This week’s video is about very high speed flashes and whether LED based flashes are practical at these speeds. The dominant flash in this space is the air-gap flash with it’s half microsecond flash duration it is about 100 times faster than a typical speedlite flash from Canon or Nikon. I do some calculations that try to figure out how many LEDs you would need, what kind of current you’d need to driver those LEDs, and the efficiency of an LED flash compared to an air-gap flash.
The downsides of an air-gap flash:
* Limited places to buy and costs $4000+
* Dangerous to make since they use 10,000 to 20,000 volts which can easily kill
I frequently have people asking about whether an LED flash will work. On the surface this seems nice because LEDs are fast and they work at under 5 volts. There are problems though and they related to the duration of light. The air-gap flash produces a huge amount of light in a very short duration.
So now I will figure out how many lumins the air-gap flash produces in 0.5 us and then figure out how many LEDs we’d need to make a compariable LED flash.
If you look in “Electronic Flash, Strobe” the peak output of a EG&G 549 flash is 5 X 10^6 candlepower, and this is quite similar to the amount of light my homemade air-gap flash generates. Wikipedia says old candlepower unit is about 0.981 candela. So the 549 is 4.9 X 10^6 candela. One candela equals one lumen. So a 549 has a peak output of 4.9 X 10^6 lumens.
Now let go to LEDs. Let’s use the SST-50-W57S-F21-GJ200 LED since it was mentioned in one of the threads on our forums. The operating current of this device is 1.75 amps and the maximum output is 500 lumens. A paper entitled “Pulsed Operation of High Power LEDs for Imaging Flow Velocimetry” C. Willert 2010 gives some data on pulsed high power LEDs. It says that for some LEDs it was possible to run them at 20X the DC rated current. The light output went up also, not one to one, but let’s be optimistic and assume that it is one to one. They also write that driving the LEDs at high current with a pulse risetime below 0.5 microsecond instantly destroyed the LED (meaning no microsecond light pulse) but lets ignore that also. So, at the most optimistic, this LED will put out 20X500 lumen or 10,000 lumen per device.
Finally, how many LEDs do we need to equal a microflash output? 4.9X10^6 divided by 10,000 gives 490 LEDs. The price for this LED at 100 quantity is 13.87$. So the price for the 490 LEDs is $6796. There is also a lot more cost for the pulse capacitors needed to store the energy for the LEDs and for the power transistors to control the microsecond bursts.
The LEDS will use 490 * 1.75A * 20 * 4V = 68600 W/sec
The airgap flash will use:
W = 1/2 C V^2
1/2 * 0.03uF * 20000V^2 = 6W
Then multiple by 1M to convert from 1us to 1 second so we have 6M W/sec
This means the LED is about 100x more efficient.
For slower speeds the LED flash might make more sense, but then it needs to compete against speedlight flashes.
I’d like to thank Alan for all his help compiling the information used in this video and post.
* Great paper about pulsing LEDs with very high currents
* The EG&G 549 Manual
* LED used in my calculations
* Camera Axe forum thread 1
* Camera Axe forum thread 2
* Camera Axe forum thread 3
* Camera Axe forum thread 4