Flash Durations - your opinion on these please. UPDATED

[Andy Taylor]

I found this excellent article which explains a lot more about the difference between different types of flash. on the second page, the author also finds that the manufacturer's figures are somewhat optimistic.

http://www.scantips.com/speed.html

 

[HoppyUK]

I found this excellent article which explains a lot more about the difference between different types of flash. on the second page, the author also finds that the manufacturer's figures are somewhat optimistic.

http://www.scantips.com/speed.html

Thanks Andy, interesting link

 

[HoppyUK]

UPDATE (and nerd alert)

A quick update, and to let the many people that kindly contributed to this thread know their time hasn't been entirely wasted I now have an oscilloscope and got everything working properly today. It's already opened my eyes to a few things with a lot more to come. I was recently commissioned by a photography website to test some flash guns so it'll be very useful there. Thanks to Gareth (Kaolin) for suggesting a Pico PC-based oscilloscope (high performance at affordable prices) and others that posted links to suitable photo-diode circuits. It seems there's nothing off the shelf to do this, so I built a little custom sensor and it works brilliantly. I chose a photo-diode with a spectral response very close to daylight and fast reaction time - like 1.5ns (that's 1.5 billionths of a second).

1) Elinchrom D-Lite One at full power. T.5 time is 1/2200sec, exactly as claimed, and...

2) ...at minimum power it still reads 1/1300sec. Flash durations with conventional studio heads tend to get much longer than this as power is reduced but I noticed in real-world testing that the D-Lite One was unusually fast at minimum power, and here's confirmation - it's smaller capacitors work faster (only 100Ws)

3) Yongnuo 600EX-RT in high-speed sync mode, showing how the light pulses to create effectively continuous light for the duration of the shutter cycle, giving even exposure down the frame. Total burn time is 1/112sec, firing in clusters of mini-pulses. It's not so visible here, but when you zoom in to examine the clusters, it's firing ten pulses each time, at 54,000 cycles per sec. I have never seen the very close detail of this before today. Another feature is the spikes at the beginning, that explain why the flash is actually triggered slightly before the shutter opens in HSS mode, to ensure brightness has settled to an even level right from the start.
CORRECTION: I had the sampling rate set too low, and that has created aliasing effects resulting in those little clusters of ten that don't actually exist! The flash is in fact firing at a continuous rate of 54,000 cycles per sec throughout, at constant brightness.

1)


2)


3) See correction in comment above

 

[HoppyUK]

More nerdy stuff

Canon 580EX firing in optical Master mode. On the left, there's a cluster of seven command pulses (clearly visible when I zoom in) followed by the brighter pre-flash used for exposure measurement. Then there's a delay of 1/180sec where presumably the camera is processing the pre-flash feedback received from remote slave guns. Then there's another cluster of nine more command pulses that tell the remote slave guns what to do, followed the main flash on the right that goes off the top of the scale.

So that's a total of 18 flashes from the master gun, over a period of 1/130sec from start to finish.

Edit: there are actually no remote slave guns firing in this example.

 

[sk66]

So that's a total of 18 flashes from the master gun, over a period of 1/130sec from start to finish.

Rear curtain sync? Would it have been quicker if there were remote flashes to respond?

 

[HoppyUK]

Rear curtain sync? Would it have been quicker if there were remote flashes to respond?

Not significantly - the major delay seems to be the camera processing time. And the native Canon system doesn't do remote rear-curtain sync, even the RT guns (though it's available in third party radio controlled alternatives).

 

[MidnightUK]

Actually I've just done some shots (for a totally different purpose) where the model was tossing her head and her very long hair was flying. I used an "ordinary" (not IGBT) flash head for that and you know what? I like the blur.

Can you obtain a similar blur by dragging the shutter?

 

[Garry Edwards]

I feel that getting this oscilloscope setup is a major step forward for you, and all credit to you for doing it, I know just how complicated it is to get it set up to produce meaningful and reliable data.
Of course, the downside of being able to produce this kind of data is that the real nerds then use that info to delude themselves that they *NEED* the fastest and most expensive toys and that they think that gear is more important than either knowledge or technique, but that won't be your fault.

 

[Phil V]

Can you obtain a similar blur by dragging the shutter?

Would depend on the balance with the ambient.

