EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: Crusty on June 05, 2012, 05:21:08 am
-
I was interested in using a scope to measure a camera's flash duration. I have a few flashes and a few strobes, and I'm curious how fast the pulse of light is, so I know what type of motion I can freeze. For example, some flashes state 1/300 of a second, and some go up to 1/10,000 of a second or higher..
I found someone that did essentially what I want to do here:
http://www.gock.net/2012/01/flash-durations-small-strobes/ (http://www.gock.net/2012/01/flash-durations-small-strobes/)
but his results seem off, they seem to be 30% below the the manufacturer specs for the few flashes and strobes I checked. In fact many internet sources have different results.. In this particular case, I'm wondering if it was the fact that he used a photo transistor instead of a faster photo diode. (the photo transistor he used has a higher turn on (rise) time than he states in his blog) I also tried to contact the author but got no response.
I am guessing I will need a Digital Storage Oscilloscope, and I was considering the Rigol 1052E.. It seems the author of the above blog also appears to be using a Rigol scope, but given the error, I'm wondering if the scope was not the best choice for this type of measurement, or did he do something wrong (stray capacitance)? Or maybe all the manufacturers misstate their flash duration times?
Also, as a secondary usage, if I did get the Rigol 1052E, I would like to test AC power to see if it was a true sine wave, modified sine wave, and/or detect bad/dirty power. I actually have a few devices that generate power for strobes (claim to be true sine waves), and it would be nice to analyze them under load, as I have some doubts. Would be nice to look at generator power as well. Would I need a special high power probe for this?
Thanks
-
Yes, phototransistors are generally slow. A photodiode + transimpedance amp should work much better. If you collect enough light a photodiode + 50 ohm terminator can even work.
The slowest scope is fast enough to capture a 100 microsecond flash pulse. Any errors are in the detection system. You can't just connect a photodiode or a phototransistor emitter to a 1 megaohm scope input, you need a load resistor.
-
you can very quickly set up a counter on a breadboard with a 1Mhz crystal clock oscillator and a binary counter chip.
You still need the photodiode and transimpedance amplifier, probably a schmidt trigger or a comparator. Use the schmidt trigger or comparator output to gate the clock into the counter.
the counter value is the duration of the flash, in periods of the clock. So a 1 Mhz clock has period 1 microsecond. A 100 microsecond flash will show a count of 100.
This will be much more accurate than reading off a scope trace. use a 10Mhz clock and a 16 bit counter and you have a much more accurate count.
you can do it with a microcontroller too. Use the internal counters.
-
but his results seem off, they seem to be 30% below the the manufacturer specs for the few flashes and strobes I checked
you need to define the threshold (much like lo-V hi-V in digital) since flash light is not a nice square shape. you can search google, i found many info last time, one like http://www.chem.helsinki.fi/~toomas/photo/flash-discharge/redwait.html (http://www.chem.helsinki.fi/~toomas/photo/flash-discharge/redwait.html)
i never heard there's any need for accurate flash duration measurement, practically by any photographers. except maybe let say if you wanna shoot fancy picture of bullet with trailing light. people use flash burst to freeze motion. for more accurate measurement, dont expect some simple phototransistor+wiring will do, you need like true light/lumen measurement sensor to do that + fancy circuitry, or some fancy physics setup such as rotating stick/disc and measure the trailing angle. but again its not so easy, since flash burst is not a nice square pulse.
if you talking about stroboskop interval, then it will be much more make sense. imho.
-
Thanks for your responses.. I've included a few questions below:
> A photodiode + transimpedance amp should work much better.
> If you collect enough light a photodiode + 50 ohm terminator can even work.
> You can't just connect a photodiode or a phototransistor emitter
> to a 1 megaohm scope input, you need a load resistor.
Thanks.. I will need to look up what a transimpedance amp and a load resistor is, though I think a load resistor is just a resistor connected to the diode.
> you can very quickly set up a counter on a breadboard with a
> 1Mhz crystal clock oscillator and a binary counter chip.
