The picture below is two LED strip (WS2815B) controllers. I have 8 of these boxes, 4 of them are plugged into an outlet of their own, 4 of them are in two pairs like this one and are plugged into the same outlet. This used to have a single cord and plug to the 120V wall socket, I changed it to 2, hoping to fix a problem, but it did not.

The problem is that one of these two boxes causes the strip to flash at irregular intervals on the 1 second scale, blinking only for an instant. I appears that a single update of the strip is being screwed up.

Some background: These are IOT boards of my design that run patterns on the LED strips. They update 50 times per second. The WS2812 uses pulse width encoded digital data, 24 bits per LED with a shift register effect to individually set each LED in the strip on each update. The update takes about 10 ms and is done by software bit twiddling, since the pulses are very fast (350 to 650 nsec). The processor is running at 32 MHz.

I think I have eliminated everything except interference from one power supply to the other. Here is what I have tried.

1. Swap the two circuit boards. No change. Problem does not move with the board.
2. Swap the two strips. No change. Problem does not move with the strip.
3. Replaced power supply on strip with the problem. No change.
4. Replaced the JST connector that I constructed with commercially wired ones. No change.

And this worked:

6. Plug one of the supplies into a separate outlet through a 25 foot extension cord. The two outlets are on the same breaker.

And this did not:

7. Plug the same 25 foot extension into the same outlet as the other box.

So, I can think of nothing else except that one of the power supplies is inserting noise into the AC line that somehow interferes with the other box.

When I look at the data stream to the flaky strip, I see many updates looking correct, but an occasional one either starting late or ending early, which I presume causing is the blinking. Bizarrely, when it starts late, one the scope, it looks like the amplitude of the pulses starts low and builds to the correct amplitude, which I am at a loss to explain since it is a digital output from a PIC microcontroller.

So, finally, my questions:

Does my interference noise theory sound correct?
Is there something else that might be causing this?

If it is noise injected onto the AC line, what can I do to suppress it?

Info that I don't think matters:

Board uses a PIC24EP256MC204 chip and and ESP01 (ESP82666) WiFi / internet board.

I see a lot of noise at exactly 500 KHz on the 12V supply to the strip. As much as 600 mV, which seems impossible to me. Presumably this is the switching frequency of the PS.

 

For interference traveling in the power line feeds there is a solution to the problem available to purchase at many electronics distributors, but not at any consumer electronics seller. Line filter chokes that place an inductance in series with both sides of the line. You can see examples of them in many old PC power supplies, if you trace out the mains input circuit. In addition, because this is a very common requirement, there is a whole lot of information available.
AND, congratulations on the success of your diagnostic process!!

 

Thanks, but it turns out the diagnosis was wrong after all. The blinking still occurs with only one of them plugged in! I cant imagine why it went away with one plugged into a different outlet.

Now I am on the cabling. It is the only thing I have not replaced through all of this.

The cable from the board to the strip is the longest one. Perhaps the length (about 18 inches) is the problem.

 

Interaction between high frequency signals in wires is certainly happen, and it has been a design consideration for manyyears, as how to avoid it. There is a lot published about that subject, and a whole lot of experience among those who have fought with noise coupling.

 

Bizarrely, when it starts late, one the scope, it looks like the amplitude of the pulses starts low and builds to the correct amplitude, which I am at a loss to explain since it is a digital output from a PIC microcontroller.

What happens to the 3.3V or 5V supply to the PIC when this happens?

I helped a friend fix a data-integrity problem on this type of LED strip - we added a low-side MOSFET driver in the data line to buffer it.

 

Haven’t checked the 3.3V supply, mostly because I would expect the PIC to reset if the. power had a glitch that long. When the signal is lost, it is typically for several ms. Will check it though.

 

One more thing about the cable and signal integrity. Two other controllers run a strip that is split in two with 5 feet of the same cable between them without problems.

I also had to fix several cables connectors on severak of the others (there are 8 in total). But in these cases, jiggling tge cable to connector made the bad connection obvious.

 

Pulsing quickly at 1-second intervals sounds like one or more of your power supplies is powering a load in excess of it's designed output. Many modern power supplies interrupt their output (protection mode) and then "hiccup" with a restart of power to check if the excessive load is still connected (and disconnected again as soon as the excess load is sensed). The hiccup is typically on a one-second repeat.

these rgb lights draw a lot more current than most people think.Each chip draws 30mA per color. If you set a 1-meter strip with 80 chips per meter to full brightness, they will draw 7.2A. Check the power level of your power supplies. Also, if your LEDs are only 40 per meter and you connect two strips in series, you'll have the same 7.2A load.

finally, most manufacturers of these LED strips (at least the competent ones) recommend putting a 1000uF capacitor at each meter of lights - this is especially helpful (detrimental) if you go from full off (0-amp load) to full on (7.2 amps) because the power supply may glitch and cause your microcontroller to reset. The 1000uF cap will buffer the power demand and help the power supply work through the rapid load change.
Good luck

 

The pulses are not regular, they are at random intervals of about 1 second, but it could be 1/10th of a second or 10 seconds between blinks.