 

[HoppyUK]

I feel that getting this oscilloscope setup is a major step forward for you, and all credit to you for doing it, I know just how complicated it is to get it set up to produce meaningful and reliable data.
Of course, the downside of being able to produce this kind of data is that the real nerds then use that info to delude themselves that they *NEED* the fastest and most expensive toys and that they think that gear is more important than either knowledge or technique, but that won't be your fault.

Cheers Garry. Yes, it's a cool little device. It's better than I hoped, and a lot cheaper than I feared - about £120 for the Pico oscilloscope and £30 worth of bits to make the sensor. Pico were very helpful speccing key components, and after I discovered they're only half an hour down the road from me, I went to make sure I'd got everything properly optimised. It's actually very easy. I think they're going to do something similar for their training blog, as it's an interesting way to demo certain aspects.

It's so easy to collect tons of quality data in very little time and it will be handy for my work for sure. The danger is getting drawn into the technicalities and loosing sight of the relevance, but it's a great way to cross-check and confirm things. The detail it'll go in to is astounding. Want to know the delay between triggering an optical slave and the flash firing? I always thought that was instant, but it's not and I can now measure it to the nanosecond

 

[sk66]

It's better than I hoped, and a lot cheaper than I feared - about £120 for the Pico oscilloscope and £30 worth of bits to make the sensor. Pico were very helpful speccing key components, and after I discovered they're only half an hour down the road from me, I went to make sure I'd got everything properly optimised.

Excellent!
If empirical test data were included with all reviews maybe some of the manufacturers would become a little less free with their claims...

 

[HoppyUK]

Excellent!
If empirical test data were included with all reviews maybe some of the manufacturers would become a little less free with their claims...

Thanks Steven

Now to come up with a formula to properly convert t.5 times into real shutter speed equivalents that works with all flash types at all power levels. I have an idea for that - something along the lines of t.5 multiplied by the graph area of brightness above 50% vs below 50%. That should work, as it's the brightness below 50% that creates the blurring, but getting the levels and ratios right will take a lot of trial and error

 

[sk66]

Thanks Steven

Now to come up with a formula to properly convert t.5 times into real shutter speed equivalents that works with all flash types at all power levels. I have an idea for that - something along the lines of t.5 multiplied by the graph area of brightness above 50% vs below 50%. That should work, as it's the brightness below 50% that creates the blurring, but getting the levels and ratios right will take a lot of trial and error

I'm not understanding... t.1 is the effective SS irregardless of type/power (and it appears to be close to 2ms, 1/500 at both min/max for the RX 1)

 

[HoppyUK]

I'm not understanding... t.1 is the effective SS irregardless of type/power (and it appears to be close to 2ms, 1/500 at both min/max for the RX 1)

That's the rule of thumb, or 3x t.5, and if it turns out to be that simple I'll be very happy. I have no wish to reinvent the wheel. But very few manufacturers publish t.1 times so there's no way of knowing.

 

[Faldrax]

Of course, the downside of being able to produce this kind of data is that the real nerds then use that info to delude themselves that they *NEED* the fastest and most expensive toys ...

No, REAL nerds know they don't NEED the new toys, but WANT them anyway

 

[sk66]

That's the rule of thumb, or 3x t.5, and if it turns out to be that simple I'll be very happy. I have no wish to reinvent the wheel. But very few manufacturers publish t.1 times so there's no way of knowing.

You are directly measuring it (t.1) just as you are measuring the t.5 times (that's what the vertical dashed lines represent right?)... 10% is pretty easy to calculate

AFAIK t.1 is the accepted/industry measure, many don't specify it (or lie) and the ROT 3x may not be accurate, but I wouldn't try to convolute things (unless you come up with something new/entirely better). In actual use t.1 time may not be the only determining factor if that's what you are getting at, the power at t.1 is also relevant... but IMO that is more a function of ambient light levels/exposure settings and not a flash characteristic per se (i.e. power and speed are "flash characteristics," but the combination relative to ambient is "situational" and highly variable).

