I sometimes notice that I get inconsistent exposure even when using flash on manual. This inconcistency is not a huge problem, but it has been confusing at times as it makes me worry about technical issues when I should focus on the subject.
I was doing measurements on some flashes the other day, and I thought I’d check consistency while I was at it. On full pops, variation was less than 0.1 stop. But interestingly, none of the flashes could give a full pop at the time the ready light turned on. They all needed another 4-5 seconds or more (dependent on the state of the batteries). I think this phenomenon is what I’ve been experiencing. When shooting quickly, the flash output may vary by as much as a full stop.
Here are my measurements:
The chart shows a fully charged flash compared to the flash you get if you press the shutter button at the moment the ready light turns on. I checked some different flashes, from the high end Nikon and Canons down to the cheap ebay flash. For the SB900 and 580EX, the max power is ƒ36, but the ready light turns on at ƒ25.3 on average. The difference is 0.8 stops (or a bit more than 2/3 of a stop).
As a general rule of thumb; if you take the image right when you hear the beep, you get 2/3 stops less light than you expect. You have to wait twice the recycle time to get the full amount of light.
- Nikon SB900: 0.8 stops
- Canon 580EX: 0.8 stops
- Vivitar 285: 0.9 stops
- Metz 36: 0.6 stops
- Youngnuo: 0.5 stops
There are several reasons why these numbers vary: how easy the different test buttons are to press, variations in my reaction speed, and of course where the trigger point is set by each manufacturer.
So why does it work this way?
I don’t think there is anything particularly nefarious about the fact that the ready light turns on before the flash is truly ready. My guess is that this has historical reasons. Old flashes like the Vivitar has very simple electronics. Because of component variations, temperature drift, ageing and whatnot the ready lamp is set to light up a little bit before the capacitor is fully charged. If not, there is a risk that it never triggers.
The new high-end flashes like the SB900 or the 580EX has a microcontroller and there should not be any technical reason to turn on the ready lamp 0.8 stops before the flash is actually ready. But who wants to do that, when it will make the specs look a lot worse. 10 second recycle time or 4 second recycle time?
The numbers above are from the 50mm setting. In theory they should equal the GN number, but they don’t. I’m not sure what exact setup the manufacturers use to measure the GN number. I would be interested to know.
By the way, here is the output from the SB900 and the 580EX. They are so similar, it’s friggin scary.
All measurements are done manually but averaged over a number of readings, at least 5 for each flash and zoom combination. I could rig something to trigger automatically and take out the human response time. My guess is that 2/3 would increase to a full stop. Let me know if there is any compelling reason to do that. I measured what flashes I had at hand, and they are all well used. Measurements were taken in a single session and the instrument used was a Sekonic L-758.
Edit: There is an interesting discussion on this topic over on the Flickr Strobist group.