And, I have watched the supply on the scope. Lots of noise at 500KHz, but they never drop out. The PS is 12V because that is what the strips take and the 3.3V for the micro and ESP8266 is a switching regulator module on the main board.

These strips are newer ones (WS2815B) that are 12V and each chip is internally limited to about 10 mA. There are 284 LEDs on the strip, so 2.84A max. The supply is rated fir 5A.

I will take a look at the 3.3V supply.

 

Here is a scope image. The strip is disconnected here.

Yellow is the data line to the strip. The busts of data are spaced at 20ms. Strip is disconnected.

Blue is the 3.3V power.

Probes are both at 10X, and scope is set correctly for 10X.

I cannot believe the noise I am seeing on the 3.3V power. How can the micro run if the power is that noisy? I think I am seeing up to 4V spikes up and down from the 3.3V.

And you can see the missing data in two updates here. It is not this frequent, I had to capture maybe 10 times before catching a missing one.

 

What is the current rating of your power supply? What is the calculated current draw of the strip? What happens when you program the micro to set all the LEDs to 32 out of 255 (as in a 24bit color number 0x404040.

 

stated in post 9, but I will repeat;

PS rated at 5A, max draw of strip 2.85A.

 

I built 8 these boxes. Seven work, 1 does not. I have swapped the board, the power supply and the strip, and it still fails. The failure is intermittent, though it never runs long without blinking. When I scoped the 3.3V the first tine, it stopped blinking After trying some other things, then scoping it again it did blink. Very frustrating.

 

I built 8 these boxes. Seven work, 1 does not. I have swapped the board, the power supply and the strip, and it still fails. The failure is intermittent, though it never runs long without blinking. When I scoped the 3.3V the first tine, it stopped blinking After trying some other things, then scoping it again it did blink. Very frustrating.

It would be interesting to scope the 3.3V lines on some of the other boards and see if the extreme noise spikes are consistent on all of them. I'm wondering if there could be a bad decoupling cap, or bad solder joint, on just that one board. If all of the boards have the same spikes on the 3.3V lines, then that's probably not your problem (although it still might warrant further investigation!)

I am curious about what parts you've swapped while troubleshooting. It sounds like you've swapped basically everything at one point out another (including which LED strip was being driven) and the problem is staying with the box itself? Like, the inert plastic housing that couldn't possibly be the problem? If so, that's incredibly strange!

If it is staying with the box, not with any internal or external components of the circuit, the other question would be location. Does this box, regardless of which components you have in it, stay in the same spot? Could it be getting noisy mains power because of a bad receptacle, or picking up RF interference beyond the levels it's immune to because of where it's located? If you move the bad box to a location where other boxes work consistently, does it work there?

 

I have swapped in a board and strip with ones from a working box, and the working one continues to work and the blinking one continues to blink.

I have swapped in a brand new PS and the blinking continues.

The only things that renain are the plastic enclosure and the cabling. I guess I should make up a new cabling harness and try that. I cannot easily swap them because the cables have to run through holes in the box before the connectors are added. I can tug on and wiggle the connections without affecting the problem one way or the other.

 

There may be a flaw in the cable that is not related to a mechanical intermittent defect. That might include a shorted pair or an accidental connection to a shield, or even a gap in the shield. It might also be an error in a connector, such as an un-seated connector pin, if the connector pins are the push-into-place type. Or a socket that does not grip correctly.
The substitution method is certainly a valid diagnostic method.
Generally, when you have eliminated so many possible guilty portions, the problem is someplace else.

 

Yeah, an intermittent short of the data line to ground would certainly explain the problem. One problem is that I have not mastered making these damn JST connectors. I have about a 50% chance of getting a good crimp each time I try it. This cable, however has my crimped connector replaced with the one that came from the end of the strip, twisted and soldered and covered with heat shrink, so it should be good. I also remade the other end of the data connector, which is a two pin header going to the board, and that did not fix it. I suppose it could be a flaw between two conductors of the four conductor ribbon cable that goes to the strip.

 

I am not familiar with that kind of connector, but I have soldered the MIL style crimp pins very successfully. The benefit being that this makes them re-usable, which is handy when no pins and no crimper are available. But if one remaining unchanged part is a cable that is different it might be a guilty item. Not ope or shorted, but a discontinuity of impedance leading to reflection of the signal at times. (admittedly this is a guess)