 

[HoppyUK]

You are directly measuring it (t.1) just as you are measuring the t.5 times (that's what the vertical dashed lines represent right?)... 10% is pretty easy to calculate

AFAIK t.1 is the accepted/industry measure, many don't specify it (or lie) and the ROT 3x may not be accurate, but I wouldn't try to convolute things (unless you come up with something new/entirely better). In actual use t.1 time may not be the only determining factor if that's what you are getting at, the power at t.1 is also relevant... but IMO that is more a function of ambient light levels/exposure settings and not a flash characteristic per se (i.e. power and speed are "flash characteristics," but the combination relative to ambient is "situational" and highly variable).

We're on the same page Steven. If t.1 works out to be decently accurate in all situations, then happy days. I don't know yet, simply due to the absence of published t.1 data to compare to my real-world findings based on actual shutter speeds. What I do know is that the t.5 x3 = t.1 conversion is not good enough. With a lot of flash units it ends up under-stating t.1 at one end of the power range, and over-stating at the other. But if a true and accurate t.1, measured as you say, takes away that conversion error then that'll be sweet.

Yes, there will be other factors that skew things according to some individual flash unit characteristics, but I am not looking for absolute accuracy (impossible anyway, as the t.x formula is only a simulation). A +/-10% error would be hard to detect visually, even +/-20% would probably be okay in practical terms, eg if 1/1000sec turns out to be 1/800sec or 1/1200sec, then I don't think there would be too many complaints

Edit: For the benefit of others - t.5 is the length of time the flash stays above 50% of peak brightness. In the first two graphs posted above, t.5 is shown by the vertical dotted lines. T.5 is the figure quoted by almost all flash manufacturers, but is often misleading as it substantially over-states the action-stopping potential of the flash, compared to actual shutter speeds, typically by 2x or 3x. In other words, a quoted t.5 flash duration of 1/1000sec freezes movement like an actual shutter speed of 1/330 to 1/500sec.
T.1 is the length of time the flash stays above 10% of peak brightness, and is widely regarded as a much closer approximation of actual shutter speeds, but very few flash manufacturers give it.

 

[sk66]

IDK... thinking about how to come up with an empirical measure more accurate than the accepted t.1 seems to get rather convoluted to me (or I'm overcomplicating things).

I think you would have to work in a blacked out environment w/ "bulb" settings and not rely on "dark frame" exposure settings. That's because an ambient level just below recording/display is not quite the same as "no light" reaching the sensor. And you would need some kind of "neutral target" because light colors overpower/blur more and blacks remain black no matter how much light they receive (essentially). Then you would have to do black BG comparison shots at various SS's and make an educated guess/comparison for SS equivalence... I suppose you could align images in PS and set the mode to "difference" in order to get a fairly accurate measure/comparison? And then there will be ramp up curve (leading blur) and tail curve (trailing blur) variables that will only be accountable for "subjectively."

Even if you do all of that and determine that t.1 is not accurate, you will have to set/establish/defend/etc your "new standard" which I think is a bit of a moot point... I don't think t.1 time is entirely accurate anyway (i.e you could record enough t.05 tail so that it adds up to some t.1 portion). And I don't think it needs to be any more accurate as there are other variables that might push the results around quite a bit in actual application (like how water drops can record as moving backwards/up). Unless you do find that t.1 is entirely inaccurate by a large margin... but then I doubt it would be the accepted standard (then again, I spent almost my entire life "knowing" that the brontosaurus existed).


Having said all that, if I were in your place I would probably chase that rabbit until exhausted...

 

[HoppyUK]

IDK... thinking about how to come up with an empirical measure more accurate than the accepted t.1 seems to get rather convoluted to me (or I'm overcomplicating things).

I think you're right. The idea I have in mind, that I briefly mentioned above and I'm pretty sure would work, is far too complicated TBH. I just need to try a few things, when I find some time.

I think you would have to work in a blacked out environment w/ "bulb" settings and not rely on "dark frame" exposure settings. That's because an ambient level just below recording/display is not quite the same as "no light" reaching the sensor. And you would need some kind of "neutral target" because light colors overpower/blur more and blacks remain black no matter how much light they receive (essentially). Then you would have to do black BG comparison shots at various SS's and make an educated guess/comparison for SS equivalence... I suppose you could align images in PS and set the mode to "difference" in order to get a fairly accurate measure/comparison? And then there will be ramp up curve (leading blur) and tail curve (trailing blur) variables that will only be accountable for "subjectively."

