Design LED driver for machine vision flash

[Yaroooo]

Hello,

I'm looking for tips about suitable products to use in a machine vision flash light. Actually I'm using an LM3409HV at 24V (rarely at 48V) driving about 2 / 3 A output (depending on led panel configuration). I'm not really happy about it because it's too slow for some applications.

Mains problem are:
- Between EN signal and power ON passes about 11ms;
- Between power ON and stable LED lighting passes about 250us;

PCB design was designed by datasheet tips and examples. Value are also from datasheet.

I'd like alternative product suggestions to use in this application as alternatives of LM3409HV. I've only found one LT3932 but I'm looking for others to make a comparison.

Here's a picture of what I mean about first problem (about second one it's difficult to show it by images):

[ajb]

What switching frequency are you targeting for the LM3409? What sort of on times and frequencies are you targeting for the LED?  Can you show your schematic?

The datasheet shows much faster rise times than you're seeing here with PWM, so it could be a problem with component selection, or it could be that since you're presumably providing a fairly long off-time the low-power shutdown feature is being activated, and the long startup time is due to the part coming out of that mode and more or less starting from cold rather than just enabling the output.  The datasheet doesn't really indicate what startup time from shutdown should be, or if it's different from the specified rise time in PWM operation.  You could test this by applying a PWM signal and see if the behavior changes. 

In some cases, using external PWM dimming may be preferable to using internal dimming with a switching converter.  When using the EN pin for PWM dimming the converter has to bring the inductor current up from zero and then back down, which means the rise and fall times depend on the transient response capability of the whole thing.  If you put a FET in parallel with the LED, you can turn the LED off by turning the FET on, effectively shorting out the LED.  That allows the converter to continue operating and inductor current to continue flowing while the LED is off, allowing for much faster PWM, and since the output voltage is near zero when the FET is on very little power is consumed.  However this is probably not necessary in your case unless you need very fast rise/fall times, and it certainly seems like you should be able to get far better performance with the converter using the enable pin than what you're seeing now.

[Marco]

Does energy efficiency really matter? If not why use a switching controller?

[StillTrying]

Is this something to do with the slow enable.

Nope.
Going by this plot you should be able to use EN to get the LEDs to reach full brightness in around 4us.

https://www.ti.com/product/LM3409HV
http://www.ti.com/lit/pdf/SNVA390

[Siwastaja]

If you can accept slightly lower efficiency (like 70-80% instead of some 90%) and somewhat poor current regulation (i.e., depends on LED die temperature), use plain old regulated voltage source + series resistor. The more you drop the voltage over the resistor, the better the current regulation but worse the efficiency.

This way you can switch with simple low-side N MOSFET and get some 20ns switching time no problem. Using this classical circuit, I have successfully modulated up to 20MHz in 3D time-of-flight imaging system.

Using a current source and shorting out the LED is another option.