LM3409EVAL: incorrect switching frequency

Hello Felipe,

What is the input voltage?  Output is 35V (10 X 3.5V).

Best Regards,

Hi Irwin,

The input voltage is 48V. Furthermore, my calculations were done assuming a Vo=41V, in order to compensate for the change in LED forward voltage with temperature.

I have a few more issues which I can't explain...

Let me give you more information to help you understand better what is happening.

1) using the LM3409HV EVM, with the changes I mentioned in my initial post, I get the frequency "issue" already discussed;

2) I have another 2 custom prototypes using similar component values as the EVM mentioned in the initial post. One operates in the same way as the EVM, whereas the other switches the LEDs (output) on as soon as the power supply is switched on. In addition, with the latter prototype, Vo=Vin and the enable input has no effect (no modulation at the gate of the pmos either). When I had this issue in the past it was either because:

a) Roff/Coff had incorrect values;

b) pmos was connected incorrectly;

c) pmos was damaged;

I checked that Roff and Coff are correct and replaced the pmos (SiR873DP-T1-GE3), just in case. There was no change in operation. Replaced the LM3409HV...no changes observed also.

The fact that one prototype is working in the same way as the EVM suggests to me that there can't be any PCB layout/footprint errors, and I checked the same components are populated in both prototypes. 

The ONLY difference between the 2 prototypes is that in the one that is not working I'm driving Iadj input using another external power supply (with the RC filter connected), and the enable input is connected to a wire that is unconnected (floating), which I can then connect the enable input to either GND or 5V.

Independent of where the enable input is connected to (GND or 5V) when I switch the LED driver on, the result is always the same: the LED comes on fully and no modulation is present at the gate of the pmos.

However, if I toggle the ENABLE input between GND and 5V, this starts the modulation at the gate (1.7MHz), but if I try to increase the output current (via Iadj) then the system goes back to the previous condition, i.e. fully on and no modulation at the gate of the pmos. 

Any thoughts?

Regards.

Hello Felipe,

Generally the Vf goes down with temperature.  If the LED Vf is specified at a higher operating current the Vf will probably be higher at room temperature but going up to 41V seems like a lot.

If I use 35V for Vout I get 840 KHz with your values (assuming Coff is 470 pF), if I use 41V for Vout I get 531 KHz.

This has very low ripple and the MOSFET is a rather large MOSFET so I'm wondering if you are running into a leading edge blanking issue.  This would also cause the frequency to be higher since the on-time would be short.  Average current calculates to 2.95A, peak to peak current ripple is 0.166A.  This is 8.3% current ripple.  On the current sense this deviation is 13.6 mV.  Noise, leading edge turn-on current spike can cause the peak current trip to trip early.  Also note that the LM3409 alternates the input on the error amplifier to increase current regulation accuracy.  A requirement is that the peak to peak current ripple should be above 24 mV (being lower won't damage the part it will just appear to run at half frequency).

If you have a way to look at the current across the current sense resistor look to see if it is above the current trip threshold after the leading edge blanking time, 115 ns typical.

The MOSFET chosen may end up being too large for your design, optimizing switching loss and conduction loss.  Also I would use a lower voltage MOSFET since RDSon X gate charge tends to go up with VDS voltage rating increase.

As for the board that turns on immediately, unless it is damaged, I'm guessing that the power source is current limiting and the LM3409 MOSFET is full turning on but cannot reach the peak current trip threshold.

Best Regards,

Hello Irwin,

Thanks for your reply. I meant to say with to compensate Vf due to If. The LED I'm using can vary from 33V to 39V due to If.

Perhaps my understanding of the calculations might be mistaken...

I used 41V as I want the driver to be able to drive the LEDs at up to 2.9A ILED average, and an ILED ripple of 140mA peak-peak. Should I be using 35V instead? If I do that, won't the LED driver be limited to a maximum output voltage of 35V?

With regards to the Fsw, shouldn't it stay constant?

My understanding was that only the Duty cycle would change to adjust the output voltage (and as a consequence the output current). If the Fsw is changing with Iout, is that what's causing the issue?

I had considered the 24mV constraint you mention, but if this were the case I'd expect both prototypes to fail. Looking at the current across the sensing resistor will be tricky, but I'll give it a go. Perhaps I will recalculate component values to have a higher delta_i_L_pp and use an output capacitor to smooth the ILEDpp.

With regards to the MOSFET... I don't fully understand your comment: "being too large for the design, optimizing switching loss and conduction loss". Note that this MOSFET has an RDSon and gate charge smaller than the MOSFET in the LM3409HV EVM, which I would expect to give me the same switching performance with less heat generated (losses).

Finally, what are the common failure modes for the LM3409HV? In other words, what conditions can I expect to observe which would indicate driver IC failure? (besides the trivial "no output")

Regards,

Hi Irwin,

Thanks for all your help and for putting up with me thus far. Ironically, I found the section in the datasheet that explains the variable frequency, right after I had replied to your previous message.

With regards to the non-working prototype...

I tried replacing the LM3409HV, the mosfet, the diode, Coff.

Regardless of what I try the result is the same... LED comes on as soon as Vin is applied and the enable / Iadj inputs have no effect whatsoever. 

Considering your last note... The power supply I'm using is 48V / 2.1A. To decrease the peak current further I replaced Rsns for a 100 mohm resistor, but that didn't make any difference. 

I have run out of ideas as to what to try. I tried measuring the voltage across Rsns, but no luck (too noisy). I wanted to measure the inductor current but I don't have a current probe available, unfortunately.

If you have any other suggestions I'm all ears. What really bugs me is that the other prototype, with the same PCB layout / components (original values), is working correctly.

Regards,