OK I am sure that Ray will take care of my TC strings, but he has gone silent. I have emailed him several times now with no response! Not that it matters when I get them, I just hate being ignored. It's been 4 days now.
timon wrote:Ray and I suppect the regulator might be an issue so we're going to be running some extensive temperature tests using a probe that will be attached to the regulator and then molded.
I was just curious what type of voltage regulator is it? I thought most voltage regulators just shut down when overheated then reset when cooled? Something like the 78L05 is good for up to 100ma, but with the 3 LEDs it should be drawing less than 60ma. At that rate it would be about (7*0.06=0.42W), which would likely be within the design window. Has anyone seen a circuit diagram they can share?
I have taken some apart. Mine have the 78L05.
From Ray:
".... for the problem technicolor pixel, i am re-designing the new pcb to make improvement:
we found where is the problem. it is the LDO(78l05), amount the bad nodes, almost all of the LDO on the pcb have been burnt,
the current through the pcb is about 70ma,the led only takes around 0.22W(so 3.2V*70ma=0.22W)but the LDO takes 0.35w(5V*70MA=0.35W), so the LDO get lots of heat by the current, and finally burnt if the pixels keep working all the time.
i am re-designing the pcb, to put the LDO on the down side, and make the pcb larger than before,more copper sheet,
and add a resister between the DI and DO,to avoid the signal reflection
and add a resister on the red channel.
also make the pcb with EMC approved.enlarge the soldering pad."
Sounds like you may want to power these through a buck convertor @ 8-9VDC anyway to help with excess power loss through the regulators.
pythan wrote:From Ray:".... for the problem technicolor pixel, i am re-designing the new pcb to make improvement:
we found where is the problem. it is the LDO(78l05), amount the bad nodes, almost all of the LDO on the pcb have been burnt,
the current through the pcb is about 70ma,the led only takes around 0.22W(so 3.2V*70ma=0.22W)but the LDO takes 0.35w(5V*70MA=0.35W), so the LDO get lots of heat by the current, and finally burnt if the pixels keep working all the time.
i am re-designing the pcb, to put the LDO on the down side, and make the pcb larger than before,more copper sheet,
and add a resister between the DI and DO,to avoid the signal reflection
and add a resister on the red channel.
also make the pcb with EMC approved.enlarge the soldering pad."
Not sure you need a regulator feeding a regulator when the design is corrected.. but for now I have been telling everyone to turn the voltage down.
.... turn them all on white full power via software... dial down the supply until you either start to see the last pixels start to dim or change color, or until you hit about 6.5v at the string. Which ever comes first.... then bump it up about .2v just for good measure.
Those regulators for the most part are ok at 6.5v.. but 7 volts is the specific highest required voltage to allow regulation.
I have just received 5 Technicolor pixels from MindTrik, I did quick analysis on them. Here is high level what I found;
#1 - minor water ingress, came in through 12V input wire, actually on the strands of wire, appears to have attacked the via on the PCB feeding data through PCB to WS2811 Data in pin. Probably electrolysis helped. Apply makeshift jumper wire to WS2811 on data in, pixel works fine. Draws 38mA @ 12V while all bits high (full on). 90% duty method.
#2- WS2811 draws much current-about 127mA while idle. 5V regulator is working, puts out 5V during this event, seems ok. Clearly package had been hot, conformal coating was melted over the WS2811 internally. If regulator is regulating, why did WS2811 die? Still investigating...please note, the LED itself works fine, no degradation of color.
#3, #4- Same behavior; WS2811 Data in pin sinks too much current while data is sent. >2mA. Good pixel is <10uA. once I drive data pin with enough signal to overcome sink load, pixel works fine, 38mA @ 12V full on.
#5 - works as expected. I will continue to examine.
General observation;
Pixel design uses tried and true internal regulator, LM7805. data in will need to be between 0V and 5V, +/- 0.3V. pixel draws 3mA idle, as much as 38mA at full on ( I would think at 18.5mA per leg, should be around 56mA). Water proof design is actually quite good, appears to we well sealed. Seems robust in mechanical design. LED is larger package, handles heat well. IMHO, these did not fail due to heat or regulator failure. In all cases, the regulator is in full regulation. One flaw in this design (IMHO) is the Data in threshold. Data sheet states the data in voltage shall not exceed VDD + 0.5V, and since this is powered by LM7805, VDD is 5V. BUT the string voltage is 12V. therefore, if you happen to touch data pin to 12V, you risk damaging the WS2811. I dont know about Technicolor users, but I do this on GECE all the time. However, I use 5V, not 12V.
It should be noted, it appears this pixel will work fine at >7V input. It does not need 12V. The pixel draws 38mA at 7V or 12V (as expected), but power (heat) would be about 40% more at 12V. The LED brightness appears to be exactly the same. If heat is an issue, this would reduce it. Data sheet also says the output voltage is 12V, but the VDD input supply max is 6.0 to 7.0???? not sure what this means....
It should also be noted, if the data in pin is held low, the current draw goes up from 38mA to 55mA. Data low for >50US is considered a "reset" condition. not sure why it soaks so much more current during this period. The LED does not get any brighter during this period...
besides the PCB via failure, the others appear to be electrical overstress on WS2811 package. The two that load the data in pin could be a clue, perhaps signal on that pin was out of spec for some reason. The one drawing 127mA at idle, not certain failure mode.
how does technicolor pixel differ from this one, electrically? http://www.aliexpress.com/store/product/DC12V-input-WS2811-pixel-node-100pcs-a-string-IP68-rated/701799_735789822.html It appears these may be same EE design (similar PCB) ...If they are the same, why does this one not seem to have same issue?
