you left out a key piece of the puzzle though: what distance were you shooting at? try again further away or closer perhaps.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
What the green square can teach us about underexposure...
- Thread starter james-montana
- Start date
]
But with all things staying pretty constant (including my distance from the subject) shouldn't the Program mode have compensated, say by altering the shutter speed or apeture to compensate for the change in simulated film sensitivity?
Doug Kerr
Forum Pro
Hi, Gang,
Today I ran a series of tests to investigate the performance of the flash metering system of the 300D, as applied to the onboard flash, with respect to changes in ISO sensitivity.
For the test I used a flat gray target, which was arranged (by change of focal length) to always fill the frame. The target supposedly has an reflectance of about 18% (although this was not verified by measurement. A small black crosshair was provided in the center as a focus aid. The camera was always aimed at this crosshair.
All shots in this series were taken in M mode at 1/125 seconds and f/5.6. All focus points were active and FEL was not used.
A set of shots was taken for each of seven distances, flash-to-target, ranging from 24" to 96" in 12" steps. For each disatnce, shots were taken at ISO settings of 100, 200, 400, 800, and 1600. At least 10 seconds was allowed between each shot to be sure that the flash dould fully recharge.
The histograms of the resulting shots were all examined in VuePrint, and the location of the histogram peak recorded. Those locations were converted, in several stages, into terms of stops down from the top of the luminance scale ("255"). Then. for each set, the luminance value for the exposure at ISO 100 was used as the reference, and the differece in luminance from that(in stops) determined for each of teh other ISO values.
The ambnient environment was such that the metering system would recommend 1.3 second at F/5.6 at ISO 100, 1/12 second at ISO 1600. Thus, with a shutter speed of 1/125 in the flash test shots, the ambient component should be negligible
Those results are summarized here:
Distance: 24"
ISO Difference in stops
100 (ref)
200 -0.1
400 -0.1
800 +1.1
1600 +2.4
Distance: 36"
ISO Difference in stops
100 (ref)
200 +0.4
400 +0.2
800 +0.6
1600 +1.5
Distance: 48"
ISO Difference in stops
100 (ref)
200 +0.2
400 +0.8
800 +0.8
1600 +0.8
Distance: 60"
ISO Difference in stops
100 no data
200 (reference for this set)
400 0.0
800 -0.1
1600 0.0
Distance: 72"
ISO Difference in stops
100 (ref)
200 0.0
400 +0.2
800 +0.4
1600 no data
Distance: 84"
ISO Difference in stops
100 (ref)
200 -0.5
400 -0.3
800 -0.4
1600 +0.1
Distance: 96"
ISO Difference in stops
100 (ref)
200 0.0
400 +0.1
800 +0.3
1600 +0.9
It should be noted that there was ample opportunity for experimental error in this series of tests. Thus care should be taken in drawing inferences from small differences between the results for different ISO values. Additionally, any given shot may have a signficant error.
I make no claim to undestand the signficance of these results, other than clearly the flash metering system is more consistet in its behavior over a certain modest range of distances.
It appears that shutter speed has some influence in this matter, even less well understood. I discovered this when, at one point, I indavertently shifted the shutter speed from 1/125 to 1/160 rather than bumping the ISO setting.Having done so, I noted that the peak of the histogram was at a slightly higher luminance with 1/160 seconond than at 1/125 second, no other exposure factors having been changed. Go figure! I plan to run some tests tomorrow to explore the sensitivity of flash "exposure" to shutter speed cahnges under the 1/200 sec FP threshhold.
Best regards,
Doug
Today I ran a series of tests to investigate the performance of the flash metering system of the 300D, as applied to the onboard flash, with respect to changes in ISO sensitivity.
For the test I used a flat gray target, which was arranged (by change of focal length) to always fill the frame. The target supposedly has an reflectance of about 18% (although this was not verified by measurement. A small black crosshair was provided in the center as a focus aid. The camera was always aimed at this crosshair.
All shots in this series were taken in M mode at 1/125 seconds and f/5.6. All focus points were active and FEL was not used.