I like this idea to test things out initially (on the cheap), though it would be tricky to figure out the t0.1 and t0.5 numbers most flash manufacturers quote.. Plus, this is challenging since I am more of a computer science guy.. I need to learn.. Any help would be appreciated.. Also, you mentioned a micro controller.. Like an Arduino board? Would it be fast enough and have what I need?
> you need to define the threshold (much like lo-V hi-V in
> digital) since flash light is not a nice square shape
Usually the standard is t0.1 and t0.5.. Not based on voltage.. I like the link you provided. His solution was unique as he used a special probe for opitical fiber cables.. I will need to look into that as well.. I wish he went into more detail on how he did it..
> for more accurate measurement, ... , you need like true light/lumen
> measurement sensor to do that + fancy circuitry
Sounds discouraging...
-
Phototransistor should be fine at these speeds - probably want a fairly low value pullup (1K maybe). Not sure to what extent PTs can saturate at high light levels, possibly extending the on-time but you could check this by attenuating the light,
A flash puts out enough light that almost anything will detect it - a photodiode in voltaic mode would probably be fine - put maybe 10K across it to discharge self-capacitance.
-
for more accurate measurement, dont expect some simple phototransistor+wiring will do, you need like true light/lumen measurement sensor to do that
No, a photodiode is all you need for measuring the flash waveform. "lumens" are only relevant if you need to do radiometric measurements, i.e., calibrated measurement of total amount of light. Phototransistors may also work, but I am prejudiced against them and avoid their use for anything but low speed digital use.
-
No, a photodiode is all you need for measuring the flash waveform
light measurement is quite ambigouos to me though. say if you want to measure ambient light, or projected/incident/focused light. from my experimentation with simple photodiode+speaker cable+recorded to PC software, the closer you put the diode, the more saturated it will be, i have to filter the incoming light with darker plastic film to get nicely nonsaturated signal. so it depend on the distance from light to sensor. we (i) need more definitive standard here.
-
> No, a photodiode is all you need for measuring the flash waveform.
That's what I thought.. What's the benefit of just a photodiode vs a photodiode + transimpedance amp?
I google'd and found an OPT101 which is a chip that has both the photodiode and Transimpedance amp on one chip.. Seems easy.. Though, I question the amp, as the flash will create a lot of light which should cause the photodiode in voltaic mode to create a lot of measurable voltage.. And if it is too much, I can move the flash away from the photodiode until I get the response voltage I want..
I am most concerned about accuracy though.
Thanks
-
> so it depend on the distance from light to sensor. we (i) need more definitive standard here.
The flash or strobe can be moved close or away from the sensor so that I get the voltage response I want.. I want to test the flash at different power levels (as the power affects how fast the flash burst is).. Full, vs 1/32, etc..
So if I have a scope, I can use that to determine distance so I don't peg (saturate, clip, etc..) the sensor.
Thanks
-
and again, might be relevant. my bro used to compete in car-woofer competetion. they use term-lab as sound pressure measurement. in each competetion, there's a guide on what distance the sensor will be placed from the speaker/woofer. because different distance will give different reading. i guess its the same as projected light. you put your sensor closer, you'll get different measurement profile.
I want to test the flash at different power levels (as the power affects how fast the flash burst is).. Full, vs 1/32, etc..
remember the simple rule.. light intensity = (1/distance^2). you move your sensor 2x farther, you'll lose 3/4 of the light.
Usually the standard is t0.1 and t0.5
what is t? how do you define the starting and ending of t?
-
t is time.. so t0.1 is 90%, t0.5 is 50%.. If you look at the link I supply in the original post, it has a diagram of exactly what is t0.1 and such..
Remember, I'm only interested in time, intensity I will measure with a different setup (actually a camera and light meter).
I don't see how the distance will affect anything other than voltage.. In the extreme it may make the flash look faster or slower, but I won't push it to the extreme, I will move the flash as close as I can without clipping/maxing the voltage output by the photodiode..
