LM3409HV: Component failure at power-on / start-up

Hello Robert,

• What is the number of LEDs and the Vf at the output current setpoint?  Do you have a range in the number of LEDs? 
• I noticed you are using a 3A diode, that doesn't seem correct given an operating current of 4-5A.  I don't know what the Vled range so I can't calculate the duty cycle of the switcher and the rms currents through the diode.
• It can support 5A operation but that depends on the input and output conditions along with components used.  The P-channel FET driver doesn't have a strong gate driver, 2 ohms source, and 2 ohms sink.  That means depending on the fet used (ones with large gate charge) it might have a hard time to drive it quickly and turn on the FET fully before the blanking period.  That shouldn't make it damaged but it might give you much smaller currents or inconsistent operation.  
• Did you capture waveforms of the failure event?  It would be nice to see the inductor current, the SW node, Vled, and p-fet gate drive.  Maybe look at COFF signal too, to make sure you reach the off time.  We need to figure out now it get damaged and why before we can address how to fix it.  It this something that is fairly repeatable?  
• How are the supplies controlled?  Is 48V applied first or 12V?  It would be good to have power up signals of 12V and 48V rail
• Are you shunt fet dimming using Q2? What is PWM frequency and duty cycle range?  You schematic seems to have this as a no load.  How are you controlling the dimming?  PWM?  Analog (IADJ)?    
• Resistive load is ok, but using the actual LED is better.  LEDs are dynamic impedances, unlike resistors.  
• Yes,you can parallel current sources to get high currents.  

I really think you need to try and capture the even of when it gets damaged.  Try and use all the signals on your scope to see what is happening.  SW node, inductor current, Vled, and some other signal like gate drive, Coeff, etc.  This a very simple device and usually you can figure out what is going on once you get those waveforms during the failure event. 

 -fhoude

Failure occurs infrequently and I have not yet been able to catch any scope waveforms when the failure occurs. I have attached 2 waveforms: one during steady state operation (wfm1) at 2.5A and another at 4A at startup (wfm4). At startup the 48V and 12V are on, then the scope is triggered as the waveforms rise from an enable assertion. Yellow = LED volts out, cyan = D1,L1,Q1 node.

I do not have information on the number of LEDs in the string we are using from our customer. At the 4A drive current the voltage is 31V (max 35V).

On our test bench, the 48V and 12V are from separate supplies and are turned on manually. In the actual application, 48V will come up first, the 12V is derived from the 48V.

The board only has the analog dimming circuitry installed. The PWM dimming circuit is not populated. We intend to use analog dimming only in our application. For testing the board is currently using the pot to set the IADJ pin.

Robert D.

Hello Robert,

You need to look at all the nodes and currents I told you to look at during the power up cycling and trying to find the event that might be causing the damage.   You have 4 channel scope, please use all of them to get as much information as you can when the startup happens, see my comments before about what might be good node to monitor.  I hope you have a current probe, it would be useful to see the inductor current.   

I noticed you enable the EN pin at like ~4V, isn't there any reason you are doing that.  I would adjust that such that it used as UVLO for the enable tied to 12V.  

Do you have your design calculations for the component values you have selected?  I could review them.  

-fhoude

Attached are 2 scope traces at start-up. (yel=48V VIN, blue=LEDOUT current 100mV=1A, magenta=12V, green=LEDOUT volts)

Wfm6 shows that the VIN from our bench supply at start-up is unstable and spikes up to 100V (which is over the 75V max LM3409HV rating). Please comment and confirm that this is likely our issue. The board did not fail, but given the waveforms, I did not continue with this power supply. This test was done with the enable pin being held low at start.

Wfm12 is a start-up using a different power supply. The 48V DC supply is powered on with the enable asserted and the IADJ set for 4.25A. The 48V DC supply is a stable ramp up at start and the LM3409 output appears to perform as expected. Please comment.

We also intend to connect two of the LM3409HV board designs in parallel to one light strip setting the current for each board to 3.75A. You mentioned in a previous post response that the boards can be paralleled. Please confirm and comment.

Thank you,
Robert D.

wfm6_12.zip