[snostorm]

Hi Noam,

Excuse me for probably just being dense, but what practical application is there for a shallow flash head angle? I'm no flash wonk, but neither am I a rank beginner. I own quite a few guns, and use them as needed -- and am one of those that have had few problems with P-TTL over the years. I own and use an AF 360 FGZ, an AF 540 FGZ, and a newly acquired Metz 24 AF1 in P-TTL, and either shoot them with direct flash or bounced straight up, with or without a multitude of modifiers. I haven't found much use for any of the intermediate angles, and I don't use Omnibounce type diffusers as I find there are much more effective ways to soften the light (one of the few common applications where an intermediate angle is recommended).

The only relevance that I can see your tests having for me is that I now know to check to see if the head has been inadvertently tilted a touch if I start getting anomalous overexposures in the field. This is good to know, and I thank you for this, but IMO, you're spending a lot of time and effort to point out a fault in the P-TTL flash protocols that Pentax will probably have little or no motivation to correct.

It's pretty commonly accepted that Pentax should dramatically revamp their flash system to at least be competitive with the big two in versatility and ease of use. With backward compatibility in mind (hopefully), they should retain compatibility with current P-TTL flashes when developing the new system, but I would think that their priorities would be to put the work into developing a new system, not to patch up the one that would be replaced.

I'm not trying to discourage you. I encourage experimentation because that's how we learn new tricks to get better images from our gear, and any discoveries that might get me better results are appreciated. With P-TTL flash, I know that I've got a pretty wide range of exposure latitude with both Ev comp and Flash comp to work with. I can work around just about any metering problems within the system that exists as long as the work-around is consistently effective.

Scott

[Noam]

Here are some if the images from the test. Please refer to my two previous (lengthy) posts for detailed explanations of what you see and how it was created. The composite image was creating with LR's "Book" feature and the photos were all developed equally with the default settings.

Each row is a set of images shot with identical parameters. The captions describe the parameters and the "tilt" operation (as described in aforementioned posts). The numbers in parentheses indicate the amount of exposure discrepancy between the photo and the leftmost photo (taken in normal straight fashion). This is the amount of post-processing exposure reduction done in LR until the histograms were equalized.

Note that ISO and aperture were chosen for the sake of testing conditions variety, not as best choices for real-world use (of course). At this point, I wouldn't want to argue about the semantics of whether the bright photos are overexposed or the darker photos are underexposed (For example, the ISO80 set is clearly too dark in the straight shot and actually better in the "tilted" shots). The most important fact here (for the purpose of acknowledging and analyzing the bug) is that a consistent and very large exposure discrepancy exists between the shots in each set.

[dosdan]

Wasnt talking about the focal length or anything else, just the ISO, with ISO200 you double the GN, how would you say it then?

Doubling the ISO does not increase the GN rating. It increases the effective distance, hopefully by 1.4x.

It's like increasing the f-number from the rated f/1->f/2 (-2 stops or 1/4). This does not change the GN rating either, but it reduces the effective distance by 4x.

Changing the lens FL can change the GN rating due to flash reflector zooming (ie. changing the Angle-of-Illumination).

Dan.

[dosdan]

My pet hate when using bounce flash at an angle of 60°, is when the top half of a person ends up brighter than the bottom. Whick is OK for bust shots, but a problem with full-length shots. What do you guys do to reduce this? Set the flash to straight up? Flash towards the rear ceiling if the rear wall is close enough and light coloured?

Dan.

---

Last edited by dosdan; 12-12-2012 at 02:44 PM.

[Noam]

Hi Noam,

... what practical application is there for a shallow flash head angle? ... I haven't found much use for any of the intermediate angles,...
The only relevance that I can see your tests having for me is that I now know to check to see if the head has been inadvertently tilted a touch ... IMO, you're spending a lot of time and effort to point out a fault in the P-TTL flash protocols that Pentax will probably have little or no motivation to correct.
... I would think that their priorities would be to put the work into developing a new system, not to patch up the one that would be replaced.
... With P-TTL flash, I know that I've got a pretty wide range of exposure latitude with both Ev comp and Flash comp to work with. I can work around just about any metering problems within the system that exists as long as the work-around is consistently effective.