I have a good set up for getting comparison images at actual shutter speeds. The problem is basically with voltage-regulated studio heads (rather than IGBT) and the interpretation of results. That's what this thread is about really, asking what other people think of my estimates. Luckily they seem to tie in well, but it's very subjective and there's no way round that - when does an image change from sharp to unsharp? At what point? Results are also affected by exposure levels and contrast settings that both impact on the visual impression.

Even if you do all of that and determine that t.1 is not accurate, you will have to set/establish/defend/etc your "new standard" which I think is a bit of a moot point... I don't think t.1 time is entirely accurate anyway (i.e you could record enough t.05 tail so that it adds up to some t.1 portion). And I don't think it needs to be any more accurate as there are other variables that might push the results around quite a bit in actual application (like how water drops can record as moving backwards/up). Unless you do find that t.1 is entirely inaccurate by a large margin... but then I doubt it would be the accepted standard (then again, I spent almost my entire life "knowing" that the brontosaurus existed).

I don't know what you mean - I saw a brontosaurus in the garden only the other day. I think my mission here, if that's the right word, is to examine t.1 and see how well it fits. My hunch is that it won't be too far out, in the main. In which case, I can now measure that very easily This is the big change, as all I've had previously is my actual shutter speed tests and claimed t.5 times that are clearly miles apart.

Having said all that, if I were in your place I would probably chase that rabbit until exhausted...

Haha

 

[mike weeks]

The basic 3x to go from t.5 to t.1 can be explained by the standard RC discharge circuit, a quick read of http://www.electronics-tutorials.ws/rc/rc_2.html shows that we drop to 50% after approx 0.7 time constant and 10% after approx 2.1 - 2.3 time constants (based on a perfect capacitor which does not exists) - I did this at O level 36 years ago and nothing has changed. What the theoretical model does not include is that there is a short delay to get to 100% as indicated in Richards traces and that capacitors have leaks which means they do fully discharge whereas the theoretical model says they generally will not.

Mike

 

[mike weeks]

3)
View attachment 90001

Most interesting diagram as it suggests the system is working at absolute maximum as the charge is decreasing over time, would be interesting to compare to some other brands

would also be interesting to know if you can compare voltages with a standard pulse

Mike

 

[sk66]

The basic 3x to go from t.5 to t.1 can be explained by the standard RC discharge circuit

And there is no remotely accurate conversion for IGBT heads/speedlights. Plus the measures I have seen are not even close to linear... i.e. t.1 might be around 3x at max power, 1x in the middle ranges, and 2x towards min power. The only way to get reliable numbers for them is direct measurement.

 

[HoppyUK]

Most interesting diagram as it suggests the system is working at absolute maximum as the charge is decreasing over time, would be interesting to compare to some other brands

would also be interesting to know if you can compare voltages with a standard pulse

Mike

Yes, should be easy enough - voltage is up the Y-axis and the photo-diode's response is linear. Just have to make sure the test scenario is identical as voltage/brightness is very sensitive to distance (obviously enough).

I'd be interested to see how your various Godox flash units measure up, particularly as I don't have a really powerful IGBT flash unit to hand ATM

 

[mike weeks]

@HoppyUK another good thing would be looking at the difference between low and high power on a standard IGBT flash compared to a traditional studio flash i.e. we should see the IGBT cut off and the increase in time for the studio flash.

Mike

 

[Cistron]

For the benefit of others - t.5 is the length of time the flash stays above 50% of peak brightness. In the first two graphs posted above, t.5 is shown by the vertical dotted lines. T.5 is the figure quoted by almost all flash manufacturers, but is often misleading as it substantially over-states the action-stopping potential of the flash, compared to actual shutter speeds, typically by 2x or 3x. In other words, a quoted t.5 flash duration of 1/1000sec freezes movement like an actual shutter speed of 1/330 to 1/500sec.
T.1 is the length of time the flash stays above 10% of peak brightness, and is widely regarded as a much closer approximation of actual shutter speeds, but very few flash manufacturers give it.

Wouldn't the area under the curve be way more useful, e.g. when was 80, or 90% of the light dumped? Just timing t.1 or t.5 times really doesn't tell you that much if the tail end still dumps a lot of energy. Unless I'm mistaken and all waveforms follow them same kind of pattern.

