4
\$\begingroup\$

Been a while since I have been on here but I do not know where else to turn for help! This is more of a hardware issue than anything so i do hope that I am in the correct forums to be asking, if anyone knows of somewhere else I should be looking for help please let me know!

Anyway, this circuit simply drives a 10W RGB COB LED, using an esp8266 as the micro which then controls 3 MOSFETS via PWM that drive each channel of the RGB LED.

LED: (This is not the same manufacture that i am using but specs are the same) http://www.ledguhon.com/u_file/images/14_11_25/57d628dfd6.pdf

The input voltage is 12 - 15V AC as these are powered by low-voltage AC landscape transformers. The input is then ran through a full wave bridge rectifier followed by a 1000uF capacitor in order to produce DC voltage. This is then followed by a buck converter that drops that voltage down to 10V DC in order to drive the LED. 3 Constant current drivers are then used for each channel of the RGB LED in order to produce 350mA for each channel.

Constant Current Drivers: https://datasheet.octopart.com/AL5809-50P1-7-Diodes-Inc.-datasheet-76802127.pdf

The issue I am having is with the Red channel, in my last batch of 50 of these boards I have more than 10 that the red channel has stopped working on. They work at first and then over time some of them the red channel begins to flicker than just shuts down. Oddly enough if i press on the LED it self the red channel will turn back on but again fails as soon as i let go of pressure. ( I have checked all solder connections and everything seems to be just fine, ive even re-soldered some) I am not sure if this issue is due to a bad batch of LEDs or if i have an issue with my PCB design / circuit. Below are my schematics, the resistor R6 is 10 Ohm, 3W. The second buck converter is used to drop the 10V DC to 3.3V for the esp8266. Also A1 - A9 are the constant current drivers. A1-A6 AL5809-50P1-7 & A7-A9 AL5809-150P1-7 (PN's Located in data sheet link above).

LED Power Schematic Controller Schematic Controller Top PCB LED Bottom PCB Bill of Materials

These two PCBs then clip together to form the final assembly, The reason for the two separate boards is the heat generated by the LED needs to be far from the esp8266.

If you need more information or have any questions on the board in order to assess the problem please let me know! I could write an entire book on the function of this board but I am trying to keep it short and sweet.

\$\endgroup\$
  • \$\begingroup\$ I can't see any constant current generator. Where is it? What are A1 to A9? If they're resistors they should have 'R' designations unless you're working in another language. The rotated Q3 and associated earth symbol make the schematic needlessly difficult to read. \$\endgroup\$ – Transistor Jun 27 '18 at 21:27
  • \$\begingroup\$ What is the power rating of R6? \$\endgroup\$ – Ahmed M.Zahran Jun 27 '18 at 22:15
  • \$\begingroup\$ @Transistor A1-A9 are the constant current drivers, here is the link to the datasheet (datasheet.octopart.com/…) A1-A6 are AL5809-50P1-7 & A7-A9 AL5809-150P1-7 . Bringing each bank of 3 constant current drivers to 350mA. I apologize for the messy work of the schematic it was originally made for my eyes only \$\endgroup\$ – Mr. Pivirotto Jun 27 '18 at 22:24
  • \$\begingroup\$ @AhmedM.Zahran R6 is 3W Power rating \$\endgroup\$ – Mr. Pivirotto Jun 27 '18 at 22:27
  • \$\begingroup\$ Sorry Let me replace CC chips with CC in sim. but warning they need 2.5V drop minimum which if you dont get from bridge and LED drop will cause flicker. V IN(min) = V LED_CHAIN + 2.5V \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 16:53
2
\$\begingroup\$

enter image description here

Red forward voltage @ 350 mA = 7.5
R6: 350 mA x 10Ω = 3.5V

Where is the AL5809 going to get its 2.5Vin-out min.?


Max temperature is 85°C.

Red: 350 mA x 7.5V = 2.625 watts
Blue: 350 mA x 11.5V = 4.025 watts
Green: 350 mA x 11.25V = 3.938 watts

Total watts = 10.6 watts

Without substantial thermal management there is no way these LEDs will be running below 85°C.


if i press on the LED it self the red channel will turn back on but again fails as soon as i let go of pressure.

This is a bit of a mystery. The LEDs should be too hot to put your finger on. Unless they are not getting the expected 350 mA. If one or two of the AL5809 fail because they do not have the 2.5V min then the current will be less than 350 mA dropping the voltage across R6 giving the AL5809s their required 2.5V.

Measure the voltage across R6 to get the current.


I would try disconnecting one of the AL5809 and or short out or lower R6 to 2.8Ω for a 1V drop.


Also consider using a different CCR. The On-Semi NSI50150ADT4G has a minimum 0.5Vak and 350 mA max.


UPDATE

In the comments there seems to be some confusion as to why I made the above suggestions.

The function of the CCR is to supply a dynamic current limiting resistance in the same manner as a fixed resistor. A CCR will have a minimum and a maximum anode to cathode voltage (Vak). The actual operating Vak is strictly a function of the supply voltage minus the LED's Vf.

THis confguration is poor due to "LDO dropputs of 2.5V in both Linear and CC regulator . and thus very sensitive to any changes V overhead on NSI50150ADT4G is not 0.5Vak but 1.8 V Typical – Tony Stewart older than dirt

A CCR is a linear CC regulator.
The NSI50150ADT4G's 1.8V "typical" is only specified at a very specific current and thermal conditions. Its purpose is to specify thermal performance not the normal operating Vak as previously described. See Figure 2.

The reason I suggested the NSI50150ADT4G is becasue it is a single part that can replace three. A CCR requires sufficient thermal management for stable operation and a DPAK package is heatsink capable. Also the current is adjustable for 150 - 350 mA and 350 mA is likely too high for thermal reasons.

The purpose of R6 is to alleviate thermal stress on the CCRs (if needed). R6 is not a current limiting resistor. 10Ω is not likely to work well due to the supply voltage likely being insufficient with a minimum Vak of 2.5V, Vf of 7.5V @ 350 mA and supply of 10V. The addition of the voltage across R6 will likely take the CCRs and or LEDs out of their normal operating conditions.

If the Vf is indeed 7.5V (likely less), any voltage drop across R6 would exceed the minimum voltage required for the CCRs and LEDs. When powered on, LED Vf starts out a about 5.5V and R6 zero. The voltage of the CCR is ambiguous as the minimum 2.5 Vak is a recommendation but likely 1.5V which is specified in the electrical characteristics as the minimum Vin.

The recommended minimum VINOUT voltage of 2.5V...
--AL5809 datasheet

As current rises toward 350 mA the voltage across R6 will increase to 3.5V if the current were to unlikely reach 350 mA.
Three red LEDs in series are likely to reach 6+ volts at over 100 mA.
As VR6 approaches 3.5V it will be unlikely the CCRs will have sufficient voltage operate correctly. The solution is to raise supply voltage, or reduce (or eliminate) the voltage across R6 to give the CCRs sufficient voltage overhead to operate correctly.

As R6 is the single component with a predictable voltage, measuring the voltage across R6 will give a fairly accurate indication of the amount of current flowing for troubleshooting purposes.

Vf is low so CC is accurate but too hot and misunderstood’s analysis to reduce R will make R cooler but CC (red) hotter.
– Tony Stewart older than dirt

Vf is ambiguous but the datasheet IV curve indicates it should be about 7.5V @ 350 mA. This a CC source so the value of R6 would have no effect on current (or red's temperature) if CCR's are working correctly.

\$\endgroup\$
  • 1
    \$\begingroup\$ My client used to spend 5 seconds soldering 3mm below base of 5mm LED. Causing same issue of sheared wirebond. I reinforced Mfg specs 3 sec max. Problem fixed. I agree on CC undervoltage causing flicker on BG and commented recently he needs to reconsider a 12V reg. and 2mF cap low ESR \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 19:04
  • \$\begingroup\$ As I have said to you before, I would NEVER use Chinese LEDs. The OP said only the red had an issue. Now that you point it out it is surprising the BG work at all. 12V may not be enough with the Vf of B & G being over 11V @ 350 mA. 14V supply would be required. 11.5 Vf + 2.5V for the regulators. The actual current needs to be measured. I am doubting that the LEDs are getting 350 mA. \$\endgroup\$ – Misunderstood Jun 29 '18 at 19:35
  • \$\begingroup\$ These were not Chinese LEDs, Not only hot Red resistor but flickering (BG colors implied by design) Depends on Transformer R, MOSFET R but 12V works on my smulation with 2mF or you can say 14V with 1mF but not without simulation. with 1050mA load \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 20:03
  • \$\begingroup\$ THis confguration is poor due to "LDO dropputs of 2.5V in both Linear and CC regulator . and thus very sensitive to any changes V overhead on NSI50150ADT4G is not 0.5Vak but 1.8 V Typical \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 20:15
  • 1
    \$\begingroup\$ @TonyStewartolderthandirt Keep the discussion technical. \$\endgroup\$ – W5VO Jun 30 '18 at 1:09
1
\$\begingroup\$

The fact that the LEDs flicker and then quit after some time is a classic indication of a solder joint which is failing due to thermal issues. That pressing on the pcb (and flexing the board) fixes the problem is another classic indicator of a failing solder joint.