Scott

Hi Scott,
Thanks for your comments. There is no particular practical application to using a slight tilt as I did. It was for the purpose of the test. I opted for as similar as possible shooting conditions (and that includes flash position and direction), in order to eliminate variables and help isolate the real problem (apparent firmware bug in camera or flash). I demonstrated that everything being equal (including flash angle) we get very different exposures, which is a bad thing.

The flash should not overexpose as a result of being "inadvertently tilted a touch". It should simply work, and it is realistically capable of doing so, as has been demonstrated in the well-exposed shots (straight forward, tilted down, and even bounced in wireless TTL!). P-TTL is a good and capable system apparently ruined by a silly bug. The relevance of my test is that it helps exposing and isolating that silly bug, which can and should be fixed by the manufacturer, for the benefit of us all. We pay good money for it and It should work as advertised. I don't think it's a fundamental fault or limitation of the aging P-TTL protocol. It's a bug that hopefully can be easily fixed.

Perhaps I should stress the importance of fixing this bug. It is not limited to my convoluted test shots. I decided to perform the test after using the flash with my new K-5II and shooting normal subjects at "normal" bounce angles (e.g. ceiling) like you do, and getting lots of ill-exposed shots. Fixing this bug will make P-TTL the consistent and dependable system that it should be and has been with older DSLR models (according to many people's testimonies, the bug does not occur with K10D, K20D and older models).

Exposure and flash compensation are important features that have many good uses - from dealing with dark or bright objects, reflective objects, to creatively balancing tones and shadows. Working around P-TTL bugs is not one of them. True, I will probably be able to live through it with constant chimping and tweaking of flash compensation, but it's not how it should be, and it's not something I, you, or any other customer, should put up with. And I think the manufacturer's priorities should definitely include supporting existing customers rather than leaving them out in the cold and fixing problems only in subsequent models.

I spent hours performing the tests and writing these lengthy posts not (god forbid) to pick up a fight, argue, or bad-mouth Pentax or Sigma. On the contrary. I’m doing this as a service to myself, the Pentax user community, and even Pentax and Sigma themselves, helping expose and diagnose a problem, and hopefully motivating them to fix it.

I urge other users to contribute. Any further information and experience will help complete the picture. Please do not suffice with anecdotal evidence (“mine works great” or “Yea, I had this happen to me too once”), if you can. Perform a methodical test and share your findings.

With enough data, I hope we will be able to make a difference - approach the manufacturers and get some answers and hopefully solutions.

[Noam]

Oh, and on a lighter tone, here is the image printed on the canvas seen in the test images. This is, after all, a photo forum It was taken with a K-7.

500px / Photo "Mourning Cyclamens" by Noam Gordon

[Anvh]

Doubling the ISO does not increase the GN rating. It increases the effective distance, hopefully by 1.4x.

It's like increasing the f-number from the rated f/1->f/2 (-2 stops or 1/4). This does not change the GN rating either, but it reduces the effective distance by 4x.

Changing the lens FL can change the GN rating due to flash reflector zooming (ie. changing the Angle-of-Illumination).

Dan.

I think you're missing what you're saying here.

The aperture and ISO does change the effective GN rating.

GN is not just about the distance but about the effective flash power.

[Quicksand]

Noam,

Your contributions here are very valuable. I appreciate the effort you have put into testing.

Unfortunately, your results are entirely, 100% consistent with my observations using a K-5 "classic" and a Pentax AF360 flash. They are precisely what I see.

I have hammered on this a number of times in prior posts over the last year or more, and I see you are now receiving some of the same objections:

Maybe you are trying to operate the flash outside of design parameters, too close, ISO wrong, or whatever. NO. The direct flash almost always exposes correctly. In edge cases, bounce flash should underexpose if anything.

Maybe it's operator error. NO. My old K20d could handle this no problem.