A set of shots was taken for each of seven distances, flash-to-target, ranging from 24" to 96" in 12" steps. For each disatnce, shots were taken at ISO settings of 100, 200, 400, 800, and 1600. At least 10 seconds was allowed between each shot to be sure that the flash dould fully recharge.
The histograms of the resulting shots were all examined in VuePrint, and the location of the histogram peak recorded. Those locations were converted, in several stages, into terms of stops down from the top of the luminance scale ("255"). Then. for each set, the luminance value for the exposure at ISO 100 was used as the reference, and the differece in luminance from that(in stops) determined for each of teh other ISO values.
The ambnient environment was such that the metering system would recommend 1.3 second at F/5.6 at ISO 100, 1/12 second at ISO 1600. Thus, with a shutter speed of 1/125 in the flash test shots, the ambient component should be negligible
Those results are summarized here:
Distance: 24"
ISO Difference in stops
100 (ref)
200 -0.1
400 -0.1
800 +1.1
1600 +2.4
Distance: 36"
ISO Difference in stops
100 (ref)
200 +0.4
400 +0.2
800 +0.6
1600 +1.5
Distance: 48"
ISO Difference in stops
100 (ref)
200 +0.2
400 +0.8
800 +0.8
1600 +0.8
Distance: 60"
ISO Difference in stops
100 no data
200 (reference for this set)
400 0.0
800 -0.1
1600 0.0
Distance: 72"
ISO Difference in stops
100 (ref)
200 0.0
400 +0.2
800 +0.4
1600 no data
Distance: 84"
ISO Difference in stops
100 (ref)
200 -0.5
400 -0.3
800 -0.4
1600 +0.1
Distance: 96"
ISO Difference in stops
100 (ref)
200 0.0
400 +0.1
800 +0.3
1600 +0.9
It should be noted that there was ample opportunity for experimental error in this series of tests. Thus care should be taken in drawing inferences from small differences between the results for different ISO values. Additionally, any given shot may have a signficant error.
I make no claim to undestand the signficance of these results, other than clearly the flash metering system is more consistet in its behavior over a certain modest range of distances.
It appears that shutter speed has some influence in this matter, even less well understood. I discovered this when, at one point, I indavertently shifted the shutter speed from 1/125 to 1/160 rather than bumping the ISO setting.Having done so, I noted that the peak of the histogram was at a slightly higher luminance with 1/160 seconond than at 1/125 second, no other exposure factors having been changed. Go figure! I plan to run some tests tomorrow to explore the sensitivity of flash "exposure" to shutter speed cahnges under the 1/200 sec FP threshhold.
Best regards,
Doug
Would it be possible to hold the focal length constant, say at the wide side (18mm for kit lens) throughout all distances ? This is because the focal length is one important factor which determines the GN which gives an idication of the strength of the light reaching the object..
Zoomed at the tele range, the light will be diminished at longer distances and not sufficient for flash to expose correctly.
The size of the target should be slightly larger than the center AF point when viewed through the view finder at the longest range (96") If possible, the background should be a bland wall, curtain draping, or even the portable projection screen to fill up the rest of the frame...
Great job !
http://www.geocities.com/lasm.rm
http://www.igo4mac.com
Ben Seese
Senior Member
Remeber, John, shutter speed only affects ambient light. No matter the shutter speed, all of the flash's light is being captured, because the flash is only lit on the order of 1/10000 of a second.
Aperture, yes, would effect the flash exposure. Thing is, in program mode, the camera has to pick the aperture size before it sees how much of the flash's light is bouncing back.
So what it boils down to is that the camera guesses some basic aperture & shutterspeed values, then just watches during the exposure for the amount of flash light coming back, and adjusts the flash's output accordingly. But, the tiny little built-in flash has only so much flexibility in its output levels, and that's why some shots are underexposed and some are overexposed.
Aperture, yes, would effect the flash exposure. Thing is, in program mode, the camera has to pick the aperture size before it sees how much of the flash's light is bouncing back.
