Digital Signal Issues - How is the PSU Causing This?

Thread Starter

mriksman

Joined Aug 31, 2010
113
Hi,

I have 17 LED strips, and a 20A, 5V PSU. It has 3 terminals for COM, and 3 for V+. I have run 4 sets of automotive power cable, so my strips are powered by different cable runs. 3 of the cables power 4 LED strips, and 1 powers 5, for a total of 17 LED strips. For simplicity, my diagram shows just 9 LED strips, with each power cable powering 3 LED strips each.

1617167772375.png

I've already discussed using crappy hookup wire can cause signal reflections and mess up my digital signalling. It's why I have put the ESP32 controller close to the first LED.

I have found an issue that I haven't been able to work out. On a scope, if I look between points #1 and #2, I get the expected (albeit messy due to using hookup wire instead of a known impedance data cable)

1617167560424.png
If I measure between #3 and #2, I get the same signal.

However, if I measure between #3 and #4 (note the COM is now from a different terminal on the PSU and a different cable), I get this
1617167685329.png

Considering that #3 and #2 is different than #3 and #4 (where #3 is the data line, and #2 and #4 are COM), the only difference must be caused by a difference in the COM wires #2 and #4. So I put that on the scope.

1617169070936.png

I thought maybe because the COM wire is on different terminals on the PSU. But even when I measure #5 and #6
(power cables are now on the same COM terminal), I get the same messed up signal.

Every other LED strip looks fine. It's only when there is this change/transition from one power COM to another power COM. This distorted signal is causing havoc to my pixels, and only by adding a 2.2kohm resistor in series with the data line can I dampen it out enough to remain stable.

But why is this occurring? It must have something to do with the long power cable length?
 

Thread Starter

mriksman

Joined Aug 31, 2010
113
Do have decoupling caps on the power pin?
I haven’t any decoupling caps anywhere. Where would you put them, and why?

Both COM/GND wires are connected to the same terminal on the PSU, so this noise is being generated along the cable (about 3-6m)?
 

Thread Starter

mriksman

Joined Aug 31, 2010
113
Where would I put this bypass capacitor? At the start of EACH LED strip? Or just the LED strips that are having this issue? Between DATA_IN and GND?

Why is this happening when the signal leaves a LED strip (DATA_OUT) referenced to one ground wire, and enters the next LED strip (DATA_IN) reference to a different ground wire.
 

Deleted member 115935

Joined Dec 31, 1969
0
Scope earth I bet is the main cause, pickup ,
Try to get both the scope earth and the tip within a few mm of each other, no long wires on the earth.
 

MrChips

Joined Oct 2, 2009
30,706
The issue is with the long wires on both the probe tip and the ground clip. Every inch of wire is an antenna picking up unwanted noise.
 

Thread Starter

mriksman

Joined Aug 31, 2010
113
How does this fix my problem?
I know there is a problem (without the scope attached) because on the first LED strip connected after a new + and - lead, the pixels flicker/flash; the data is corrupted.
The signal is regenerated on the previous LED strip (each pixel recreates/buffers the signal and transmits it onwards), referenced to the previous GND wire. When it arrives at the next LED strip, the GND wire is at a very different voltage than the previous GND wire (as shown in my last scope picture).
 

MisterBill2

Joined Jan 23, 2018
18,167
OK, there is a whole lot more to this arrangement than we are told about. If this is a DC supply and there are no electronics creating that signal, why are we seeing a sort of square wave signal??? Or is it a variable duty cycle LED driver supply? Or a failing switch mode supply?? We are not getting a big part of the story just yet.
 

crutschow

Joined Mar 14, 2008
34,280
It's difficult to be dumped in the middle of a problem without knowing all the details (whether you think they are pertinent or not).
Please give us a complete diagram of the the devices and how they are connected.
Otherwise it's a game of 20 questions and guesswork.
 

Ya’akov

Joined Jan 27, 2019
9,068
OK, there is a whole lot more to this arrangement than we are told about. If this is a DC supply and there are no electronics creating that signal, why are we seeing a sort of square wave signal??? Or is it a variable duty cycle LED driver supply? Or a failing switch mode supply?? We are not getting a big part of the story just yet.
From various clues, I surmise these are WS2812 neopixels. The schematic, minimal as it is, shows and ESP32 at the LEDs.
 

MrChips

Joined Oct 2, 2009
30,706
As far as I can see there is an ESP32 sending a PWM signal at 803.2kHz.
High power LEDs are switching at this frequency.

Turn on an AM radio and tune it to this frequency and you will observe the havok you have created on the airwaves.
 

Ya’akov

Joined Jan 27, 2019
9,068
As far as I can see there is an ESP32 sending a PWM signal at 803.2kHz.
In #9:

"The signal is regenerated on the previous LED strip (each pixel recreates/buffers the signal and transmits it onwards), referenced to the previous GND wire. When it arrives at the next LED strip, the GND wire is at a very different voltage than the previous GND wire (as shown in my last scope picture)."

Part of the reason I think these are addressable LEDs.
 

Thread Starter

mriksman

Joined Aug 31, 2010
113
Apologies, I didn’t mention that. But you’ve all guessed correctly. These are addressable LEDs (SK6812; very similar to the WS2812). Indeed, the data signal is an 800kHz signal, generated by the ESP32 microcontroller. Once the signal reaches the first LED on the first strip, the LED recreates/retransmits the entire signal to the next LED.

Apart from that, my diagram, whilst simplified, shows everything that is connected.
 
Top