You can work around the problem by dialing in negative flash exposure compensation, or whatever. NO. This is P-TTL, it's supposed to figure this out by itself!

It's a bug. It's clearly a bug, and I have said so repeatedly. It works fine in direct flash mode. It works fine wirelessly. It's only the combination of external flash, on-camera, bounced that results in overexposure.

I am TREMENDOUSLY disappointed to learn that this is still a problem on the new models. I'm losing hope for a fix. I have basically given up on bounce flash at this point, and in cases where I really need it, I either go with full-manual settings, or take the flash off the camera and use wireless P-TTL.

Some of my prior posts on the subject:

https://www.pentaxforums.com/forums/pentax-k-5/200387-k-5-pttl-bounce-flash-a...ml#post2116767

https://www.pentaxforums.com/forums/pentax-k-5/191116-some-insight-into-flash...ml#post2027171

https://www.pentaxforums.com/forums/pentax-k-5/171474-firmware-v1-12-23-20-a-4.html#post1780208

[Anvh]

My pet hate when using bounce flash at an angle of 60°, is when the top half of a person ends up brighter than the bottom. Whick is OK for bust shots, but a problem with full-length shots. What do you guys do to reduce this? Set the flash to straight up? Flash towards the rear ceiling if the rear wall is close enough and light coloured?

Dan.

Increase the flash distance, the longer the distance the smoother the gradation.
Think about the square law.

[Anvh]

Here are some if the images from the test.

Looks to me the camera is metering the reflection in your image.

[IchabodCrane]

Let me throw out a wild idea.... with the flash pointed straight on, the camera controls the flash power based on what the AF system calculates for distance to subject. When the flash is tilted up, it goes out of this mode as it assumes (right or wrong) that the lighting will be from a bounce source. The flash power is no longer governed by the distance to subject calculation from the AF system. Perhaps in this non-AF governed mode, something is wrong in how Pentax is measuring and treating the flash power.

[dosdan]

I think you're missing what you're saying here.

The aperture and ISO does change the effective GN rating.

GN is not just about the distance but about the effective flash power.

But changing the f-number and/or ISO is not changing the flash power. The GN rating is the same, but the effective flash illumination distance is changed.

Take a GN of 53, rated with a 105mm FL lens, f/1, ISO100, 25% subject reflectivity. Assuming you had a subject of the correct reflectivity and you were spot metering it, if you had a 105mm f/1 lens, at ISO100 and max. flash power, you should be able to correctly illuminate a subject at 53m.

Now close the aperture to f/2. Since 1/4 of the previous light is now reaching the sensor, with the inverse-square law the effective distance is distance/f-number ie. 53m/2 = 26.5m. So the effective distance is reduced, but the GN rating is still 53 (the flash power produced is still the same, but the lens is letting though less reflected subject light.)

Now increase the ISO: ISO100->ISO800. You have not increased the flash power at all. The same amount of reflected light as before is reaching the the sensor. But now you are boosting the sensor output signal, thus increasing the rendered brightness. So a dimmer captured image is now displayed, after post-sensor gain, at the "correct" brightness level. Effectively you are increasing the acceptable flash illumination distance. Since ISO100->ISO800 is 8x, the square-root of this for the change in effective distance is root-8 = 2.8x approx.

So with a 105mm lens at f/2 and ISO800, the effective flash illumination distance is 53m/2 x 2.8 = 74m
approx.

BTW, GN is not just about absolute flash power either, as it also involves the directing effect of the head reflector: the flash is a beam, rather than omnidirectional. The directivity of the flash reflector changes in a "zooming" flash unit and this alters the GN rating at different FLs.

Dan.

[Ray Pulley]

Here are some if the images from the test. Please refer to my two previous (lengthy) posts for detailed explanations of what you see and how it was created. The composite image was creating with LR's "Book" feature and the photos were all developed equally with the default settings.

