[sTi]

Hi,
after some testing with the K-5II and indirect (i.e. tilted) flash I have made some interesting observations that I'd like to share. Plus, I've also got a theory about the cause of the overexposure problems.

For a long time I've used a flash with only an "A" setting and only recently bought a P-TTL flash (the Metz 58 AF-2). Before investing in this flash for my K-5II I contacted Metz (who are renowned for their great customer service) for any news on the notorious indirect flash overexposure problem that has plagued the K-5 and K-5II (at least with early firmware). The Metz technician who answered my inquiry confirmed the misbehaviour of the K-5 but told me that the K-5II (with latest firmware) behaves very differently and tends to slightly underexpose with indirect flash (compared to direct flash). Furthermore, he claimed that the exposure behaviour of genuine Pentax flashes is exactly the same.

Here are my observations for Metz 58 AF-2 and K-5II, both of which are running on the latest firmware, by the way (4.1 and 1.06, respectively)

My first pictures with the new flash and 2 lenses (DFA 100/2.8 Macro WR and DA 18-135) completely confirmed the tendency of slight underexposure described by the Metz technician - with indirect flash the camera tends to underexpose about 1/3 to 2/3 stops (compared to direct flash). What is more, it seems to do so pretty consistently, so by dialling in +1/3 to +2/3 flash exposure compensation for indirect flash the results are good. Problem solved, I thought.

That is until I used I tried the flash with my DA*55/1.4. There it behaves differently and still can suffer from overexposure. Why can? I've found that it depends on the subject you're photographing. For subjects without very bright parts, the exposures are OK (very little difference between direct and indirect flash). The trouble starts if there are larger bright areas (e.g. white walls) in your picture. Then I usually got about +2/3 stops overexposure with indirect flash (compared to direct).

After some more testing I have a theory what's causing the problem: The pre-flash exposure measurement. To be more specific, a mismatch between the strength of the pre-flash and the light sensors used to measure the reflected light (and thus calculate the flash exposure). The light sensors probably have a certain working range, e.g. they can record the measured luminosity range on a 0 to 100 range.
When using direct flash, this limited range is not problematic because the possible range of reflectivity given a specific pre-flash can be known pretty well in advance, so pre-flash exposure and measurement can be well matched in advance.

The problem with indirect flash is that there are great unknowns about the probable reflectivity range. It can be
(1) similar to direct flash (when flash is bounced on a white ceiling that is not far away) or
(2) much lower (when flash is bounced off something darker and further away).

In the latter case, a too weak pre-flash might not be strong enough to get useful reflectivity data (i.e. the sensors might just get data between, say, 0 and 1). My guess is that the camera thus uses a much stronger pre-flash when the flash reflector is tilted to avoid running into the "too little data" problem in the case of (2). However, it can now run into the opposite problem, the "100" ceiling, if the conditions are like (1). So, even if light with the strength of 130 hits the sensor, it only sees "100". Thus, it underestimates the reflected light (because the data from some of the sensors may be clipped) and thus overestimates the needed flash power, leading to overexposure.
The theory of different pre-flash strength for tilted and non-tilted flash head might explain why the overexposure problem has always already been present by just tilting the flash head slightly so that the camera thinks it's using indirect flash while in reality it's more like direct flash.

So, my guess is that the overexposure is a mixture of hardware limitation of the flash system shared by K-5 and K-5II (sensitivity range of light sensors) and a not very intelligent use of pre-flash strength.
A further guess is that Pentax has partly addressed this problem in the latest firmware for the K-5II so that pre-flash strength and measurement are better matched. However, it still doesn't seem to be matched perfectly to the speed of the lens. A f/1.4 lens lets in 4 times as much light from the pre-flash as a f/2.8 lens and thus can seemingly still run into the upper limit of the light sensors.

To sum up, the tilt flash overexposure seems to be fixed at least for lenses 2.8 or slower, at least when using genuine indirect flash (and not slightly tilted direct flash). Faster lenses might still suffer from overexposure in certain conditions, but the amount seems to be a lot smaller than the +1 to +2 range reported from K-5 and early firmware K-5II.

I hope that Pentax addresses this remaining problem, then we'd (finally!) have a flash exposure system that can be called "reliable".

Clarification: I presume that the pre-flash is of course measured with the lens at wide open aperture (and not the set aperture), hence the different amount of light that enters through a very fast vs. a slow lens. The overexposure issue with DA*55 was regardless of the used aperture (I tried f/2 through f/5.6) (and also regardless of ISO or zoom reflector settings, by the way).

---

Last edited by sTi; 10-07-2013 at 04:31 AM. Reason: Clarification

[Ray Pulley]

Hi,
after some testing with the K-5II and indirect (i.e. tilted) flash I have made some interesting observations that I'd like to share. Plus, I've also got a theory about the cause of the overexposure problems.

For a long time I've used a flash with only an "A" setting and only recently bought a P-TTL flash (the Metz 58 AF-2). Before investing in this flash for my K-5II I contacted Metz (who are renowned for their great customer service) for any news on the notorious indirect flash overexposure problem that has plagued the K-5 and K-5II (at least with early firmware). The Metz technician who answered my inquiry confirmed the misbehaviour of the K-5 but told me that the K-5II (with latest firmware) behaves very differently and tends to slightly underexpose with indirect flash (compared to direct flash). Furthermore, he claimed that the exposure behaviour of genuine Pentax flashes is exactly the same.

