Still trying to get flicker to go away.. do I have to buy an oscope?

Status
Not open for further replies.
T

tetsuo

Well-known member
Project thread here:
http://forum.pjrc.com/threads/24370...ng-for-a-quot-sanity-check-quot-and-a-answers

I have tried every combination of fixes suggested here on the forums, yet am still plagued with flicker issues. about 1/3 of the strips tend to flicker. I can get WHICH strips flicker to change based on how things are grounded, what resistors I use (if any), how power is distributed (4 teensys, 6 PSUs, 30 strips [120 segments] == lots of options), etc, but I feel like I'm just in over my head and flailing with trial and error.

All I can determine with my DMM is that 6' to 15' of 16AWG wire appears to reduce voltage from 5.15V at the PSU to 3.98V-4.18V at strips, but that any of the strips work OK with that voltage by themselves :-/ Beyond that, I just don't know how to properly diagnose what's wrong. Do i have to invest in a scope to do meaningful diagnosis or is there more testing I should be doing on my own? Changing grounding points or teensy-to-PSU alignment causes changes in behavior, but I don't know how to interpret those changes to effect a solution.
 
Just some general ideas (I've never used LED strips): Have you measured the current a given strip draws (if you've never measured current, the meter goes in series, not parallel like with voltage), and determined whether each PSU can handle five strips? If the strips use a voltage converter, they will draw more current when their supply voltage drops as you've measured (they keep V*I constant, so <V means >I). Have you tried some decoupling capacitors on each strip, in case there are spikes when switching? If the strips pulse-modulate the LEDs in any way, capacitors could help because they would be drawing little spikes of power, and if two strips' spikes were in sync it might draw too much instantaneous current.
 
Thanks for the tips, blargg. I can't seem to get my DMM to provide an amp reading. It was purchased used, so I may be dealing w/ a blown fuse.

Regarding decoupling caps, I presume that would simply be a cap across the +/- of each powered segment of the strip? What kind of farad rating would be appropriate? Each 41-led segment ~should~ be drawing about 2.1amps @ 5v.
 
Sorry for barging in, particularly if I am very wrong or something, but I want to suggest 100uF as a starting point - electrolytic caps right across +/- of each powered segment might be the right medicine indeed.

I should probably try to get a peek at the circuit diagram(s) involved before being too sure tho ;)
 
tetsuo said:
All I can determine with my DMM is that 6' to 15' of 16AWG wire appears to reduce voltage from 5.15V at the PSU to 3.98V-4.18V at strips,
Click to expand...

Oh my, that's most certainly not good. In fact, it's terrible.

I know #16 wire seems large, but truly it's not, relative to the 200+ amps of current your 4500 LEDs require if they're fully on (white).

My test for OctoWS2811 (the one shown in the photos on the website) used 32 pairs of #22 wire. One #16 wire is approx the same as four #22 wires. So for only 1920 LEDs, I used the equivalent of eight #16 wires (8 for ground, 8 for power). My wires were also only about 1 foot long. The longer the wire, the larger the size it needs to be.

You need a LOT more copper between your power supplies and the LED strips. Ideally, you should get those power supplies as close as possible, so the wires are short. Since you're running 2.3X as many LEDs as I did, if you can get the power supplies within about 1 to 2 feet, I'd say use 18 ground and 18 +5V wires, all #16 size.

For the power supplies 6 to 15 feet away, you're going to need an pretty incredible amount of wire to not drop so much voltage.
 
As a design goal, you should see less than 0.1 volt drop between your power supply and the LED strips. Even 0.1V is quite a lot. If you have a multimeter with a peak hold mode, use that. If not, output data that turns on the most LEDs you intend to ever use at once and simply leaves them turned on while you measure voltage.

On my 1920 LED test, I saw approx 50 mV drop from the power supplies to the LED strips. I saw about 0.25 volt drop across the 60 LEDs (1 meter), but just measuring the voltage at both ends of the strip.

If you have very long strips, you'll probably need to supply power to both sides, running very heavy wire to get it to the other side with much drop. The newer WS2812B strips are supposed to conduct better, so this may be less of an issue. But always measure yourself. Remember, these products are made be no-name Chinese merchants competing pretty much only on price. Don't assume the strip is a great conductor for the power. It probably isn't.

If all this seems crazy and extreme, consider 4500 LEDs using 50 mA each is 225 amps of total current. Just ask any licensed electrician what size conductors they would be required for electrical/building codes to install for a service running at 225 amps of current. Large scale LED projects involve pretty incredible currents. If you don't use big enough conductors, you'll get voltage loss that will cause the whole thing to be terribly unreliable.
 
"As a design goal, you should see less than 0.1 volt drop"

Thank you Paul! This is precisely the type of specific guidance I was seeking.

The thing is, I tested the setup including worst-case (longest) wire lengths at the start of this project and found that by following the signal quality best practices I was able to get a reliable flicker-free setup for the first 1600 LEDs. It was only at this point that I bought the rest of the LEDs and things started falling apart :( It's interesting that some of the strips are more content to work at only 4 volts than others (about 1/3 of the LEDs flicker at the moment, the rest are great).

Anyway, I can't move the PSU's any closer to the strips, but I CAN move the teensys closer if absolutely necessary (though this will mean longer USB cables and having teensys installed where I wont have physical access to them). I think I have a solution for the PSU issue.

Here are my constraints:
- The area between the PSUs and the strips is only 1/2" deep, but I have 5.5" of width in which to work.
- the farthest strip segment will be 15'-8" from the PSU
- power is provided at one end of each 41-led segment (I'm using 30 LED/m strips)
- signal runs through 4 41-led segments (164 LEDs total)

Proposed solution:
- run a single pair of 2AWG conductors per PSU/ that will cover most of the length (12' for the most distant set of strips)
- use distribution block to break out the 2AWG into multiple 16AWG near the strips: one pair per 41-led segment
- adjust the PSU so that voltage at each distribution block is precisely 5.0V (SP-320-5 PSU is rated up to 5.5v -> http://www.meanwell.com/search/SP-320/SP-320-spec.pdf)

Here is a partial diagram (sorry - it's the best I could do w/ google docs). This diagram represents what each of the four teensys controls (1312 LEDs). It doesn't show how the teensys appear in the mix, but I just want to get the OK on the power aspect of things.

Am I on the right track here?
LED power solution.jpg
 
Last edited:
Update: the above diagram (with minor improvements) has been implemented and under normal load I now have 5.0V at the strip! Hooray! At max load, it still drops to 4.6V, but that's a massive improvement over what I had before (3.9-4.0 at moderate load, and as low as 3.5 at full-white). Clearly I could still use some more copper if I want to run at full power, but since this is for positional ambient lighting, I'll almost never be above 20%. I plan to put a software cap in at 80% to be well under the rated capacity of the PSU's and conductors.

It's actually much more compact now since there are now six pairs of 2AWG instead of 64 pairs of mixed 16AWG and 18AWG for most of the haul.

Most of the flicker is gone even with the teensies still in the preferred location (many feet from the farthest strip). I think I can clean up what's left my experimenting with grouding and how the teensies and 74HCT245 are powered (e.g. currently teensies are powered off USB but 74HCT245 +5v is currently coming from the same PSUs that are powering the strips).

Thanks again for the advice Paul! Sorry I've been so lax about photos... I WILL start taking some, I promise!
 
Status
Not open for further replies.
Back
Top