How Fast Is Your Flash?

What’s cooler than learning about timers and interrupts on AVRs? Well, if you’re like [Matt], you can use that learning experience to build something useful – in this case, a timer for various camera flashes.

There are two ways to measure the speed of a flash. The first is the lag between when a button is pressed and when the flash goes off. As long as this is consistent, everything’s okay. The second type of speed is the pulse width. When looking at a xenon flash as time vs. brightness, they have a large spike at the beginning followed by a significant amount of decay. LED flashes are pretty much one cycle of a square wave.

To measure both types of flash speed, [Matt] used a $0.50 photodiode an a 3.5mm jack that ties into the flash remote. These bits are wired up to an Arduino, a little bit of fun work with timers and interrupts happens, and [Matt] learns how fast his flash is.

15 thoughts on “How Fast Is Your Flash?

    1. As always, he’s in the grey zone between experimental measurement and experimental measuring tools and all their transfer functions, but it looks like he’s measuring down into the single digit µs range (3µs seems the lower limit of the Arduino setup) and not squawking about accuracy relative to the Thorlabs detector/oscilloscope reference bench setup.

      More important, this seems like a good way to grab high speed data for other applications (shock data from accelerometers etc.). Also lots of other good stuff in the website about homebrewed ultra-high speed flash devices and the Vela One prototype (~500ns pulse) looks like a heap of fun.

  1. It is really refreshing to see someone using Arduino to it’s greater potential, beyond AnalogRead() and Delay() functions. Probably won’t get much feedback because regular Arduino guy now wonders ‘what the f* are comparators and timers, there are no libraries for that’, but from articles like this one people can learn a lot about time-critical applications of Atmel’s cheap micros, so please post more stuff like this.

    1. Yeah, my favorite stobes are some old potato masher Brauns, they top out around 250V; measuring that scared everyone in my apartment when I measured live to ground instead of trigger to ground (or might have been live to trigger), because even set to 10KV my little meter discharged it much faster than the caps expected. I’ve got an old hotshoe strobe from West Germany that runs on 2 AA and meters near 500V at the shoe.

  2. Seems like this is somewhat fundamentally limited by the lack of a known reference point for the comparator. He said he set his for 200mv, but when trying to digitize a complex waveform like a flash this seems problematic. Then again I suppose most people considering building something like this (as opposed to using some sort of oscilloscope, heck even a sound card based one is accurate down to a few tens of microseconds) are using LED flashes which are somewhat easier to characterize.

    One cool trick would have been to use a PWM out to generate the reference voltage (simple analogWrite on a PWM capable pin, and then through a low pass filter into the analog ref pin), then fire the flash multiple times with varying reference voltages to create a ghetto sampling-ish oscilloscope.

    Better yet would be to use the flash for its intended application and take a picture of a quickly moving object. Something like an old hdd with a white line drawn on it would work well, for a normal 7200rpm hdd (roughly 10ms per revolution) you would get a 10ms record time with resolution determined by the width of the line. If you used a 0.1mm thick line (doable with a half decent printer) on a 3.5″ hdd you would get about 2000 points, or 5us resolution (just about what he claims is possible with the arduino). Then simply note the intensity of the streak, which will be a direct measurement of the intensity of the flash as a function of time.

    1. That is similar to one of the ways described in old text books to time the shutter of the camera. Aim the camera at a tube TV tuned to static, and snap a picture at 1/1000, 1/500, and 1/250 and once you develop the film you could count the number of horizontal lines, and the distance between the first and the last, to determine the shutter speed.

      Only problem with counting the spins of an HD platter is that you end up measuring what the oscilloscope measures in this guy’s post, both the initial flash and the fall off. While it’s beneficial for someone selling LED strobes to claim that xenon strobes pulse for 140ms, that’s not the majority of the light and it’s not what you are usually capturing. That large spike on his graphics just spikes to the saturation point of the photodiode, it doesn’t measure lux.

