older 681s and just added the j1sys 4ch balanced driver. This wouldn't apply to the 6804 because there is only one output per cluster.


Reply
Page 7 / 7
Share: