Recommendations on flash for stacking?

[rjlittlefield]

Here's another indication of how far the color balance shifts.



This is using the raw converter in Photoshop CS5 Extended, Version 12.0.4 x64. I've manually tweaked the settings for temperature and tint so as to match histograms for frames 5198 and 5214, the same ones discussed earlier.

--Rik

[Charles Krebs]

That thumbnail page demo really shows it!

Seems like it might be worthwhile to try and ascertain that the the camera sensor/electronics are not the culprit. Do you have a second camera you could sync with the T1i and shoot something like a white wall (simultaneously with the single flash) to see if they both show the same shift?

[ChrisLilley]

What's interesting in the data I'm showing here is that the overall exposure is consistent even though the color balance shifts. In one frame the RGB values are [189,188,186] while in another frame they are [188,190,189]. This is averaged over the same 51x51 area in both images. It's the change in RGB ratios that is surprising.

Two things strike me.

One is that the shifts - by one to three units - are small in comparison to the resolution available in an 8-bit per component system. Computing hue angles based on such measurements (where the saturation is so low) could well be ill-conditioned.

I wonder therefore if comparing with 16 bit per component TIFF rather than 8 bit per component (with lossy compression and colour subsampling) would be instructive.

The other thing is that I assue that auto white balance is not being used here? Because that could also cause shot to shot variations depending on the content of each frame (including false colour in out of focus areas).

[g4lab]

The color temperature of the xenon arc varies with all sorts of parameters but in particular with the current density and probably with the rise time of the current pulse.

This characteristic is made use of for the purpose of getting more UV radiation out of the flash, especially in flash tubes and arc tubes made of quartz which are able to transmit the UV. When they are pushed hard the UV radiation gets into the ozone producing range. Some short arc and flash tubes have doping to prevent the or reduce the formation of ozone.

It seems pretty reasonable then that there could be variations in the color temperature since the amount of UV and probably other wavelengths at the blue violet end of the spectrum can be varied.

On the other hand E G & G now Excelitas have successfully promoted the guided arc xenon flashlamp as a highly repeatable (around a couple of per cent variation flash to flash) continuous duty flashlamp, when the power supply is properly designed. These lamps have multiple trigger electrodes that go through the envelope. These are used for all kinds of analytical purposes and are pretty repeatable. If you have ever played with a StroboTac it is that kind of bulb. Similar lamps are marketed by Hamamatsu.

http://jp.hamamatsu.com/products/light- ... ex_en.html

http://www.excelitas.com/downloads/app_ ... 071004.pdf

http://www.excelitas.com/downloads/DTS_1100Series.pdf

http://www.photonics.com/Article.aspx?AID=44487

These lamps and their power supplies appear on ebay all the time and someone could probably use them to build a very repeatable flash for stacking.

I probably could help if somebody was interested I might have some of this stuff in my basement.

[ChrisR]

It was the variation from exposure to exposure ( all manual, same settings) which was the problem. There was certainly some control of power through duration. They (Elinchrom, at least) made a big thing of being able to do short flashes. It was a bit of a gimmick for a while - flicking hair and splashing champagne.. Perhaps that was the start.

I see one at least of today's manufacturers of studio equipment making a big play of good control, at a price , $12500 for a Broncolor power pack. They do have more space for better-controlled switching power supplies, in a box that big .
There's control of duration to what looks like 1/8000th and (do they means an ability to separately?) control colour temperature - but I haven't seen them telling how, exactly.
1000:1 power output range - is that all in one head? A bit much for voltage control alone.
http://strobist.blogspot.co.uk/2011/12/ ... fight.html
Some numbers here: ( I suspect "1/12,000th" should be "1/2000th")
http://www.broncolor.com/broncolor/prod ... D=Products

A fellow whose name appears everywhere is Paul C Buff. He seems to be connected with Alien Bees, which come on for some serious colour-temperature criticism, but they are downmarket.