Here are my observations for Metz 58 AF-2 and K-5II, both of which are running on the latest firmware, by the way (4.1 and 1.06, respectively)

My first pictures with the new flash and 2 lenses (DFA 100/2.8 Macro WR and DA 18-135) completely confirmed the tendency of slight underexposure described by the Metz technician - with indirect flash the camera tends to underexpose about 1/3 to 2/3 stops (compared to direct flash). What is more, it seems to do so pretty consistently, so by dialling in +1/3 to +2/3 flash exposure compensation for indirect flash the results are good. Problem solved, I thought.

That is until I used I tried the flash with my DA*55/1.4. There it behaves differently and still can suffer from overexposure. Why can? I've found that it depends on the subject you're photographing. For subjects without very bright parts, the exposures are OK (very little difference between direct and indirect flash). The trouble starts if there are larger bright areas (e.g. white walls) in your picture. Then I usually got about +2/3 stops overexposure with indirect flash (compared to direct).

After some more testing I have a theory what's causing the problem: The pre-flash exposure measurement. To be more specific, a mismatch between the strength of the pre-flash and the light sensors used to measure the reflected light (and thus calculate the flash exposure). The light sensors probably have a certain working range, e.g. they can record the measured luminosity range on a 0 to 100 range.
When using direct flash, this limited range is not problematic because the possible range of reflectivity given a specific pre-flash can be known pretty well in advance, so pre-flash exposure and measurement can be well matched in advance.

The problem with indirect flash is that there are great unknowns about the probable reflectivity range. It can be
(1) similar to direct flash (when flash is bounced on a white ceiling that is not far away) or
(2) much lower (when flash is bounced off something darker and further away).

In the latter case, a too weak pre-flash might not be strong enough to get useful reflectivity data (i.e. the sensors might just get data between, say, 0 and 1). My guess is that the camera thus uses a much stronger pre-flash when the flash reflector is tilted to avoid running into the "too little data" problem in the case of (2). However, it can now run into the opposite problem, the "100" ceiling, if the conditions are like (1). So, even if light with the strength of 130 hits the sensor, it only sees "100". Thus, it underestimates the reflected light (because the data from some of the sensors may be clipped) and thus overestimates the needed flash power, leading to overexposure.
The theory of different pre-flash strength for tilted and non-tilted flash head might explain why the overexposure problem has always already been present by just tilting the flash head slightly so that the camera thinks it's using indirect flash while in reality it's more like direct flash.

So, my guess is that the overexposure is a mixture of hardware limitation of the flash system shared by K-5 and K-5II (sensitivity range of light sensors) and a not very intelligent use of pre-flash strength.
A further guess is that Pentax has partly addressed this problem in the latest firmware for the K-5II so that pre-flash strength and measurement are better matched. However, it still doesn't seem to be matched perfectly to the speed of the lens. A f/1.4 lens lets in 4 times as much light from the pre-flash as a f/2.8 lens and thus can seemingly still run into the upper limit of the light sensors.

To sum up, the tilt flash overexposure seems to be fixed at least for lenses 2.8 or slower, at least when using genuine indirect flash (and not slightly tilted direct flash). Faster lenses might still suffer from overexposure in certain conditions, but the amount seems to be a lot smaller than the +1 to +2 range reported from K-5 and early firmware K-5II.

I hope that Pentax addresses this remaining problem, then we'd (finally!) have a flash exposure system that can be called "reliable".

Clarification: I presume that the pre-flash is of course measured with the lens at wide open aperture (and not the set aperture), hence the different amount of light that enters through a very fast vs. a slow lens. The overexposure issue with DA*55 was regardless of the used aperture (I tried f/2 through f/5.6) (and also regardless of ISO or zoom reflector settings, by the way).

All metering is done with the lens wide open.

I can confirm that the bounce over-exposure in the K5 is directly related to the amount of white or very light colored things in the scene.

I do not have a K5II to test your theory on, however. I do have 1.4, 2.8 and f 4 lenses, however, so maybe if I get a chance, I can test them on my K5 to see if the maximum aperture makes any difference.

Ray

[rawr]

I have made some interesting observations that I'd like to share.

Nice investigative problem solving. Thanks for sharing this.

[sTi]

I can confirm that the bounce over-exposure in the K5 is directly related to the amount of white or very light colored things in the scene.

I do not have a K5II to test your theory on, however. I do have 1.4, 2.8 and f 4 lenses, however, so maybe if I get a chance, I can test them on my K5 to see if the maximum aperture makes any difference.

Ray

Thanks for your feedback and offer to test. It would be interesting to see whether the K-5 with latest firmware behaves similar to K-5II now.
I have made further tests with 2.4 (DA 35/2.4) and 2.8 (DFA 100/2.8 Macro WR) lenses. Exposure differences between direct and indirect seem to be a bit more variable than at first suspected, also for slower lenses. It is still possible to trigger the overexposure even with 2.8 lenses, but it's only with very bright subjects, and then only a third of a stop or so. What is more, I've found that it's sometimes also possible for fast (1.4) lenses to slightly underexpose, e.g. with very dark subjects.

From my tests I'd say the range of differences between direct and indirect flash (mostly depending on lightness of subject) are as follows:
For 1.4 lenses: indirect flash -1/3 to +2/3 stops different from direct flash, with stronger tendency to overexposure
For 2.8 lenses: indirect flash -2/3 to +1/3 stops different from direct flash, with stronger tendency to underexposure

So it seems we're dealing with a 1 stop range whose beginning and end (and its tendencies) are determined by the speed of the lens used.
As Pentax has now confirmed in the interview on the Forum front page that the issue is caused by a hardware problem of the exposure system, I am skeptical that this remaining difference will ever be solved by a new firmware. I guess they've by now done all they can possibly do to work around the hardware issue.
As I shoot only Raw, the 1 stop difference range is not a big problem for me and can easily be corrected, but for JPEG only shooters it must still be pretty frustrating to get the exposure right.