So what it boils down to is that the camera guesses some basic aperture & shutterspeed values, then just watches during the exposure for the amount of flash light coming back, and adjusts the flash's output accordingly. But, the tiny little built-in flash has only so much flexibility in its output levels, and that's why some shots are underexposed and some are overexposed.
Doug Kerr
Forum Pro
lasm,
Recall that the concept of the guide number relates to a fixed output from the flash. It is a way, as you know, to manually calculate an appropriate aperture based on the distance to the subject (and of course embracing numerous assumptions, likely unknown to us, about appropriate exposure, etc.!).
With an automatic flash the light output is throttled "on the fly" based on the light returned from the subject as measuredf by a sensor on the flash, thus (hopefully) accommodating both the distance from the flash to the subject and subject reflectance as well (the latter not always being what we want, of course). The guide number is not involved in this process (although we can in some cases determine the maximum "reach" of the flash in this mode, for a particular aperture, if we have a guide number that is consistent with the maximum output of the flash in that mode, which most often is the same as its output in the "manual" mode.
With E- TTL metering, the joint effect of distance to the subject and subject reflectance are determined by a measurement of the light returned from the metering pre-flash, as observed through the lens at its wide-open (metering) aperture. We would then expect that the flash metering system would, taking into account the ISO setting chosen for the shot and the aperture chosen for the shot, choose a flash output that should produce "approipriate exposure", and command the flash to fire with that output.
Again, here, the concept of guide number has no direct implication on the process. As before, if we knew the guide number that would correspond to the maximum available output from the flash, we could from that perhasp estimate the maximum "reach" of the system for a given aperture.
If all this worked as we might expect, the result would indeed be that, regardless of the ISO setting, the result in the final image would be the same, in terms of the distribution of coded luminance (assuming we don't outrun the "reach" for that ISO and aperture). Note that I don't say "the exposure would be the same" in the final image; the exposure would be less for a higher ISO (and exposure is not a property of the image, but rather of the - - "exposure"). If we were dealing with negative film, we could say that we would expect the negative density to be the same.
So the fact that we see this uniform result over part of the span of our tests is not at all surprising. What is surprising is that we don't observe it over the entire span of our tests.
Best regards
Doug
Recall that the concept of the guide number relates to a fixed output from the flash. It is a way, as you know, to manually calculate an appropriate aperture based on the distance to the subject (and of course embracing numerous assumptions, likely unknown to us, about appropriate exposure, etc.!).
With an automatic flash the light output is throttled "on the fly" based on the light returned from the subject as measuredf by a sensor on the flash, thus (hopefully) accommodating both the distance from the flash to the subject and subject reflectance as well (the latter not always being what we want, of course). The guide number is not involved in this process (although we can in some cases determine the maximum "reach" of the flash in this mode, for a particular aperture, if we have a guide number that is consistent with the maximum output of the flash in that mode, which most often is the same as its output in the "manual" mode.
With E- TTL metering, the joint effect of distance to the subject and subject reflectance are determined by a measurement of the light returned from the metering pre-flash, as observed through the lens at its wide-open (metering) aperture. We would then expect that the flash metering system would, taking into account the ISO setting chosen for the shot and the aperture chosen for the shot, choose a flash output that should produce "approipriate exposure", and command the flash to fire with that output.
Again, here, the concept of guide number has no direct implication on the process. As before, if we knew the guide number that would correspond to the maximum available output from the flash, we could from that perhasp estimate the maximum "reach" of the system for a given aperture.
If all this worked as we might expect, the result would indeed be that, regardless of the ISO setting, the result in the final image would be the same, in terms of the distribution of coded luminance (assuming we don't outrun the "reach" for that ISO and aperture). Note that I don't say "the exposure would be the same" in the final image; the exposure would be less for a higher ISO (and exposure is not a property of the image, but rather of the - - "exposure"). If we were dealing with negative film, we could say that we would expect the negative density to be the same.
So the fact that we see this uniform result over part of the span of our tests is not at all surprising. What is surprising is that we don't observe it over the entire span of our tests.