I suspect the culprit is R6. It's clear that the voltage drop (3.5 volts) is intended to drop the voltage at the current regulators to match the other two channels. It's also clear from the size of R6 that you realized that it will dissipate 1 watt. What I don't think you've realized is that it's sitting on a hot pcb due to the LED dissipation (about 9 watts) on the other side. I suspect the combination is getting your connections too hot, and one of the connections is opening up.

EDIT - Tony Stewart in comment has (strongly) suggested that the problem is in the LED unit, and on reflection I'm inclined to agree with him.

The test should be pretty straightforward. Solder a test wire to the via where the red LED cathode swaps sides. Fire up the circuit. When the LED goes out, connect the test wire to ground (briefly) with a 100 ohm 1/2 to 1 watt resistor. This will provide an alternate current path of about 60 mA. If the LED turns on, you know it's good, and the problem lies with the R6/current regulators. If the LED stays off, you know that Tony is right and you're cooking your LEDs. Use a thin wire (30 gauge would be good) - the point is to keep from providing much of an alternate cooling path via the wire.

As I say, I suspect he's right. You have no heat sink on your LED, and it's getting hot. As a matter of fact, if the LED stays off in the long run, only to turn on again if you let the unit cool off, you know that just the heat from the blue and green channels is enough to give you problems.

Why are you getting a 20% failure rate. and not 0% or 100%? Cheap vendor with inconsistent process controls.

END EDIT

\$\endgroup\$
  • 1
    \$\begingroup\$ No he’s thermally stressing the Red internal wirebond joint. Major overstress in this design with spec violations \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 27 '18 at 23:20
  • 1
    \$\begingroup\$ Pressing on LED could be LED lead but with redundant leads and micron wirebond , more likely inside thermal shock failure, CTE issues \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 27 '18 at 23:55
  • \$\begingroup\$ @TonyStewartolderthandirt Ignoring the CTE comment... So im F***ed with this design basically, not enough heat dissipation I get that, I'm already limiting (internally with the pwm output) these guys so they only reach about 3/4 their full power potential. Looks like ill be dropping that to around 50% or lower until I redesign, luckily even at 50% they produce quite a bit of light so I can make use of them somewhere... say i separated the led and all its driving components entirely, mounted the led to the aluminum structure of its enclose for heat dissipation. \$\endgroup\$ – Mr. Pivirotto Jun 28 '18 at 0:30
  • \$\begingroup\$ @TonyStewartolderthandirt kept the mico and all the LED driving components completely separate somewhere far from the LED. only running 4 wires back to the LED, for the common anode and RGB channels. Would this solve my issue? \$\endgroup\$ – Mr. Pivirotto Jun 28 '18 at 0:33
  • 1
    \$\begingroup\$ I also tested the theory of connecting the red channel directly to ground through a 100ohm resistor and no it did not fix the issue, so yes i do agree with Tony that heat is the issue. Now I need to fix it.. which is not going to be easy as keeping this design small and compact was very important. \$\endgroup\$ – Mr. Pivirotto Jun 28 '18 at 0:42
0
\$\begingroup\$

Why you have used a current limiting resistor just for the red channel? as stated by kirchhoff's voltage law: for a closed loop series path the algebraic sum of all the voltages around any closed loop in a circuit is equal to zero. This is because a circuit loop is a closed conducting path so no energy is lost.

enter image description here

Photo credits go to this Web Page

I can guess that red channel is not working due to the voltage drop across its branch prevents the voltage on the red diode to be turned ON.

Try removing R6 and short circuit its connection in one of your items and check if that makes a difference.

