LED7708 Random LED switch off issue with driver

[Shapeshifter]

Our requirement was to create a LED driver circuit to power 12 strings of 7 LEDs from a 12V DC supply with a nominal current of 20mA per string.
Using the STEVAL-ILL035V1 Development board data for PCB layout guidance and the LED7708 design tool spreadsheet for component selection we successfully built a prototype driver PCB and produced firmware to setup the on-board registers and control the LEDs for our application.

Initial results were good and all appeared to be working as expected until we noticed that after a random period (5 to 10 minutes) the LEDs would go off completely and wouldn't recover normal operation untill the device was disabled (DEVCFG0.DEN = 0) and then re-enabled.

When the LEDs go off the boost converter goes into a skipping mode of operation but no errors are indicated either from the XFLT pin or the status registers.

After disabling/enabling the device, the LEDs would come back on normally but would then go off again some time later. Interestingly this would be a much shorter period than the initial illuminated period. I'm unsure if this indicates a heat issue but temperature measurements (taken at 25DegC ambient) of the LED7708 device and PCB don't show anything outside expected running temperatures.

If the LED brightness is lowered from 100% to below 75% or the number of enabled strings is reduced this appears to have an effect on how long they will stay illuminated. Sometimes they will stay on indefinitely at 50% dimming.

If anyone has had issues like this in the past, maybe you could tell us about your experiences or possible solutions you can suggest.

Many Thanks,
Shapeshifter

[Peter BENSCH]

Welcome @Shapeshifter, to the community!

I am quite sure that you have a heat problem. As you describe, the tests (including long-term tests?) with the STEVAL-ILL035V1 were successful. What are the differences between the EVAL and your board?

Which components do you use for the boost stage (L, T, D, C output)?

What does your layout look like, especially under the LED7708?

Regards
/Peter

[Shapeshifter]

Hi Peter, many thanks for picking up our design problem

To give you a bit more general info on our design, I've attached the schematics of the Drive cct, together with the PCB layout.

The main components are:
Input Caps: 56uF Elec together with 2x 4.7uF 50V ceramics
Inductor: 22uH Isat 1.6A (Bourns SRR6038)
FET: 30V 8A (Diodes DMG4466SSS)
Diode: Schottky 60V 1A Vf 0.5V (ST STPS1L60)
Output Caps: 2x 10uF 50V ceramics

Driver Schematic

Layout

Copper bottom

Copper top

Main design parameters, as used in the spreadsheet:
- Input voltage 12V +/- 10%
- 12 channels, each comprising a string of 7 LEDs, Vf 2V - 2.6V, Iled 20mA
- Expected output voltage; Vdrive 14V - 19V
- typical Dccm ~0.3
- Output power; Pout < 4.5W

The problem symptoms, of the LEDs turning off after a period, feel temperature related. However none of the calculations and measurements suggest there is a power dissipation problem.
The spreadsheet suggests a Pdchip of 480mW, Tj 42degC. Measurement with a thermal camera gives ~44degC.

[Peter BENSCH]

If I take a closer look at your layout, I can see the 13 thermal vias on the blue rear side, which transport the heat from the LED7708 there. However, the GND area there is separated by two tracks, which reduces the heat transport in the area to perhaps 25% of the possible value. The STEVAL-ILL035V1 also has a few tracks around it, but at a greater distance, so that the heat flow is less impeded. Additionally the EVAL uses 16 thermal vias instead of 13. However, as your thermal imaging camera only measures 44°C, this may not be a problem.

BTW: what is the status of XFLT and the DEVCFG0 register in the event of the shutdown?

[Shapeshifter]

The register settings are:
- DEVCFG0; 0x2130
- DEVCFG1; 0xD840

With a 'debug mode' implemented using the host microcontroller, the status of the XFLT and control registers are continually monitored. No faults are seen (OVFW, OTAF, CFP and OVRS), and this is the case whilst functioning normally and after the shutdown issue where the LED arrays are turned off.

After a shutdown, the host microcontroller interface is still fully functional.

Will any of these faults cause the XFLT output to latch?

From the datasheet, an 'overcurrent' occurance is not flagged up by the XFLT or any register. It describes this fault condition 'skipping subsequent switching cysles' until the fault is removed.

[Shapeshifter]

An 'overcurrent' situation seems to be the listed error state which best matches the shutdown situation that we're experiencing.

An 'overcurrent' cause may also make sense, in that the situation is made worse by increasing the brightness level and/or the number of strings being driven.

For our setup, Iocp is recommended as 0.75A. It is set to 0.9A with Rclim at 75K. This value has also been doubled to 150K but appears to make no difference.

Attached is a waveform of CSNS measured across the 68mR sense resistor. Could the switching transients be causing an issue? Although the datasheet suggests this is not a latched fault and so should self-recover.

CSNS waveform

[Shapeshifter]

As an aside, comparing our design with the eval board. We've noted that the evaluation design has an additional Schottky power diode; 'D2' across the inductor and switching diode (L1, D1), What is it's functionality, this device is not mentioned in the datasheet?