Best regards
Doug
Doug Kerr
Forum Pro
lasm,
What does figure into the process is aperture. For a given aperture (in f/-number terms), regardles of focal length, the relationship between the brightness of any part of the scene and the illuminance on the sensor for the corresponding part of the scene is determined only by aperture. (That's not quite precise - in fact if we are focused at a relatively short distance, there is a correction that comes into the picture.)
A projection screen would be very inappropriate as a background. They are typically retro-reflective, and the light return from them is very dependent on the precise location of the light source and the "receiver". For one thing, for a flash located very near the lens axis, such a screen can have an apparent reflectance much greater than 100%. That doesn't mean that it reflects more than all the light that hits it, of course. It is just that with a normal "diffuse" object, the light reflected from any source is spread over all exit angles, and the concept of a percentage reflectance takes that into account. If the reflected light is concentrated toward the source (as with any type of retroreflector), then the apparent reflectance can be gigantic.
Best regards,
Doug
Focal length does not have any influence on guide number (which in any case is not a concept dorectly related to the process of interest here).
What does figure into the process is aperture. For a given aperture (in f/-number terms), regardles of focal length, the relationship between the brightness of any part of the scene and the illuminance on the sensor for the corresponding part of the scene is determined only by aperture. (That's not quite precise - in fact if we are focused at a relatively short distance, there is a correction that comes into the picture.)
Of course, for any given light output from the flash, the illuminance of the subject diminshes at greater distances, but the focal length of the lens not involved in that.
I didn't have a wall-size gray target!
A projection screen would be very inappropriate as a background. They are typically retro-reflective, and the light return from them is very dependent on the precise location of the light source and the "receiver". For one thing, for a flash located very near the lens axis, such a screen can have an apparent reflectance much greater than 100%. That doesn't mean that it reflects more than all the light that hits it, of course. It is just that with a normal "diffuse" object, the light reflected from any source is spread over all exit angles, and the concept of a percentage reflectance takes that into account. If the reflected light is concentrated toward the source (as with any type of retroreflector), then the apparent reflectance can be gigantic.
Thanks, and thanks for your thoughts on this matter.
Best regards,
Doug
billybob_jcv
Senior Member
How about a direct comparison between the on-board flash and a well-known old-fashioned auto thyristor flash like a Vivitar or Sunpak? It might be interesting to see what is lost or gained by simply bypassing the fancy E-TTL algorithms for the same situation and letting the flash sensor take control of the situation.
Take a look at the camera manual on page 81. The specs for the internal built-in flash says GN is 13. To maintain the same GN, canon must have done their own development testing, and the range of effectiveness of the flash does depend on focal length... There are two tables there, one for 18mm for the kits lens, and the other for 55mm on the kit lens.
Take the first line of the table, at iso 100 for example...
1) At 18mm zoom, the GN 13 will be maintained at approximately a range of 0.7m to 3.7m... This also means the effective aperture you can set at this zoom is from f/3.5 to f/18. If the aperture is set at f/4.0, the flash is within the range.
2) At 55mm, the distance is from 0.7m to 2.3m, the corresponding aperture must be f/5.6 to f/18. If the aperture is set to f/4.0 the flash will be out of range.
I am trying to suggest a small target enough to fill the center AF box, not to flil the whole frame.
Yep, the projector screen itslef is probably a bad idea... something needs to be hanged over it, a piece of cloth maybe,... just a suggestion..
http://www.geocities.com/lasm.rm
http://www.igo4mac.com
xvrbx
Well-known member
If this is really the case (EOS300D not taking into account ISOs different than 400) that would also explain why setting FEC to +2/3 or so solves the problem for ISO 100.
Regards,
xvrbx
Regards,
xvrbx
I am also wondering, how many times did you try it? are the
results always the same?
--As for EXIF, I hope it's still there alsoResized in PS 7. If
not I'll reshoot them here.