Thanks
-
I don't see how the distance will affect anything other than voltage.. In the extreme it may make the flash look faster or slower, but I won't push it to the extreme, I will move the flash as close as I can without clipping/maxing the voltage output by the photodiode..
right! you can move the flash back and forth then until you get the time indicated by the manual ;)
-
> right! you can move the flash back and forth then until you get the time indicated by the manual
haha! I don't think it will be that easy, but without experimentation I can't be sure.. If it did show it to be easy to manipulate, I would use a light meter to make sure the same amount of light always hits the photo diode, much like metering for the camera.. Another option is a constant distance, using a neutral density filter to reduce the intensity of the flash (again probably metered to know how much filtering to use). And, in any case, I would be able to tell something is off by looking at the oscilloscope display, if it was a characteristic pattern for the flash technology used, for example a conventional flash with a tapered fall-off or one controlled with an IGBT (Insulated Gate Bipolar Transistor), which has a sudden drop in voltage.
I think if I made sure the top of the signal was not clipped, but came close, it would capture everything I want to know.. At least I hope!
But you raise a good point.. Thanks!
-
Phototransistor should be fine at these speeds ...
Some phototransistors are really slow, like 20-30 microseconds turn on time. Some are faster, at 2-5 microseconds. A flash lasting 1/10000th of a second duration is only 100 microseconds. So he would have 20-30% accuracy, or 2-5% accuracy if he selected the fastest PT available. Then when he goes to measure a faster flash that is 1/25000ths of a second, or 40 microseconds, the accuracy drops to 50% or worse, and at best 5% with the fastest PT he can get.
I think a photodiode is best. Plus they don't saturate at high light levels.
I just did some calculations, a shoe mounted SLR flash can put out 250,000 to 750,000 lumens. Some higher power ones can put out Mega-lumens. a typical photodiode generates about 400uA in photodiode mode @ 4500 lux. That's lux, not lumens. converting lumens to lux depends on distance and spread (angle) of the light. but lets say your're at 5 meters, and the flash has a spread of 90 degrees. I don't know if these are reasonable, but lets say..
400uA / 4500 lux = 88.88nA / lux. let's derate it, and say 50nA per lux.
at 5m distance from the flash spreading in a cone at 90 degrees, a 250,000 lumens device will produce what lux?
Lux level = total light output (lumens) / area (square meters) => 1 lux = 1 lm/m^2
So in my example, it's the area of the base of a cone with height 5m, spreading at 45 deg angle from the centerline.
the radius of that base is 5m * tan(45) = 5m
the area of that base is pi*r^2 = 3.14*5^2 = 78.54 sq.m. (m^2).
then the lux level at any point on that surface (the base of the cone) is 250000 lm / 78.54 m^2 = 3183 lux
so at 5m, you can expect 3183 lux * 50nA / lux = 159uA.
at 159uA you need a transimpedance amp.
So lets move it to 1m distance from the flash. what now?
1m * tan(45) = 1m. area = 3.14*1 = 3.14 m^2. lux = 250,000 lm / 3.14 m^2 = 79618 lux.
now the current from the diode is going to be 79618 lux * 50nA / lux = 3.98 mA approx 4ma.
4mA across a 1k resistor will drop 4v. that's detectable and can trigger an input to a micro to start a counter.
So you won't need a transimpedance amp, but you will need to be 1m or closer if you don't. Brighter flashes can be farther away.
With a transimpedance amplifier it will be easier to operate and less frustration. You won't need to move the detector nearer or farther from the flash source for different flashes in order to get enough current from the diode.
Feeding the output of the transimpedance amplifier into a comparator will allow you to adjust the trigger level with a setpoint voltage on the other level. The comparator output will be a square wave for the duration of the flash. This can be used to gate a counter, like I said in a previous posting. The rising edge of the comparators output can trigger a counter reset, the actual square wave output will gate a 1Mhz clock into a digital counter.
right! you can move the flash back and forth then until you get the time indicated by the manual ;)
The distance will affect the current output from the diode (and thus voltage ) and this might affect the trigger level.
i.e. at what voltage level do you consider a flash has started ?? when does it end ?? distance will affect that, and thus will vary the starting and ending points of a counter (i.e. the width of the pulse output from the comparator) This is where the oscilloscope has the advantage. You can clearly see the waveshape (the voltage levels don't matter) and you can estimate the starting and ending times.
-
Wow, thanks for the well thought out response CodeBoy.. I'm not sure what to focus on, as you raise many good points..