Each row is a set of images shot with identical parameters. The captions describe the parameters and the "tilt" operation (as described in aforementioned posts). The numbers in parentheses indicate the amount of exposure discrepancy between the photo and the leftmost photo (taken in normal straight fashion). This is the amount of post-processing exposure reduction done in LR until the histograms were equalized.

Note that ISO and aperture were chosen for the sake of testing conditions variety, not as best choices for real-world use (of course). At this point, I wouldn't want to argue about the semantics of whether the bright photos are overexposed or the darker photos are underexposed (For example, the ISO80 set is clearly too dark in the straight shot and actually better in the "tilted" shots). The most important fact here (for the purpose of acknowledging and analyzing the bug) is that a consistent and very large exposure discrepancy exists between the shots in each set.

Noam,

i was waiting to comment until I saw the subject of your tests, and now that I have, it seems to confirm what I have observed with the bounce flash over-exposure issue.

A little background:

I have a Pentax 540, and an older Sigma 500 PTTL super. They worked fine in bounce on my Pentax K10s and K20s (well, at least as good a PTTL ever is, but no overexposure in bounce).

The K7 showed a similar problem, as does the K5 (I have not yet tested my K01 for this, although something is different on the K01 as the Sigma 500 does not even work at all on the K01 but does on the K5 ).

My normal settings for indoor bounce with a white ceiling are ISO 400, AV mode, f5.6, usually with a -.5 FEC and I normally have a Demb diffuser attached. I cannot count the number of shots I have taken this way with many Pentax bodies over the years, and the results were very consistent body to body.

What I have found is that the over-exposure is a result of having a significant amount of white or light colored area in the scene. I can shoot a person in non-light clothing, bouncing as I normally do and all is fine. If I shoot another person with a white shirt in the same exact position and with the same settings, I get serious over-exposure.

I have tested this in my kitchen where I have oak cabinets and a white dishwasher set into the cabinets. If I do not include the dishwasher in the scene, exposure is fine when bouncing off my white ceiling. If I include the dishwasher, I get over-exposure. I have also seen this over and over with white tables and other shots with lots of white in the scene as I use bounce quite often indoors. This is opposite of what should happen with traditional metering of a white scene, as it should expose to push the white down towards mid-grey, just as it should push scenes with lots of black up towards mid-grey.

I have not found any settings that make the K5 flash work well in PTTL/bounce, and direct flash does not have the same problem.

It is hard to understand why this has not been fixed.

Ray

[Noam]

Looks to me the camera is metering the reflection in your image.

I'm not sure what you mean - there is nothing reflective in the scene (both canvas print and piano are matte finished), and anyway all images in the set should have been affected equally.

Let me throw out a wild idea.... with the flash pointed straight on, the camera controls the flash power based on what the AF system calculates for distance to subject. When the flash is tilted up, it goes out of this mode as it assumes (right or wrong) that the lighting will be from a bounce source. The flash power is no longer governed by the distance to subject calculation from the AF system. Perhaps in this non-AF governed mode, something is wrong in how Pentax is measuring and treating the flash power.

It's not a wild idea at all, and indeed this is one of my own speculations. Some other flash systems (Sony and maybe Nikon) utilize focus distance in straight flash power calculations. I don't know of such a feature in P-TTL but it's possible. It should be easy to test this by taking several identical shots on a tripod, where the AF is fooled to focus to various different distances.
Even if it's not distance-based, at the least it is quite plausible that the system uses two different power calculation logics - for straight and bounced positions - and for some reason the two are highly uncorrelated.

[Noam]

Quicksand and Ray, thank you very much for your comments.
Quicksand, your words are both encouraging (reinforcing my findings) and discouraging (they may never fix this...). Thanks for the thread references. I read through at least some of them in the past.
Ray, can you confirm that you had the same misbehavior with the Pentax 540 flash? This is very important. Thanks for the tip on bright objects. I'll investigate it (although my test scene is not overly bright. The wall is painted green).

If anyone could upload some test images it would be awesome.
Two images will be enough: Just put the camera on a tripod, set appropriate manual settings, and shoot one straight-flash and one tilted-flash shot in P-TTL.