-James
Daniella
main gallery: http://www.infrareddream.com
http://www.photosig.com/go/users/userphotos?id=26918
http://www.pbase.com/zylen
C7OO FORUM: http://www.c700uz.com
c7OOuz, Dimage-7, Tcon14tele, C210tele, Cokin-173, Grad-ND,
Hoya-red-Intensifier, Hoya_R_72.
xvrbx
Well-known member
Maybe the root cause of the problem is not underexposition but exposition calculated for ISO 400 (the ISO speed used in the auto modes). That, of course, will have the consequence of underexposition at ISO 100.
Best regards,
xvrbx
Best regards,
xvrbx
Ben Seese
Senior Member
nice thought, but this has already been shown later in the thread that this is not what is happening.
http://forums.dpreview.com/forums/read.asp?forum=1031&message=6668055
It's a long thread, so it's understandable...
http://forums.dpreview.com/forums/read.asp?forum=1031&message=6668055
It's a long thread, so it's understandable...
It is quite confusing here, and I am trying to grope around and learn too..
Thinking very simplistically and thinking out loud, subject to correction and input from others:
1) I tend to believe, iso, like aperture and shutter speeds are fixed for the ambient light exposure before the flash exposure comes into play..
So, we should logically see an increase in exposure value going from iso 100 to iso 1600. This is generally the case for the wine bottle test and the two gray card tests... except for the fact that in the grey card test, the difference between iso 100 and iso 200 is so small that the eyes can hardly pick it up..
Is there a problem here ? Is the iso 100 and iso 200 behaving in the way as to be almost the same ? Much like in some models, iso 50 and iso 100 are actually the same, they don't increase the exposure by one full stop, but much less..
2) The tricky built-in flash now comes into the picture.
Nobody knows how it works best except Canon. But I assume that since the flash is so tied in to the AF point, the camera should measure the light reaching the AF meters and cut off the flash output once it is sufficient.
It shouldn't be concerned about the overall pic brightness or exposure, which should be taken care of by the ambient exposure process...
Which means, very simplistically, if I put the flash target just under the AF box (not the whole frame of the sensor) I would expect the flash to cut off once there is sufficient light there..
This is difficult to determine by the flash card tests and the wine bottle tests... because the flash target does not distinguish itself from the surrounding objects and is affected by the overall exposure increase as a result of raising the iso..
The white cup test however, is more revealing... We only have two shots at iso 100 and iso 400 to compare, but it is evident at iso 400 the ambient exposure is raised by two stops from the iso 100
The cup however, is deemed to be properly exposed by the flash at iso 100. It will appear to many people as under-exposed though... So at iso 400, the flash should not increase the exposure for the cup, if anything, it should be reducing exposure by one stop, because the brightness of iso 400 would have lifted up the overall exposure of the background and the cup at the same time..
http://www.geocities.com/lasm.rm
http://www.igo4mac.com
Thinking very simplistically and thinking out loud, subject to correction and input from others:
1) I tend to believe, iso, like aperture and shutter speeds are fixed for the ambient light exposure before the flash exposure comes into play..
So, we should logically see an increase in exposure value going from iso 100 to iso 1600. This is generally the case for the wine bottle test and the two gray card tests... except for the fact that in the grey card test, the difference between iso 100 and iso 200 is so small that the eyes can hardly pick it up..
Is there a problem here ? Is the iso 100 and iso 200 behaving in the way as to be almost the same ? Much like in some models, iso 50 and iso 100 are actually the same, they don't increase the exposure by one full stop, but much less..
2) The tricky built-in flash now comes into the picture.
Nobody knows how it works best except Canon. But I assume that since the flash is so tied in to the AF point, the camera should measure the light reaching the AF meters and cut off the flash output once it is sufficient.
It shouldn't be concerned about the overall pic brightness or exposure, which should be taken care of by the ambient exposure process...
Which means, very simplistically, if I put the flash target just under the AF box (not the whole frame of the sensor) I would expect the flash to cut off once there is sufficient light there..
This is difficult to determine by the flash card tests and the wine bottle tests... because the flash target does not distinguish itself from the surrounding objects and is affected by the overall exposure increase as a result of raising the iso..
