Very fast LED driver for flash applications

[Yaroooo]

Hi all,

I'm asking for advices about an LED Driver IC suitable for vision application. What I'm trying to archive is a flashlight with ON time of 0.5ms that is enough to take a photo in a short time.

Using LM3409 I can get a good quality illumination, but LEDs are fully ON only after 10ms. In this case I need a faster IC driver. Do you have any advice?

[ledtester]

Microchip has some specialized LED drivers for camera flash applications:

https://www.microchip.com/ParamChartSearch/Chart.aspx?branchID=99889

Such as the MIC2870:

https://ww1.microchip.com/downloads/en/DeviceDoc/MIC2870-1.5A-Synchronous-Boost-Flash-LED-Driver-with-I2C-Interface-DS20006078A.pdf

Have a look at Figure 2-13 for the chip's flash mode timings.

Similar chips are available from other manufacturers.

[mawyatt]

Not a single chip solution, but LED driver design for use with macro photography.

https://www.photomacrography.net/forum/viewtopic.php?f=25&t=40999

Best,

[Yaroooo]

I forgot to specify that it's an industrial application. So I have a 24V to 48V power supply. These chip I've already seen, they are specific but unfortunately they have a low voltage range.

[mawyatt]

The design mentioned above supports higher voltage & current LED COB modules. We've used it with 36V LED COBs, with proper sized components can support over 60V.

Be sure to read the entire thread, and study the schematics. It's a simple design that uses cheap COTS parts, the last design at the end is more complex and involves a RPi. We never ended up building the last design tho, got side tracked with other projects, but did build the initial version and it worked quite well and has been in use. If you are interested in macro use, check out the modified Jansjo LED with much higher output LED COBs, these were used with the controller.

https://www.photomacrography.net/forum/viewtopic.php?f=25&t=41464

Best,

[georges80]

I use the LT3756 on a few of my drivers. The LT part has a dedicated PWM control input that drives an external FET to turn on/off drive current. The on/off delay time of the output current is a couple of microseconds from full drive to off or off to full drive. Some of my customers have used one of the drivers for high speed photography.

The LT part has a max 100V operating voltage. Obviously you need to choose the external components (fets, caps, diodes) with appropriate ratings.

You just need an appropriate trigger/signal to drive the PWM control input.

cheers,
george.

[Marco]

I use the LT3756 on a few of my drivers. The LT part has a dedicated PWM control input that drives an external FET to turn on/off drive current. The on/off delay time of the output current is a couple of microseconds from full drive to off or off to full drive.

So do you use it with an ordinary diode and film capacitor (for low leakage) and simply do one flash on circuit turn on to dial in the LED supply voltage?

[georges80]

^ ??

The LT part is a current regulated (if configured) that way buck or boost switcher chip. Basically a one chip LED driver. You just need the external components (FETs, diode, caps, resistors etc. Then drive the LT PWM pin high/low as needed to drive the output.

Check out its datasheet if you want more details.

cheers,
george.

[Marco]

When the capacitor isn't charged the LED current will ramp slower than when it is.

With continuous PWM this will never be an issue, but if you try to use the PWM input as a trigger for a flash it can be.

[georges80]

^ it's about 25usec from PWM OFF to full drive current when PWM goes to on. That's with several seconds of OFF time and a single on time pulse. Plenty fast for the OP I would imagine.

cheers,
george.

[Marco]

Even with a Schottky and an electrolytic I doubt it will be significantly discharged in a few seconds, that's not the worst case scenario. Soft start for the datasheet implementation is in ms range as is the loop compensation. Worst case scenario is first flash after start up.

Ultimately it's hard to beat simple MOSFET hard switching with a resistor for current limiting and a fixed voltage ... it's a flash, wasting an extra couple 10% of power on heat in the resistor is not an issue.

[georges80]

^ Ceramic caps throughout. This isn't soft start, this is PWM. Measured with 50MHz Tek current clamp and caps discharged as far as a boost driver will let them discharge between 'firings'. Several seconds is more than enough (measured) to drop the output cap down to 'discharged'.

So, this is measured on the bench with real multiple series LED (about 20V of Vf) from nominal 10V input.

Yes, properly charged cap to working voltage and a fast FET to switch is fine for a single flash cycle. The OP was asking about a faster responding LED driver chip, I provided the info for that.

cheers,
george.

[mawyatt]

Ultimately it's hard to beat simple MOSFET hard switching with a resistor for current limiting and a fixed voltage ... it's a flash, wasting an extra couple 10% of power on heat in the resistor is not an issue.

Agree, and if you use mostly the same components rearranged in a current mode scheme shown above in my earlier post you can achieve fast and easily controlled current mode pulse width and amplitude that are almost independent of the LED, temperature and power supply characteristics. For our macro stack & stitch work, which could involve 10s of thousands of individual images, and need to have all individual images exactly exposed the same which could span over hours, even days, current mode operation became a requirement, and exactly why we chose this method

The simple resistor in series requires a very large supply voltage to keep the LED current somewhat independent of LED characteristics, temperatures and supply voltage. Using the current mode scheme, the supply voltage requirement is relaxed, as well as the resistor, but you need to use a quality current sense resistor with low TC and low inductance for faster pulses. This concept could easily achieve sub-microsecond LED PW with faster components, but our goal was to try and duplicate the effects of a Xeon strobe (although not in intensity) so 10 us was consider acceptable as the faster pulse range.

I know this wasn't a single chip solution, but it's not complex, is cheap to build, works will LEDs of 2~50V @ current pulses of many amps and works very well indeed for macro photography. The circuit is easily extended to higher voltage and currents with appropriate components & heat sinking.

Best,

[StillTrying]

If you've already got up to 48 VDC for the LEDs what's wrong with a 555 driving a mosfet.

I was surprised at how clean the simulated 7A 0.5ms pulses (white trace) are even without any low R in series with the LED. to suit.

[jonpaul]

Hello all the title...VERY FAST LED driver for FLASH...prompts the historical note:

In 1940s, Dr Harold Edgerton at MIT invented the fast xenon flash technology used for military and defense use eg night photo mapping of targets and nuclear device testing photography.

With special construction of the xenon flash and trigger with spark gaps and krytrons, etc, Edgerton was able to achieve 1-10 usec  flash and jitter.

http://www.nasonline.org/publications/biographical-memoirs/memoir-pdfs/edgerton-harold.pdf

By the 1950s and 1960s,  developments brought the duration and jitter down to nanoseconds!

Enjoy.

Jon

[Terry Bites]

http://td0g.ca/2016/07/28/ballistic-chronograph/

[Terry Bites]

if you use an application specific MOSFET driver life will be a lot easier. Many have nanosecond rise times. I simulated this with LCW H9GP LEDS. Its a starting point.

[amishasingh]

What sort of power consumptions does these led drivers generally do ?

[Doctorandus_P]

I think I would first go for (something like) that 555 driven circuit, with an extra current shunt and transistor to turn the fet into a current sink. Maybe even simplify to just a resistor for current limiting. It probably switches in less then a microseconds. Maybe a cap over R2 to compensate for the gate capacitance of the fet can make it a bit quicker still.

I would avoid any switching topology for this. Inductors need time to get some decent current through them, which will make your driver slow. Efficiency (as for regular lighting) is also total irrelevant for this.