LM3429 LED driver demo board is malfunctioning..why?

[treez]

Hello,
We used the LM3429 evaluation board (Boost LED driver) with PWM dimming of LEDs and it mis-funstions..why?
The Board runs OK when it is not being PWM dimmed, and gives the correct 153mA of LED current, but when we apply the PWM dimming signal, the error amplifier output voltage (“COMP” voltage) rails high and the LED current rises to 190mA. (190mA when it is in an “ON” interval of PWM dimming)

Why does the error amplifier get railed high when PWM dimming is taking place?

LM3429 evaluation board:
https://www.ti.com/lit/ug/snva404b/snva404b.pdf

LM3429 datasheet:
https://www.ti.com/lit/ds/symlink/lm3429.pdf


Spec of our LED driver in use on the LM3429 evaluation board..
Vin = 12V
Iout = 0.153A
V(LEDs) = 54.4V (16 leds in series)
Off time = 656ns (F(sw) = 332KHz)

We changed components on the board to suit our spec, which is as follows..
R8 to 5k1, R6 to 0.22R, R9 to 3R3, R10 to 75k, and L1 to 100uH

 

[D.A.(Tony)Stewart]

“COMP” is the result of comparing the PWM On current sense after a 1us RC filter. But you are modulating using a 332 KHz switch or a 3us cycle, so your filter is 3T which does not reach 100% of Isense within one switch cycle and will be affected by boost duty cycle.

Since 1T response is 63% of a step (=1/(1-e))
I am not sure yet how the cycle-by cycle current limit affects this.

SO examine the filter response HSP and HSN for the problem and change C12. Also look for offset voltage. WHen PWM is off, COMP out should be floating and not drifting high. C8 in BOM is a 1uF Murata X7R cap for the S&H "Comp". In my experience ceramics make very bad S&H caps. so use film instead .

And show a photo of "COMP" before and after your fixes.

BUt your results are [mA] 190/153=124%

 

[treez]

Thanks, the funny thing is, that we also modified the LT3756 demo board to drive our LEDs, and that doesn't have this problem...

LT3756 Boost LED driver demo board..
http://cds.linear.com/docs/en/demo-board-manual/dc1319BABf.pdf

with or without the Q1 fet in series with the leds, the LT3756 demo board doesn't have this problem.

With the LM3429 demo bard, the COMP pin voltage rails high, to the 6.9V

 

[D.A.(Tony)Stewart]

you will have to show more info so I can understand your problem. graphs, DSO shots, testpts.

I thought I explained the problem

 

[treez]

thanks, as you said, that RC filter is just 1us time constant, so that shouldn't be causing the COMP voltage to rail high. I also changed the components on the CSH pin to 1k and 22n and it still rails high........I admit im not using film caps, but neither does the lt3756 board and that doesn't have this problem.

 

[D.A.(Tony)Stewart]

MUst add details as requested.

It is a sample and hold output, Does it sample correctly? Can you offset the hold by looking at the input or R to input or output?

Where are waveform?
What is I vs PWM? graph..
WHat are PWM rates? Does S&H work as expected on COMP out?
etc etc