The white cup test however, is more revealing... We only have two shots at iso 100 and iso 400 to compare, but it is evident at iso 400 the ambient exposure is raised by two stops from the iso 100
The cup however, is deemed to be properly exposed by the flash at iso 100. It will appear to many people as under-exposed though... So at iso 400, the flash should not increase the exposure for the cup, if anything, it should be reducing exposure by one stop, because the brightness of iso 400 would have lifted up the overall exposure of the background and the cup at the same time..
http://www.geocities.com/lasm.rm
http://www.igo4mac.com
Doug Kerr
Forum Pro
lasm,
If we assume that the flash would be at "maximum output" at the distance limit in the table, and want to describe that maximum output in terms of guide number, then the guide number for ISO 100 (and guide number, as you know depends on ISO, and stating one is only meaningful if we state it at a particular ISO value) would be 42: 7.5 feet x f/5.6 (which is the aperture implied by their "55 mm" column heading - see the note shortly below). A GN of 42 (ISO 100 basis) is resonable for a small flash unit.
If the GN at maximum outpout were 13, then the maximum distance at an aperture of f/5.6 would be 2.3 feet!
Best regards,
Doug
Where does it say the GN is 13? Perhasp you have inferred it from the table, but not correctly.
If we assume that the flash would be at "maximum output" at the distance limit in the table, and want to describe that maximum output in terms of guide number, then the guide number for ISO 100 (and guide number, as you know depends on ISO, and stating one is only meaningful if we state it at a particular ISO value) would be 42: 7.5 feet x f/5.6 (which is the aperture implied by their "55 mm" column heading - see the note shortly below). A GN of 42 (ISO 100 basis) is resonable for a small flash unit.
If the GN at maximum outpout were 13, then the maximum distance at an aperture of f/5.6 would be 2.3 feet!
No, no - this is beacuse they are talking about a particular lens, the "kit" lens, and recognizing the maximum aperture that particular lens has at those two focal lengths - f/3.5 at 18 mm and f/5.6 at 55 mm. It's a way of simplifying the range for people using that particular lens. It is only the aperture that matters. The two columns are really for f/3.5 and f/5.6.
That would give very inconsistent metering results - the metering in any case looks at an area bigger than the focusing mark square - the square is only a symbol. (See the manual, p 84.) It is important to have a uniform reflectance over the scope of the metering system - I chose the whole frame.
Best regards,
Doug
Nope, I did not pluck it out from the air.. This spec is further down in the manual at page 134. The built-in flash is rated at 13 meters at iso 100, equivalent to 43 ft at iso 100.
No it is not reasonable.. The GN for the top end Canon ETTL Flash 550EX is rated at 55 at iso 100. To think the tiny built-in flash have the power of 42 is to grossly over-estimate its performance. In actual fact, the rated GN of built-in is only 13.
The GN for 550Ex you can check at page 121 of the 550EX manual. You will notice that the GN at 17mm coverage is only 15, and increases to 105mm which is at its maximum of 55.
Yes. They release the specs for the kits lens, since this is the lens most people will have. But the range is still valid. From my calculation for example, we know f/18 is the maximum effective aperture, but in the camera itself, you can set the apterture to 33 or 22 depending on zoom..
but at these small apertures, the built-in flash would be totally ineffective !
I am speculating that if the experiment is to derive any result of the flash exposure, by isolating away the influence of the ambient exposure, then one need to concentrate only on the small AF area where it is known that flash exposure is very strongly biased, so in effect, you can ignore any other area of the meters as far as flash exposure is concerned...
lasm
http://www.geocities.com/lasm.rm
http://www.igo4mac.com
Doug Kerr
Forum Pro
lasm,
Best regards,
Doug
Oh, you meant on a mater basis. One needs to state foot or meter basis (although I didn;t in my note!)
I meant 42 on a feet basis (as you might have concluded from my mention of the implied range "in feet"). You are right in both cases for values in meters. 13 (meters basis) is quite reasonable for a small, on-board flash unit.
Best regards,
Doug
Keyboard shortcuts
- f
- Forum