[ChrisR]

Here's a few pictures put together in Photoshop CS-2 as a "contact sheet" with no captions - saves space.
They were Jpeg source images ( not preview images).
(They're mostly sequential, it appears they're showing as number 64, then 0 to 62 )
All the same setting on a Nikon SB-900, about 1/32nd + 0.3
Canon T3i/600D, set to "flash" colour balance.
1/200th, Iso 200, no highlight or shadow control, sharpening or any of that hokum turned on, in the camera. Mirror lockup.
These 64 were taken in 3 minutes so 3 seconds apart.
I can check the other 180 if you like, they're all suitably boring too. .

I wound up the saturation to 100 on this image, and it looks OK I think.
The file linked above is without that.

Got another flashgun, Rik?

This is pure conjecture, - I wonder if
your electrodes have degraded in some way which changes the way the spark comes off ,
or if the triggering coil/electode is misplaced,
or of course if its voltage pulse is variable.
Could it be you're getting about the right amount of charge ( coulombs) but with a different duration?
Woodworking router, 30,000 rpm, 1/20,000th of a second.. that's around a millimeter at the tip of a half inch cutter ???

[DQE]

Very interesting thread, with much useful information.

---------------------------------------------------

I wonder if the heavier-duty/continuous duty flash guns sold by Quantum would work in a more stable fashion? Since they can be used at high-output in a rapid, continuous duty cycle without overheating, perhaps they would also provide more stable color balance, etc. Just a conjecture, of course. They are more expensive than Canon or Nikon flash units.

Here's a quote from their web page:

"Unlimited rapid fire of full power flashes."

and

"Flashtubes tested to 250,000 continuous full power flashes in a 1400 hour test"

Links:

http://www.qtm.com/index.php?option=com ... Itemid=301

and

http://www.qtm.com/index.php?Itemid=197

They make a radio-controlled and a "basic" (non-radio-controlled but cheaper) model of the Trio.

I've been committing technolust for one of these units ever since I burned out the flash tube of my Canon flash unit while chasing an ant around my yard. I then read the manual and learned that one must allow for cooling time after not too many full or nearly full-power flashes...an expensive lesson.

[rjlittlefield]

One is that the shifts - by one to three units - are small in comparison to the resolution available in an 8-bit per component system.

It's true that the shifts are small in comparison to the +-0.5 resolution of a single pixel. But they're pretty large in comparison to the +-0.01 resolution of the average of 51x51 pixels, which is what the measurement reported. The differences are easily visible in the histograms and show up in the raw converter also, not just in the JPEGs.

Do you have a second camera you could sync with the T1i and shoot something like a white wall (simultaneously with the single flash) to see if they both show the same shift?

Now there's a test that cuts right to the heart of the matter!

The test is harder to carry out than I had anticipated, but here are some illustrative results. Canon T1i on the top, Nikon 5100 on the bottom, both with an adjustment layer added for Saturation=97. 1/32 power on the Canon 580 EX II flash, directed at about 45 degrees to a photo gray card. Dark room, 1 second shutter speeds, Canon being clocked by a StackShot with 2nd curtain sync, Nikon triggered by hand on hearing the Canon shutter open.



I've tried a variety of other setups too, and I'm pretty well convinced that both cameras see the same shift.

Could it be you're getting about the right amount of charge (coulombs) but with a different duration?

That seems like a good possibility. Even with everything else stable, I wouldn't be surprised if there's some randomness just from gas dynamics in striking a 1/20,000 second arc.

Got another flashgun, Rik?

Yes, several. What I don't have right now is time to continue messing with this problem.

--Rik

[rjlittlefield]

Of course I underestimated my ability to be distracted by an interesting problem.

While I had the cameras set up, I checked out a few other flash configurations.

First was the Canon T1i's internal flash, running on automatic exposure with what must have been low power because I was bouncing the flash off a white wall at about a foot. That produced small differences in color balance, easily visible in the histograms and by cranking up the saturation in Photoshop.