\$\endgroup\$
  • \$\begingroup\$ I will give it a shot! The reason for the resistor is to bring the voltage down for the red channel to around 6v in hopes of saving some heat dissipation because the red channel is 6-7V while blue/green are 10-11V \$\endgroup\$ – Mr. Pivirotto Jun 27 '18 at 22:53
  • \$\begingroup\$ Is that recommended by the manufacturer? try reducing its resistive value. \$\endgroup\$ – Ahmed M.Zahran Jun 27 '18 at 22:55
  • \$\begingroup\$ There isn't much documentation from the manufacturer, in fact ive had trouble finding any at all, so no it was not recommended. And i just tried replacing that resistor with a straight line, so bypassing it, and it had the same result. the red channel turned on for a bit until the board warmed up and now its out again. but this may be due to the LED now being faulty? \$\endgroup\$ – Mr. Pivirotto Jun 27 '18 at 23:01
  • \$\begingroup\$ You have mentioned that this happens in this batch of 50, have you made another batches that have worked well before - with the same design -? \$\endgroup\$ – Ahmed M.Zahran Jun 27 '18 at 23:04
  • \$\begingroup\$ This is a new design for the current 50, I have had other batches but with slightly different components. This rendition includes the surface mount diodes, buck converters, and that added R6 in an attempt to reduce the heat produced by the constant current drivers. the last batch of lights did not have that resistor and the constant current drivers would produce too much heat and eventually shut down. I corrected this by simply limiting the red channels output within the micro but that was just a band-aid. \$\endgroup\$ – Mr. Pivirotto Jun 27 '18 at 23:14
0
\$\begingroup\$

- update:

  • The unregulated bridge needs to output a minimum of 2.5V dropout above the 7810 10V output.
  • The input Vac is unknown but if Spec given is correct the output depends on the diode Vf at 8A pulse not 1A and the 141% rms to pk conversion
  • therefore C must discharge 1050mA 80% of the time (est)
    • while the diodes pump 5x the current 20% of the time Reducing the RMS Cap ripple current to 1.4A
  • the CC regulators must have 2.5v min @350m
  • the ref. LED specs are not accurate but imply a difference if 4V between G-B and R from min to typ to max.
  • therefore for slight margin the purpose of R6 is to drop 3.5 V with 0.5V margin.
  • HOWEVER without actual VI test results for ; transformer, diodes, LEDs and Cap ESR and C everything is guesswork

But when selecting R6 or any power resistor the power rating MUST BE DERATED for worst case board temp. Which could 50% or even 10% of power rating. !!!! For R6=3.5V*0.35A = 1.225W If actual case temp >100’C it is under rated. I would suggest THT resistor raised off board 3W but verify delta V (3 sigma) between Green ~ Red to verify 3.5V drop is ideal or less.

  • As far as bandaid improvements Disregard previous suggestions , measure all above and Vcap, dV , Vmin, V avg , then all other component voltages different than expected

  • then optimal fix can be made

  • FET LDO’s “ can be “ around 0.1V

  • bigger caps and lower ESR increases Vin

The purpose of R6 is to drop the difference between Green and Red @350mA

Although heat transfer is a mechanical problem , EE’s who are self reliant must master this like any RC network.

Solder Heat profiles MUST NEVER be exceeded for rate, dwell or Tmax. Why? LED’s must reduce visibility of extremely tiny fragile gold whisker wirebond. So you must never violate thermal specs. Otherwise bi-metallic contact switch thermal faults or damage results are your fault. 9 times out of 10) This can and does even happen to Cree LEDs.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ Alright just to make sure I am understanding this correctly... 1- In your simulation are the 5 5 & 5 resistor symbols before each channel still the constant current drivers that I am using? (datasheet.octopart.com/…) 2- increasing the size of the AC-DC filter cap is no problem, hopefully this does not make the physical size of the cap giant 3- You are suggesting that I increase R6 from 10 Ohm 3W to 12 Ohm 3W? Also I tried to mark this as useful post but I dont have enough rep yet in the electronics section of StackExchange \$\endgroup\$ – Mr. Pivirotto Jun 29 '18 at 16:08
  • \$\begingroup\$ I did not use CC chip and used 5 Ohm R instead. because it would not work. I don't have a full schematic and list of test results on every node like you. but I suspect 10V is dropping and CC chips don't have 2.5V min. on the flickering RG but RED is OK but hot? I suspect you need a 7812V Reg and need to remove 50mA of Green, test, then Blue, test, then Red if too bright otherwise leave. I cant do any more for you But since Cap RMS ripple current is closer > 1A and closer to 2A from charge pulses with 2mF Cap with low ESR. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 17:35
  • \$\begingroup\$ I suggest buy these. digikey.ca/product-detail/en/panasonic-electronic-components/… They are rated for 2.4A rms ripple current at room temp ( not 80'C) \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 17:38
  • 1
    \$\begingroup\$ I meant BG not RG \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 17:43
  • \$\begingroup\$ digikey.ca/products/en/capacitors/… \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 29 '18 at 18:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.