The distance will affect the current output from the diode (and thus voltage ) and this might affect the trigger level.
i.e. at what voltage level do you consider a flash has started ?? when does it end ?? distance will affect that, and thus will vary the starting and ending points of a counter (i.e. the width of the pulse output from the comparator) This is where the oscilloscope has the advantage. You can clearly see the waveshape (the voltage levels don't matter) and you can estimate the starting and ending times.
This is the direction I'm going. I think it would be easier to figure this out with a scope, and does not preclude building, in the future, what you outlined. I also have to remember that the photo diode will have "dark current" voltage (the flat line for ambient/room light). Changing the distance of the flash to the photo diode will affect the peak voltage, and thus the 50% and 10% times (t0.5 and t0.1) which is based off the highest and lowest (flatline) voltage reading. If I can mitigate this flash to diode distance variable using the transimpedance amp, it still might be worth using, even with a scope, but thinking about it, it seems I am getting a free lunch. How can the transimpedance amp make up for the flash distance factor? If I put the flash far away, I will only see the top of the wave, and the transimpedance amp will only push it up in voltage.
The more I think about it, the more I think I will need to vary the distance to always deliver the same amount of light to the photo diode, no matter what power setting the flash is set to..
Thanks!
-
Wow, thanks for the well thought out response CodeBoy.. I'm not sure what to focus on, as you raise many good points..
The distance will affect the current output from the diode (and thus voltage ) and this might affect the trigger level.
i.e. at what voltage level do you consider a flash has started ?? when does it end ?? distance will affect that, and thus will vary the starting and ending points of a counter (i.e. the width of the pulse output from the comparator) This is where the oscilloscope has the advantage. You can clearly see the waveshape (the voltage levels don't matter) and you can estimate the starting and ending times.
This is the direction I'm going. I think it would be easier to figure this out with a scope, and does not preclude building, in the future, what you outlined. I also have to remember that the photo diode will have "dark current" voltage (the flat line for ambient/room light). Changing the distance of the flash to the photo diode will affect the peak voltage, and thus the 50% and 10% times (t0.5 and t0.1) which is based off the highest and lowest (flatline) voltage reading. If I can mitigate this flash to diode distance variable using the transimpedance amp, it still might be worth using, even with a scope, but thinking about it, it seems I am getting a free lunch. How can the transimpedance amp make up for the flash distance factor? If I put the flash far away, I will only see the top of the wave, and the transimpedance amp will only push it up in voltage.
The more I think about it, the more I think I will need to vary the distance to always deliver the same amount of light to the photo diode, no matter what power setting the flash is set to..
Thanks!
You're welcome.
The dark current is literally dark current. i.e. if the diode had absolutely no light hitting it at all. It's different from the ambient current you might get in a room.
With a scope you can measure the ambient current (once converted to a voltage) and that will form your baseline. The flash will still show it's waveshape above the baseline.
The same applies to a counter based system. Setup the detector in a room and set the trigger threshold to be above the ambient voltage.
The transimpedance amplifier will not compensate for light that doesn't reach the detector, or for too little light reaching the detector. The full flash from start to finish always has to reach the detector. You can guarantee this by keeping the detector close enough to the flash. The amount of light reaching the detector won't matter if all you want to know is the flash duration. What matters is that enough of the light reaches the detector, and you can make sure it does by keeping the detector close to the flash, not farther away.
Transimpedance amplifiers can be tricky to get right, and now that I think more about this, and the light levels involved, you probably don't need to complicate it. You can get away with reverse biasing the diode to 5v and sinking the current through a 10k resister and amplifying that voltage drop across the resistor.
100uA through a 10k will give you 1V. Amplify that with a standard non-inverting voltage amplifier of gain=5 and you can get 5V. If the light only gives you 50uA, you still get 2.5V out. at 10uA you get only 500mv, that's too little output. If the output might be that low, then change the gain to 7 or 10. You can still feed this to a scope or counter.
-
Took a white led and connected a 1k resistor across it. Flashed my camera about 2" (5 cm for you Dave) and measured a 1.5V spike about 25 usec long using my camera as a flash. Rise time was about 2 usec. Most any LED should work. For set up, mainly getting the scope to single trigger correctly, I used a flashlight with a much slower sweep speed on the scope.