Then I tried an ancient Pentax AF200T flash, running in manual 1/8 power. Again, small differences in color balance.



And finally, I tried the Nikon D5000's internal flash, as observed by the Canon T1i while the Nikon was auto-exposing itself. (Sounds kinky, eh?)

To my great surprise, the Nikon's internal flash showed no detectable variation in color balance over the 18 frames that I shot before a memory card filled up and I quit.

This is with the same Saturation=97 setting that I've been using with other tests. The Canon camera was custom white-balanced to the Nikon flash based on an earlier flash, not shown here. (Without the custom white balance, the Canon sees the Nikon flash as being a bit blue.)



I find this to be a definitely intriguing result. Now I'm inclined to try YongNuo's YN565 and see if I can get the same handy auto-exposure as the Canon 580 EX II, but without the variation in color balance. But that will have to wait for a while -- other things to do right now.

--Rik

[DQE]

Sounds like we're approaching a conclusion that one needs to dump one's Canon stuff in favor of an all-Nikon system! The intriguing Nikon 800E would fit into this theory very smoothly (grins).

Now if Nikon would only come out with a macro lens that is similar to the MPE-65...and if only I would win a major lottery!

Quoting from a Beatles song:

"Money don't get everything it's true
What it don't get, I can't use
Now give me money
That's what I want
That's what I want, yeah
That's what I want

[Craig Gerard]

Rik,

What type of power source are you using with the 580EX II?

I think we need to isolate possible inconsistencies with your particular 580EX II before drawing broad conclusions.


Craig

[rjlittlefield]

What type of power source are you using with the 580EX II?

Freshly charged Eneloop AA's. Recycle time < 3 seconds for full power, tested here at 1/32 power with constant 10 seconds between.

I think we need to isolate possible inconsistencies with your particular 580EX II before drawing broad conclusions.

No broad conclusions here -- except perhaps for the usual one that the world is a complicated place.

--Rik

[Craig Gerard]

The following attempt at discussion participation may be entirely irrelevant....

I took 15 images (10 seconds apart) of a sheet of white polystyrene (at 4') using a Canon 7D (set at Manual, WB Flash) with hot shoe mounted 580EX II (set on 'M' 1/32), powered by a set of fully charged Sanyo Eneloop 2500 mAh AA batteries.

The RAW images were converted to tiff files and imported into Photoshop (Automate>Load files into Stack). All layers were selected and (Edit>Auto-Align Layers) was initiated. A single, new Adjustment Layer (Hue/Saturation) with saturation set at +97 was placed above all the layers. Each image layer was then turned off one after the other. The intention was to monitor the values of a designated position. The 'pixel' changes colour for each displayed layer when doing this exercise. The others most likely do also, but I was only concentrating on the single 'pixel' at the intersection of the guides.

Similar results are displayed when (Edit>Auto-Align Layers) is omitted from the exercise.

Here is an example of one layer (cropped screenshot, maximum PS zoom):





Craig

[rjlittlefield]

Pixel-peeping is not a good way to look for color shifts because you'll get overwhelmed by the noise. Try looking at just the histograms in Colors mode like I've shown above. Those summarize all the pixel values. What you'll be looking for are relative shifts in the peak locations, for example in one image the peaks may all be perfectly overlapped, but in another there's a blue band on the left and in a third there's a green band on the right. These histograms (same as shown above) illustrate such shifts.



It would have been even better if I had thought to click on the little triangle with the "!" in it. That would cause the histogram to be updated from the full resolution image instead of a lower level form where there had already been some averaging done.

--Rik

[Craig Gerard]

Histogram including Hue\Saturation Adjustment Layer influence for selected separate layers taken from the example in my previous post.

From the left (our left...providing no-one is reading this upside down).
First image of 15, second image of 15, last image of 15.




*When I turn off the Hue\Saturation Adjustment Layer there are minor differences in the histogram when comparing all layers.

In same order as screenshot above.



Craig

*edit: added additional information and screenshot