K10D and P-TTL
I have yet to get good results from PTTL but haven't the time, and quite honestly I probably don't have enough knowledge to go through what appears as a pretty intense set of tests that you've done to figure out exactly why the results are so "inconsistent".
The historgrams on the web page loaded quickly, but none of the images loaded. I guess I'll download the ZIP and look at it later.
That's a lot of work Matt, I hope something comes of it.
Dan
I think it is quite apparent that the flash is not working as expected in all cases, but it adds another data point on what is going on.
Now that I have a light meter I may try to re-run the PTTL tests I ran last year... probably won't get time until this weekend or so.
It is just nice to see others are getting similar findings.
I'll let you know my findings when I have some.
Hefty,
It is not always inconsistent Especially when there is a reflective patch in the viewfinder do things go AWOL.
What I did was work out a setup where the P-TTL was definitely underexposing every time. Other scenes will expose just fine.
When I've used the AF540 for events for fill in flash I found that it will consistently expose to keep highlight detail. That was very, very helpful when you're talking about a wedding dress! (Think it was on the HS setting.)
It's just in some circumstances the flash goes completely potty - and I am seriously curious to work out what it is. I did try on the *ist-D many times, but it is easier with the K10D because you can instantly adjust the ISO rating
There are other modes of flash - slow speed sync and HS - that I need to give a try as well. Unfortunately I don't have a flash meter (But that would've been a good idea!)
Matt
(For gallery, tips and links)
Perhaps it's because I can't recall taking many (any?) pics with a reflective surface in them that I haven't noticed this sort of behaviour from my setup. I will certainly be repeating your exercise though in the next few days and I'll let you (and everyone else) know what happens.
Unfortunately, I don't own a flash meter either so I'll just have to trust in histograms after the event too.
Carefully set up a CD in the middle of the frame so that it will reflect the light. It appears to be down to reflections, for sure. Haven't yet worked out if it's just in the centre area or not, but the compensation when there is a reflection seems overly aggressive from what I can see.
The "A" mode on flash remains a better shot overall. However, the K10D on P-TTL seems to wind back the flash too hard when there's a reflection - even with flash compensation there are still only a couple of pixels blown out. I can't believe the metering is that sensitive to pin-lights.
The "Slow Sync" flash settings are all perfect (as they were with M-mode on camera with P-TTL flash). "HS" makes no difference (as I'd expect at low light levels). Tried red eye reduction mode, trailing curtain sync - no difference (again, as expected).
The flash compensation within the K10D flash menu seems to work reasonably, though I've not done any exhaustive testing on it - it may well behave the same as the "flash-compensation-on-AF540" that I did yesterday.
My current thoughts to pin down are:
* Is the flash metering on reflective lights down to pin-points of light?
* Is the metering just in one area in the viewfinder, or matrix?
* Could the flash be controlled in discrete steps (like 1Ev steps) that produces these rather "stepped" results?
Thoughts, thoughts, thoughts.... and not too many answers
Matt
(For gallery, tips and links)
My current thoughts to pin down are:
* Is the flash metering on reflective lights down to pin-points of light?
* Is the metering just in one area in the viewfinder, or matrix?
* Could the flash be controlled in discrete steps (like 1Ev steps) that produces these rather "stepped" results?
Not much to add, but...
From what I've got from Pentax when asking about PTTL behavior, the pre-flash metering is in matrix mode.
Some of my initial confusion was that I wanted spot metering but there was no effect on flash output. As a bright object broke a certain percentage of the frame, all went to hell. Which certainly makes more sense when you know the intended behavior.
This time, I had the following setup:
* CD placed reflective side facing K10D
* AF540 on a tethered lead
Now, I used the digital preview to align the flash, CD and K10D to produce a nasty reflection. Result - massive underexposure in P-TTL (like near black frame, except for the reflection).
Because I could move the camera without moving the flash, I found the following points:
1. The effect of reflection appears to be the same all over the viewfinder.
2. The size of the reflection is absolutely irrelevant. If I zoomed out to 24mm so the reflected part was a tiny dot in the bottom left corner (that's about 2mm square on the K10D LCD), then the massive underexposure still occurs.
3. The underexposure is directly related to the ISO setting. (Note: even though the without-flash-shot at the same aperture and shutter speed is waaay underexposed.)
So, it seems to me that the flash metering is working purely on the peak light level - without any decision regarding its positioning, or relative size with respect to the whole image
And the ISO rating, mysteriously, has a great effect.
(If you get some person in a group photo with a shiny brooch or badge - that will throw the whole exposure completely off).
On the one hand, it's nice to be able to say to the camera "please don't blow out the highlights" (Nikon have this setting for their newer flash systems, as far as I understand). That's a real boon for weddings (where it's so easy to blow the detail out of the dress completely). However, for general shots, there's something curiously amiss with reflected items
Might try and put up some of the images later...
Matt
(For gallery, tips and links)
Perhaps some have bore logic, but mine is old and cheap. (works great in the studio, but of course has its limitations).
Anyway, every flash read about F2.0 regardless of what showed up in the resulting image.
Anyway, Just playing around in A mode. Using the AF540FGZ
I did not change the camera settings in any way. Left that in manual mode and just let the flash adjust for its settings. (In A Mode, the camera should have little effect anyway).
Obviously consistent scene, on a tripod, blah, blah, blah...
Sorry for the odd notation below... The meter has numerical F stop readings, then a digital dial like area for partial stops in between. I've rounded to the nearest 1/4 stop.
I did each flash a few times to make sure the meter reading was consistent. In that regard, for each particular setting the metering was consistent for the flash. So once you figure out what setting to use, the flash will give acceptably consistent results.