  3. Doesn’t work on “real” stobes: the old ones that require a safe-sync for new cameras.

    Doesn’t measure lux, and assumes the total trigger to 0V time is the important part of the strobe flash. I don’t think 9 LEDs can put out even roughly 1/5 the light of a xenon strobe in that short peak time, like the saturated diode’s bias voltage leads you to believe.

    Think’s lag time needs to be low: shutters take time to open. Sure, a strobe designed exclusively for high speed work needs a fast trigger so you capture the image you want. But that Olympus 600FL has a guide number of only 180-something; that’s not a high speed flash, that’s what you put on top of the point-and-shoot to light up a small room or keep the 5-year-old on a sugar-rush still for a photo. Sure, even amateurs like me build some high speed shots out of the various little strobes we have picked up over the years, but . . .

    It’s on a blog dedicated to selling 9 LEDs for $1000. That much money would buy you Elinchrom, Quantum, Lumedyne, or even 2 LED Cube (I dunno, some similar product in photo stores, I don’t have one so I can’t compare). At the least, compare this LED strobe to something in the same field, even a Bolt bare bulb. But comparing a <$100 strobe to something they want to charge $1000 for is disingenuous and makes me question not just the results.

    1. Hi,
      You’re correct that it doesn’t give an actual lux reading, but a photodiode is quite linear in its response to incident light. I calibrated it against the DET10A photodetector. I’m not measuring total trigger time. I’m measuring time to and from 120mV, which works out as around t.1 in flash testing terms.
      Shutter lag isn’t a consideration in high-speed photography, as you’re using a long exposure and freezing using the flash. Most cameras’ shutter lag is way outside what’s acceptable for high-speed: we’re talking 50ms and up, with a variance in the tens of ms. That translates to tens of metres of bullet travel.
      You’d be surprised what a cheap speedlight can do with high-speed photography. A high GN isn’t the key factor (in fact most high-speed flashes have very low GNs). The key factor is pulse width, and that is (much) lower on speedlights than studio strobes like the Elinchrom or Quantum. Those often have minimum pulse widths in the hundreds of microseconds, or even over 1ms, as do most other LED flashes.
      This all comes down to the fact that most flashes are not optimised for high speed, as that’s not what most people use them for. Most people want high GN, which mostly comes down to larger caps to give wider pulses. With the Vela One I’ve made a flash that is optimised for high speed, and so needs very high peak luminous flux, very short pulse, and very stable lag. The reason it costs $1000 is that to achieve this I need to use very expensive LEDs and caps, as I need orders of magnitude more luminous flux than any LED flash on the market. That’s why this is the first high speed LED flash.

      1. I agree with you on almost all points. Posting the response time from trigger to flash, and comparing to a hotshoe flash in the graphics, just raises my eyebrows. I don’t expect a hotshoe flash to be a good high-speed strobe without a little tweeking of the internal circuitry. Yes, it’s an issue for high speed shots, but since the flash isn’t designed for that it seems unfair to expect it to be fast to trigger. It’s stability in that delay is what I think is important.

        Secondly, looking at your graph on the oscilloscope, it appears that the diode hit’s it’s peak saturation voltage (if I read the schematics right). Linearity isn’t the issue. Once the diode is saturated at it’s peak voltage, any data past that is lost; since testing it at a lower charge isn’t an option then some ND2 or even ND16 filter glass might provide some data above that saturation point.

        As for the big name strobes not being designed for high-speed, that I can appreciate. I haven’t gone through a catalog in a while, so I don’t know what’s out there in the xenon world that is designed for an open-shutter dark-room image capture. I have used cheap strobes for slower-that-bullet images, and synchronizing different brands of flashes is an issue I’m glad to see tackled. And I also didn’t mean to imply by “a guide of only 180” to mean that a higher guide number was desirable.

        But I do wonder if 9 expensive LEDs work better than 50 dirt cheap LEDs. The caps, not much room to budge there, but what is the color rendering or spectrum of your LEDs? What is the flux; either as lux, lumen, or guide number? You have something neat, I just think that selling it at a pro-photographer price while comparing it to budget-photog gear isn’t the way to sell. Maybe you are having great luck with that, in which case feel free to ignore me.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.