Stayed away from the forum for a day as I was busy getting the display up and running. Still have one set of lights and some cleanup to do. Hopefully I can go line Wednesday night. Anyway here are my comments on the findings.
Steven010 wrote:how does technicolor pixel differ from this one, electrically? http://www.aliexpress.com/store/product/DC12V-input-WS2811-pixel-node-100pcs-a-string-IP68-rated/701799_735789822.html It appears these may be same EE design (similar PCB) ...If they are the same, why does this one not seem to have same issue?
That's the same board current used in the Technicolor's.
Steven010 wrote:I have just received 5 Technicolor pixels from MindTrik, I did quick analysis on them. Here is high level what I found;#1 - minor water ingress, came in through 12V input wire, actually on the strands of wire, appears to have attacked the via on the PCB feeding data through PCB to WS2811 Data in pin. Probably electrolysis helped. Apply makeshift jumper wire to WS2811 on data in, pixel works fine. Draws 38mA @ 12V while all bits high (full on). 90% duty method.
#2- WS2811 draws much current-about 127mA while idle. 5V regulator is working, puts out 5V during this event, seems ok. Clearly package had been hot, conformal coating was melted over the WS2811 internally. If regulator is regulating, why did WS2811 die? Still investigating...please note, the LED itself works fine, no degradation of color.
#3, #4- Same behavior; WS2811 Data in pin sinks too much current while data is sent. >2mA. Good pixel is <10uA. once I drive data pin with enough signal to overcome sink load, pixel works fine, 38mA @ 12V full on.
#5 - works as expected. I will continue to examine.
- Did it look like the water came through on the outside of the insulation or did it somehow get into the wire and come in that way?
- Sounds like the input was damaged. A design change I sent to Ray should fix that problem.
- Data input, see #2. I don't understand the low current either. Did you test a working pixel to see what current level you get. I can try that with some of the PCBs I have here.
- Same as 3
- Did you open this one up?
General observation;Pixel design uses tried and true internal regulator, LM7805. data in will need to be between 0V and 5V, +/- 0.3V. pixel draws 3mA idle, as much as 38mA at full on ( I would think at 18.5mA per leg, should be around 56mA). Water proof design is actually quite good, appears to we well sealed. Seems robust in mechanical design. LED is larger package, handles heat well. IMHO, these did not fail due to heat or regulator failure. In all cases, the regulator is in full regulation. One flaw in this design (IMHO) is the Data in threshold. Data sheet states the data in voltage shall not exceed VDD + 0.5V, and since this is powered by LM7805, VDD is 5V. BUT the string voltage is 12V. therefore, if you happen to touch data pin to 12V, you risk damaging the WS2811. I dont know about Technicolor users, but I do this on GECE all the time. However, I use 5V, not 12V.
This is what I expected if the regulator thermal protection was working. What you found is confirming what I expected was the problem based on the design.
It should be noted, it appears this pixel will work fine at >7V input. It does not need 12V. The pixel draws 38mA at 7V or 12V (as expected), but power (heat) would be about 40% more at 12V. The LED brightness appears to be exactly the same. If heat is an issue, this would reduce it. Data sheet also says the output voltage is 12V, but the VDD input supply max is 6.0 to 7.0???? not sure what this means....
I've found another regulator that should work much better than the 78L05. It has a much higher junction temperature of 150C, lower dropout of about 0.5V instead of 1.7V, lower quiescent current of about 5ua instead of 2ma, a 0.5W rating, thermal shutdown and lastly it looks to be about the same price as the 78L05.
It should also be noted, if the data in pin is held low, the current draw goes up from 38mA to 55mA. Data low for >50US is considered a "reset" condition. not sure why it soaks so much more current during this period. The LED does not get any brighter during this period...
Yep, I really think this is either a EDS or the input got shorted to +V. When the WS2811 is taken out current goes up then the temperature goes up and the PCB starts to scorch. Classic cascade failure.
besides the PCB via failure, the others appear to be electrical overstress on WS2811 package. The two that load the data in pin could be a clue, perhaps signal on that pin was out of spec for some reason. The one drawing 127mA at idle, not certain failure mode.
As you can see there are no input and output resistors on the current board. Those resistors are there for two reasons, first to act as series terminations and second to protect the input to the chip from hot plugging and voltage spikes such as ESD. That is one of the design changes that I had Ray make.
pythan wrote:From Ray:".... for the problem technicolor pixel, i am re-designing the new pcb to make improvement:
we found where is the problem. it is the LDO(78l05), amount the bad nodes, almost all of the LDO on the pcb have been burnt,
the current through the pcb is about 70ma,the led only takes around 0.22W(so 3.2V*70ma=0.22W)but the LDO takes 0.35w(5V*70MA=0.35W), so the LDO get lots of heat by the current, and finally burnt if the pixels keep working all the time.
i am re-designing the pcb, to put the LDO on the down side, and make the pcb larger than before,more copper sheet,
and add a resister between the DI and DO,to avoid the signal reflection
and add a resister on the red channel.
also make the pcb with EMC approved.enlarge the soldering pad."
Yep, those are the changes that I had Ray make.
I'm talking with Ray right now. He's sending me the PCB files to check. Will report back later.
John
From the info above I have a possible idea as to why the WS2811 itself is failing with the regulator still in tact after the fact.. If the 78L05 over heats for some reason (as we think it is) and shuts down if you keep feeding it data, you will exceed the VDD +.5 threshold. (I assume that .5v is to keep an internal junction from exceeding the .6v or .7v threshold). Perhaps that might be something to look at?
Craig