Flash setting: Meter Reading:
ISO 100:F2.8 --> F2.8 1/2
ISO 100:F4.0 --> F4.0 1/2
ISO 100:F8.0 --> F5.6 1/2
ISO 100:F16.0--> F11.0 1/2
ISO 100:F22.0--> F16.0 1/2
ISO 200:F2.8 --> F2.0 3/4
ISO 200:F4.0 --> F2.8 1/2
ISO 200:F8.0 --> F5.6 1/4
ISO 200:F16.0--> F8.0 1/2
ISO 200:F22.0--> F11.0 1/2
ISO 400:F4.0 --> F2.0 1/2
ISO 400:F8.0 --> F4.0 1/2
ISO 400:F16.0--> F5.6 1/2
ISO 400:F22.0--> F8.0 1/2
ISO 800:F8.0 --> F2.8 1/2
ISO 800:F16.0--> F5.6 1/4 compare to F22 below...
ISO 800:F22.0--> F5.6 1/2 (yep, fired these six times each to be sure.) WTF?!?!
ISO 1600:F8.0 --> 2.0 3/4
ISO 1600:F16.0--> 4.0 1/2
ISO 1600:F22.0--> 5.6 1/4
If the flash does internal calculations for PTTL anything like it does in A mode, I have little wonder about its behavior.
I did not have any reflections, but may give that a try. I'm just out of time to play tonight and thought I'd post my findings to see if anyone else finds them interesting and/or to see if they get consistent behavior.
There will be a difference between reflected light and incident light, especially if the scene is not uniform.
You might get away with metering the P-TTL if you enable red-eye reduction The camera will separate the preflash and main flash by about 1 second (might be enough to get your meter to read the second flash).
(Think I'll try and put some of the reflection results on the web later)
Matt
(For gallery, tips and links)
Here's some more information and test results:
http://www.photon.me.uk/pu/k10dpttlreflect/results2.html
and http://www.photon.me.uk/pu/K10DFlashReflectionTestResults.zip
(The first shots I've lost the details for )
Most of the tests were done with a tethered AF540 - this is helpful so you can move the flash and camera independently and yet maintain full P-TTL control.
My observations:
* When there is a reflection the body will only let out the smallest amount of flash. The tests with manual control and 1/64 power illustrate that the change in exposure due to ISO rating is not a change in the metering, but a change in the effect of the smallest amount of flash that the gun can deliver. Effectively, I am too close
* Interestingly, although the manual power can be stepped down to 1/64 on the AF540, the camera can let out less than 1/64 when under P-TTL control. Would have been nice to have the lower rates manually selectable too!
* It doesn't matter where the reflection occurs, or really how large it is - anywhere in the frame will cut the exposure. (I didn't try reflecting out of frame though...)
* Most strangely the evidence points to the fact that the cut off for reflection is not metered on the preflash . For the last two results (6624 & 7625) I set the red eye reduction on, let the preflash fire and then moved the flash for the main burst. It showed that the preflash metering has nothing to do with the reflection issue It is almost as though there is some "cut off" control during the main exposure.... and I'm very, very puzzled!
EDIT: On the K10D I've just noticed that the red eye reduction works differently to the way I remembered on the *ist-D. The K10D fires THREE flashes on red-eye: 1) the red eye, 2) the preflash, 3) the main flash.
Consequently this makes the above statement untrue.
If you have any other thoughts to add to this then do let me know
Matt
(For gallery, tips and links)
Overall it seems the best approach with P-TTL is to bounce the light and therefore cut the probability of direct reflections (maybe this is why I generally get excellent results in practice from the AF540 )
I'll add a couple of tips onto the sticky note...
Matt
(For gallery, tips and links)
At the risk of appearing as though I'm talking to myself... Matt
Keep going Matt, you are doing a great job for the benefit of many.
Bodies: K20D (2), K10D, Super A, ME Super, Auto 110 SLR, X70, Optio P70
Pentax Glass: DA* 300, DA* 60-250, DA* 50-135, DA* 16-50, DA 70 Ltd, FA 31 Ltd, DA 35 Ltd, DA 18-55 (2), DA 12-24, DA 10-17, M 200, A 35-70, M 40, M 28, Converter-A 2X-S, 1.4X-S, AF 1.7, Pentax-110 50, Pentax-110 24
Other Glass: Sigma 105 macro, Sigma-A APO 75-300
Flash: Metz 58 AF-1 P, Pentax AF160FC ringflash, Pentax AF280T
Add Comment
To leave a comment - Log in to Pentax User or create a new account.
5866 posts
21 years
Sussex,
UK
I've spent a little while setting up a reasonably controlled experiment for the dreaded P-TTL on the K10D and AF540-FGZ. I've tabulated the results into an HTML page complete with histograms for easy comparison. However, I've not yet completed my analysis.
What comes out straight away are the following points:
* Underexposure is related to ambient light + ISO + aperture.
* The flash compensation on the gun works as expected at ISO100, but as you increase the ISO it has less and less effect.
* Using "A" mode is pretty much consistent - the P-TTL is the dud. (Also shows that the flash gun has plenty of power at all ISO ratings for the test scene)
* Using "M" mode with P-TTL flash gives excellent results AFAICS
* Camera exposure compensation has odd effects
It's almost as though the flash gun thinks the camera is always in "Auto ISO" mode, even when it's not.
I'm happy to discuss the results with anyone that can be bothered. Hopefully I can get this mess back to Pentax and help them sort it out once and for all!!!!
A slow, slow URL link is here: http://www.photon.me.uk/pu/k10dpttl/results.html
(It's slow because there a couple of hundred images to download )
And a prezip'd package: http://www.photon.me.uk/pu/K10DPTTL.zip
(Better to download the zip file, unzip to a directory and open the "results.html" file )
Matt
(For gallery, tips and links)