by Toomas Tamm and Jeremy Stein
Table of Contents
- Discharge Graphs of Electronic Flash
- Experimental Setup
- Canon 430EZ at different power output
- Canon 380EX at different power output
- Dependence of Flash Output on Charge Time
Discharge Graphs of Electronic Flash
Electronic flash units work by passing strong electrical current through ionized gas (xenon). Part of the electrical energy is thereby converted into light. The source of the electricity is a capacitor, which discharges (loses charge) very rapidly in the process. For more details on the workings of electronic flash, please read the Electronic Flash FAQ.
In February 1997, Jeremy Stein, an amateur photographer and physicist working in Albuquerque, New Mexico, measured the light output of Canon 430EZ and 380EX electronic flash units. The results are described in the following pages.
For the impatient, here is the most important result:
This is the discharge curve of the 380EX flash unit at full power. The rapid rise and a slower decay of the light output are well visible. The horizontal axis is time, and the whole graph represents a duration of 5 milliseconds (1/200 of a second).
Now follows the full story, mostly as written by Jeremy himself.
Experimental Setup
The testing started with a fresh set of batteries, and I waited until the red ready light was on before flashing the unit manually. The testing was not done in any linear fashion, and there was no evidence of serious differences in the flash timings measured as the batteries wore down or heated up. By the end of the tests, after maybe 100 flashes, the time to recharge at full power was getting to be long enough to be annoying.
I used a Tektronix probe (TEK P6701 o/e converter) intended for the purpose of converting the light in an optical fiber to an electrical signal. This unit has a bandwidth of 700 MHz, so I felt it would be plenty fast enough. I connected this unit to an oscilloscope (TEK 210) and fired the flash unit into the fiber port of the sensor. I filtered the incoming light enough to avoid saturating the sensor (overdriving it into nonlinear response), and I printed the scope traces on a computer printer directly from the oscilloscope. The scope I used was not capable of sending the data to a computer for further analysis. When I do the more accurate setup, I will be able to do numerical analysis of the data directly on the computer, but the results are rather interesting even now. I extracted the following data from the scope printouts by measuring levels and timings with a pair of dividers and a ruler.
The data were taken with a TEK 210 digital sampling oscilloscope. For a 250 microsecond per division sweep speed, the scope takes one sample every microsecond. For a 2.5 millisecond per division sweep speed, the scope takes a sample every 10 microseconds, etc. The scope has a maximum sampling rate of 1 Gigasample per second, and its overall bandwidth is 60 MHz.
Canon 430EZ at different power output
First I set up a fixture so that the conditions would be the same for every shot, at least in terms of the light pickup and its relationship to the flash head. Then I ran a series of shots at the same scope settings over the manual range of power levels, from 1/1, 1/ 2, 1/ 4, 1/8, 1/ 16, and 1/32. I waited at least 30 seconds from the time the red light came on before firing each shot, so that the power level (or at least the charge level) would be the same. Then I reshot the lower power levels at faster sweep speeds so that it would be possible to get better data on them.
The risetime of the light from the flash unit is essentially the same for any power level. If not clipped off by the end-of-flash thyristor, the light output rises in about 100 microseconds, 10% to 90%. At the lower power levels, this time is apparently shorter because the light output is clipped off before it reaches full output.
A problem with making a measurement of this type is deciding what constitutes the end of photographically useful light output. At full and half power the light output level is strictly a function of the capacitor discharge; i.e., an exponential decay from the peak level. I chose to determine the time between the levels at half peak intensity and also those for a quarter of the peak intensity. These correspond to a range of one stop down from the peak and two stops down from the peak. It might be meaningful to check the timing for lower levels also, but only for the full and half power flashes, since the flash is clipped off and the time is essentially the same for any level less than about one stop down.
These data are in the form: power level, risetime (10% to 90%, microseconds), duration for one stop down, duration for two stops down.
Power Level |
rise usec |
-1 stop usec |
-2 stop usec |
---|---|---|---|
Full | 110 | 1000 | 1810 |
1/2 | 100 | 1000 | 1820 |
1/4 | 100 | 940 | 1010 |
1/8 | 115 | 365 | 405 |
1/16 | 105 | 155 | 195 |
1/32 | 90 | 88 | 120 |
This is the discharge graph at full power. One unit on the horizontal axis corresponds to 250 microseconds; the whole graph represents a time period of 2.5 milliseconds (0.0025 seconds). The rapid rise and exponential decay of the light output are clearly visible. | |
This graph is made at 1/2 power setting. The cut-off thyristor cuts the pulse at slightly after 1 millisecond after the flash has started. | |
1/4 power. The working of the cutoff thyristor is clearly visible. At less than 0.5 milliseconds, the current is sharply cut off. | |
1/8 power. The output is clipped almost immediately after full power has been reached. | |
1/16 power. Now the current is cut off well before it reaches its peak value. Cutoff time about 160 microseconds. | |
1/32 power. Cutoff occurs long before peak light output could have been attained. |
Canon 380EX at different power output
This file consists of data taken from the same camera and flash combination, starting out with new batteries, and taking a series of shots with the flash set on H at successively faster shutter speeds. The sequence goes from 1/125 to 1/4000 of a second. The pictures show several things about this flash system. There is marked change in the behavior of the flash when the shutter speed goes above 1/125 second; at this and slower speeds the flash behaves as in the data I took on the 430EZ. That is, the light output rises in about 100 microseconds, and then decays exponentially unless the system decides that the exposure has been completed, in which case the flash output is chopped off early. At shutter speeds faster than 1/125, i.e. those for which the focal plane shutter is never completely open, the flash modulates its output light at about 40 kHz in order to maintain a nearly constant light level for longer than the time it takes for the shutter slit to travel completely across the film plane. You can see that this time is about 13 milliseconds for 1/180 and goes down to about 9.5 milliseconds for 1/4000 second. During this discharge time, the light level changes by about 1 full stop or maybe a little less.