-
100uA through a 10k will give you 1V. Amplify that with a standard non-inverting voltage amplifier of gain=5 and you can get 5V. If the light only gives you 50uA, you still get 2.5V out. at 10uA you get only 500mv, that's too little output. If the output might be that low, then change the gain to 7 or 10. You can still feed this to a scope or counter.
Won't the amplifier be slow? I looked on mouser for an non-inverting amplifier (op-amp), and what i saw has somewhat slow timings.. It also looked like it would be fun to hook up..
Took a white led and connected a 1k resistor across it. Flashed my camera about 2" (5 cm for you Dave) and measured a 1.5V spike about 25 usec long using my camera as a flash. Rise time was about 2 usec. Most any LED should work. For set up, mainly getting the scope to single trigger correctly, I used a flashlight with a much slower sweep speed on the scope.
Wow, you have a really fast flash.. If my calculation is correct that's 1/40,000 of a second.. Only a few top end flashes are that fast..
-
Won't the amplifier be slow? I looked on mouser for an non-inverting amplifier (op-amp), and what i saw has somewhat slow timings.. It also looked like it would be fun to hook up..
If you are going to buy a scope (and I think you need one to properly asses flash duration) you won't need any amplifier. Just get a photodiode with a few sq mm area and load it with 50 ohms. You could put it on the end of some 50 ohm coax and use a 50 ohm BNC terminator and T piece at the scope end.
You will get plenty of signal to look at if the flash is within a few feet. It is possible to saturate photodiodes. I doubt it will be a problem but it would be easy enough to check by changing the flash to diode distance and checking the pulse amplitude varies as expected.
On 25us being fast - a cheap flash has to be that fast in auto exposure mode when you have a close target.
On the diode maybe a BPW34 which is 7.5mm^2 and dirt cheap. Spectral sensitivity might be an issue? Don't know what the spectrum of a flash is and I imagine you want a spectral response similar to the camera. You can get BWP34 variations with integrated IR filters.
-
100uA through a 10k will give you 1V. Amplify that with a standard non-inverting voltage amplifier of gain=5 and you can get 5V. If the light only gives you 50uA, you still get 2.5V out. at 10uA you get only 500mv, that's too little output. If the output might be that low, then change the gain to 7 or 10. You can still feed this to a scope or counter.
Won't the amplifier be slow? I looked on mouser for an non-inverting amplifier (op-amp), and what i saw has somewhat slow timings.. It also looked like it would be fun to hook up..
Yes, you need a fast amp; 10V/us or better will give you a fast risetime at the edges of the pulse.
-
I did some searching and found a full writeup of photo diodes (jpg attachment and full document link below)..
I think I can hook it up, but I am not sure what values to use for C1 and RS, and why I need the 50ohm Coax. The formulas will help I suppose.. And how do you terminate the coax with a 50ohm resistor? Also the diagram shows an oscilloscope with internal 50 ohm resistance, what if mine is not 50ohm?
RL * C1 > T (where T is the pulse width in seconds) (But what is RL ???? )
RS = V bias / 10 Ipdc where Ipdc is the DC photo current (photo current coming from the photodiode I assume)
Bandwidth = 0.35 / T (Why do I need this?)
Here is the full document.. Attached is a small snip..
http://www.osioptoelectronics.no/application-notes/AN-Photodiode-Parameters-Characteristics.pdf (http://www.osioptoelectronics.no/application-notes/AN-Photodiode-Parameters-Characteristics.pdf)
-
but his results seem off, they seem to be 30% below the the manufacturer specs for the few flashes and strobes I checked. In fact many internet sources have different results.. In this particular case, I'm wondering if it was the fact that he used a photo transistor instead of a faster photo diode. (the photo transistor he used has a higher turn on (rise) time than he states in his blog)
Hello,
I'm the guy that wrote the article in question. I frequent this forum and only just come across this :)
A photo diode will give you must faster response time, and this photo transistor is all I had laying around at the time. Putting things back into reality, it is a Vishay BPV11 which has a response time of around 5-6us, one delay for turn on, one delay for turn off. In my measurements, this response time, on the fastest and shortest durations i was measuring, it results in an error maybe up to 5%.