 

[mike weeks]

Wouldn't the area under the curve be way more useful, e.g. when was 80, or 90% of the light dumped? Just timing t.1 or t.5 times really doesn't tell you that much if the tail end still dumps a lot of energy. Unless I'm mistaken and all waveforms follow them same kind of pattern.

Except that is not the industry standard measure

Mike

 

[HoppyUK]

Wouldn't the area under the curve be way more useful, e.g. when was 80, or 90% of the light dumped? Just timing t.1 or t.5 times really doesn't tell you that much if the tail end still dumps a lot of energy. Unless I'm mistaken and all waveforms follow them same kind of pattern.

Yes, it's something along those lines that I was alluding to earlier, but it's not an easy calculation to do accurately. The part of the flash that causes blurring is, predominantly, the triangular area of the tail to the right of the t.5 cut-off and below 50%. That area needs to be measured as a proportion of the total, and also weighted. Eg, the area below 10% might add up to quite a lot but is also spread over a much longer time so it would hardly register. Then the area immediately to the right of the t.5 cut-off is contributing quite a lot of light and over a much shorter time, so that will show as quite significant blurring.

That's just way too complicated, regardless of merit. There must be an easier way that is accurate enough in practical terms. I'm going to start with t.1 and see how that compares to actual shutter speeds.

 

[sk66]

Wouldn't the area under the curve be way more useful, e.g. when was 80, or 90% of the light dumped?

T.1 *is* effectively 90% of the light dumped... there is no "above/below" in the measure, there is "before/after" (i.e. "the area" is all of it between the vertical dashed lines). The only thing it doesn't tell you is "total power," which is somewhat irrelevant to the measure IMO (going back to speed and power being separate characteristics).


Edit: We are talking about times so fast that "intensity" far outweighs duration/time. I suppose it is possible to have a very long tail at/above 10% to where duration becomes more of a factor, but then you would probably have to account for shutter speed as "typical use" is not black environment/bulb exposures... and I still think it would typically be of minimal difference.

 

[HoppyUK]

Wouldn't the area under the curve be way more useful, e.g. when was 80, or 90% of the light dumped? Just timing t.1 or t.5 times really doesn't tell you that much if the tail end still dumps a lot of energy. Unless I'm mistaken and all waveforms follow them same kind of pattern.

There is merit in this. It's simple and logical, and from the little basic research I've done so far, it kind of stacks up. It's hard to measure though, time consuming, at least with the gear I have to hand. And I think the significant percentage is more like 70% than 90%... Needs more work

 

[Cistron]

There is merit in this. It's simple and logical, and from the little basic research I've done so far, it kind of stacks up. It's hard to measure though, time consuming, at least with the gear I have to hand. And I think the significant percentage is more like 70% than 90%... Needs more work

If you get RAW data output from your oscilloscope you might be able to calculate the 70% timings with a few integral equations in R (don't ask me how to, it would need some digging in my highschool maths notes).

edit: damn, integrals, not differentials, duh!

 

[sk66]

And I think the significant percentage is more like 70% than 90%... Needs more work

I think that will depend on how much ambient is in the mix... In a black room/bulb exposure lower levels will be more evident. That's my thinking anyway...

 

[mike weeks]

Wouldn't the area under the curve be way more useful, e.g. when was 80, or 90% of the light dumped? Just timing t.1 or t.5 times really doesn't tell you that much if the tail end still dumps a lot of energy. Unless I'm mistaken and all waveforms follow them same kind of pattern.

Essentially this is what they did with the Elinchrom HS heads according to some, they flattened the peak and extended the tail

Mike

 

[HoppyUK]

I think that will depend on how much ambient is in the mix... In a black room/bulb exposure lower levels will be more evident. That's my thinking anyway...

All tests done in a darkened studio. Any significant ambient would invalidate everything.

 

[HoppyUK]

Correction -

The graphs I put up in post #82, there's a slight error in the third one showing high speed sync.

I had the sampling rate set too low, and that has created aliasing effects resulting in those little clusters of ten that don't actually exist! The flash is in fact firing at a continuous rate of 54,000 cycles per sec throughout, at constant brightness - no clusters!

I've added a correction note to that post.

 

[mike weeks]

Correction -

The graphs I put up in post #82, there's a slight error in the third one showing high speed sync.