The electronics in the flash unit most likely have a transistor which completely shuts off current flow to the flashtube for part of each 40 kHz cycle, and relies on the relatively slow rate of deionization of the flash tube gas (low-pressure Xenon, almost certainly) to keep it well enough ionized to maintain a rather constant load impedance. In addition, although the current is pulsed on and off, the light emission from the still-ionized gas continues in reasonably even fashion during the time the current is shut off. The overall decay time constant seen in the H mode flashes is much longer than that seen in the 1/125 second and longer exposures. The peak intensity in the two modes is very different. In the normal flash mode, peak intensity in this series is about 2.8volts, while in the H mode, the peak intensity is from 350 to 450 millivolts. This accounts for the very low guide numbers for the H-mode flashes.
[Ed. note: “volts” means volts as measured from the output of the light-to-electrical signal converter. We expect to have a linear correspondence between the “volts” and light levels, but the numerical relationship between the two is not known to us.]
This is the discharge graph at 1/125 seconds. It is similar to those measured earlier for the 430EZ flash. | |
A small change in shutter speed from 1/125 to 1/180 drastically changes the picture. The flash unit switches to high-speed synchronization mode, the operation of which is nicely described by Jeremy above. | |
The images for shutter speeds between 1/180 and 1/4000 seconds are similar, with the duration of the pulse slowly diminishing. | |
Further changes in shutter speed (this one is taken with the camera at 1/4000 sec) cause only a small reduction of the duration of the pulse. | |
This image shows the start of the flash in the high-speed synchro mode, with a much more detailed resolution than the previous images. As soon as a certain level of output is acieved, the electronics start a rapid on-off cycle at approximately 40000 cycles per second. | |
This is the end of the high-speed synchro flash, with a rapid cut-off, apparently by an independent timing circuit, since the last wave is not complete. |
Dependence of Flash Output on Charge Time
I just finished testing (retesting) the 430EZ, and learned a lot more about its behavior. This thing is SMART!
I noticed while I was shooting the lower power levels that the duration of the pulse at any given power level seemed to vary. I thought that might be due to not waiting long enough for full charge, so I ran a very simple test. At 1/8 power, I waited a long time after the red ready light came on, then I fired eight shots at one second intervals and recorded them all on the same scope trace with no change in settings. The traces clearly shows that the flash unit lengthens the flash duration as its peak level falls off. I believe that the area under the light intensity versus time curve is held to about the same level by adjusting the flash time. This means, by the way, that stating what the duration of the flash is based on the power setting is iffy at best, and is a variable in any case. Note that all these results were obtained with no camera attached to the flash; the 430EZ decided itself what durations to use for the successive flashes.
On another day, the tests were repeated for the 380EX:
The series consists of data taken on a Canon 380EX, attached to an ElanIIE body, and flashed by the camera at various times after the red flash ready light came on. The times were: when the light came on, 5 seconds after, 10 seconds after, 15 seconds after, and 30 seconds after. The change in peak intensity is from 1.9 volts fired at the instant the ready light came on to 2.8 volts when fired 30 seconds after the ready light came on. Note also that the half peak level time for the first shot was about 950 microseconds, while the same measurement on the shot fired 30 seconds after the light came on was about 1150 microseconds. Nothing else changed during this series, so the scope data differences are due solely to level of charge. The batteries were nearly new, but I expect that the timings would be different for older or newer batteries. The data should probably serve more as a warning that it is best to wait a while after the flash says it’s ready, although I suspect the difference will never be seen in the pictures.
[Ed. note: “volts” means volts as measured from the output of the light-to-electrical signal converter. We expect to have a linear correspondence between the “volts” and light levels, but the numerical relationship between the two is not known to us.]
This image shows the output from the flash fired right after the red “flash ready” light came on. It is quite similar to those measured earlier. | |
If one waits 5 seconds after the “ready” light is on, the output is much more powerful, though. I would expect almost twice the Guide Number from this output compared to the previous one. | |
Waiting for 10 seconds adds even more power. A small “plateau” starts to form near the peak, showing energy reserves in the capacitor. | |
If you wait for 15 seconds after the “ready” light comes on, the capacitor has enough charge to keep an almost steady current through the xenon for about 250 microseconds. | |
Increasing the charge time to 30 seconds after “ready” seems not to introduce any furher interesting changes. |
Thus the observation also stated in the Flash FAQ that full GN is achieved long after the “ready” light comes on is also true for Canon flash units.