I just fixed the error about the response time in my blog, I wrote ns instead of us (although i did know it was us when i was doing the tests).
Remember in photography, lighting is dealt with using reciprocity laws, and 5% error (if it is even that much) equates to less than 0.05 stops.
On another note, the Rigol scope in question is a nice cheap scope, perfect for the hobbyist and even for pros. Worth the money with a lot more uses in the future after doing your flash measurements :)
Andy
-
Thanks Andy..
One thing I noticed about your times when comparing to the Paul Buff times for Einstein was that after a certain point , your measurements started to diverge from the ones the manufacturer reported.. For example, the Einstein times (as reported in the manual), vs yours:
| Power | Manufacture | Yours |
| FULL | 1/568 | 1/379 |
| 1/2 | 1/2041 | 1/1603 |
| 1/4 | 1/3514 | 1/3378 |
| 1/8 | 1/6050 | 1/6173 |
| 1/16 | 1/10417 | 1/7622 |
| 1/32 | 1/11050 | 1/8562 |
| 1/64 | 1/11765 | 1/7530 |
| 1/128 | 1/12579 | 1/7764 |
| 1/256 | 1/13514 | 1/7267 |
The first two and the last 5 seem off. I wouldn't be surprised that your first four might be accurate, but it seems after 1/8, your measurements really diverge.. 1/4 and 1/8, i think you nailed the times..
I am just curious, and I hope to be making my own measurements soon. I have a few flashes from over the years i hope to test, and eventually might be getting an Einstein unit myself.. So this is a mix of electronics coolness using an oscilloscope and actually having a use for the numbers, i hope to do some high speed flash photography in the future..
Please don't take this the wrong way, I'm not trying to knock your contribution, more like, I am trying to understand what is going on.. I think eventually I may contact the PaulBuff company and ask how they measure it, they seem the most likely company that would offer the testing setup..
-
can anyone recommend an off-the-shelf probe or box with bnc that has a photo-dide, battery for bias or active probe?
usb-solutions also welcome
-
Photo diode is current generator, it needs load resistor instead. Take any LED, put something like 0.1 .. 1k in parallel, connect scope @ 1V range. You shall see some voltage appearing when you strobe photoflash light into LED from close (30cm) distance.
-
If you use an LED you should not use LED with a phosphor layer. The decay of the phosphor is slower than the response of the LED itself. So white LEDs are a poor choice and some other colors seem to use a phosphor as well.
-
We did some strobe optical PW measurements awhile back using a diode and LED over at Photomacrography.
https://www.photomacrography.net/forum/viewtopic.php?f=25&t=39748&hilit=photodiode&start=15 (https://www.photomacrography.net/forum/viewtopic.php?f=25&t=39748&hilit=photodiode&start=15)
If you do a search at PM you'll find we did a bunch of these type measurements some time ago to find out what the actual optical waveform "looked like" and what the effective pulse width was for various speed lights and studio strobes.
Best,
-
This was the first thing I thought, there are so many variables here before you even consider the electronics. It is very likely that this is impacting much of the online data and testing the original poster has seen. One thing that comes to mind is filtering of the light to the visible wavelengths.
No, a photodiode is all you need for measuring the flash waveform
light measurement is quite ambigouos to me though. say if you want to measure ambient light, or projected/incident/focused light. from my experimentation with simple photodiode+speaker cable+recorded to PC software, the closer you put the diode, the more saturated it will be, i have to filter the incoming light with darker plastic film to get nicely nonsaturated signal. so it depend on the distance from light to sensor. we (i) need more definitive standard here.
-
A BPW85 phototransistor is cheap (cheaper than a typical photodiode) and has Tr and Tf <3uS. A typical photo-flash is in the mS. Messing about with transcondcuctance amps is is an uneccessary expense and they're prone to oscillation unless you match the diode and the amp circuit with great care. Faff.
https://www.vishay.com/doc?81531 (https://www.vishay.com/doc?81531)