I had the sampling rate set too low, and that has created aliasing effects resulting in those little clusters of ten that don't actually exist! The flash is in fact firing at a continuous rate of 54,000 cycles per sec throughout, at constant brightness - no clusters!

I've added a correction note to that post.

Graph to follow?

Mike

 

[mike weeks]

Had the thinking cap on last night and of course these do not show some info that would show how much of the tail is captured. As we all know the flash does not go off when the shutter button is pressed but slightly later to ensure both curtains are fully open, depending on the length of the flash duration and the length of time the shutter is open it might be a full or partial capture of the flash pulse. Pocket wizard exploited this with what they called Hypersync http://wiki.pocketwizard.com/index.php?title=HyperSync_Setup which can be thought of as Peak sync (the info seems to have disappeared from their wiki pages) or Tail Sync - interesting stuff

Mike

 

[HoppyUK]

Graph to follow?

Mike

Here you go Zooming in on the detail, it's clearly strobing at 54,000hz. Interestingly, a Canon 580EX strobes at 40,000Hz but has slightly faster flash durations at same power settings, more so at lower power.

It seems that TP won't allow me to upload the full data graphs (including all settings and notes etc) but only JPEGs like this. I need to convert them I guess.

 

[HoppyUK]

Had the thinking cap on last night and of course these do not show some info that would show how much of the tail is captured. As we all know the flash does not go off when the shutter button is pressed but slightly later to ensure both curtains are fully open, depending on the length of the flash duration and the length of time the shutter is open it might be a full or partial capture of the flash pulse. Pocket wizard exploited this with what they called Hypersync http://wiki.pocketwizard.com/index.php?title=HyperSync_Setup which can be thought of as Peak sync (the info seems to have disappeared from their wiki pages) or Tail Sync - interesting stuff

Mike

There's no easy way of overlaying these graphs with the actual shutter opening, but there are some slo-mo videos on the web that also include approximate shutter timings. The rule of thumb I use is about 3ms for one shutter blind to run from top to bottom though of course that varies. Some of the faster shutters run at more like 2.5ms, and slower ones as in the Canon 5D Mk2 video below are nearer 3.5ms. I suspect lower spec cameras with 1/180sec x-sync speed will be more like 4-4.5ms. That's a significant difference but I can't find much really accurate data on many cameras.

Yes, Pocket Wizard's optimised HHS feature adjusts itself to maximise brightness when customised to individual camera models. The really clever trick* is PW takes the wasted power (at faster speeds, after the second blind has closed) and pushes it back into the active brightness zone. It's a shame this doesn't seem to have caught on as with some cameras (notably the faster ones like Canon 1D4) it can increase effective brightness by over one stop. On the other hand, it could be this feature that caused so many Canon 580EX guns to burn out expensively. I believe Canon made the 600EX-RT a bit more durable in this department, and the new 600EX-RT Mk2 takes it a step further with an uprated heat-sink. HSS really gives the flash tube a hammering.

In the graph posted above, shutter speed is 1/320sec and the total burn time about 9ms. Working to the lowest common denominator and a shutter blind time of say 4ms, plus another 5ms for 1/200sec shutter speed, that's about right. As shutter speed is raised, the burn time reduces slightly to about 8ms at 1/8000sec, but brightness does not change. And Mike, I've also done an identical voltage comparison between normal sync and HSS at full power, and HSS runs at a constant voltage/brightness of exactly 10% of peak normal sync. My Yongnuo 600EX-RT and Canon 580EX guns are both the same in this respect. Measuring the 'area under the graph' HSS covers 70-80% of the normal x-sync area.

*PW's optimised HSS feature only increases brightness with Canon. With Nikons, the unused power is recycled as a faster recharge time, but not extra brightness.

Edit: PW has four modified flash sync timing options. I'm describing Optimised HSS above. Then there's Peak Hypersync that adjusts normal x-sync timing by camera model and squeezes a slighly faster speed with no loss of brightness, eg Canon 7D can go from 1/250sec to 1/400sec, 1D4 can reach 1/500sec. Then there's Tail Hypersync which is the faux HSS mode, and optimised second-curtain sync that takes up the slack in native timings to get the flash firing as close to the second curtain closing as possible. From memory, I think that's right.

View: https://www.youtube.com/watch?v=ptfSW4eW25g