Timescapes - Digital Timelapse Discussion

Hey,

I have seen before time lapses show the transition from day to night or vice verse. When ever I try to get a time lapse like this it always ends in failure. When I try normal time lapse I either start or end in over/under exposure. I have tried everything I could think of to expand my dynamic range; auto metering(all modes), manual setting transition, HDR, post-shot, bracketing and many combination of these. Most end with useless flickery movies, or are impractical.

There is about 13 f/stops between a full moon lit scene and a typical sunrise, that number goes up to 21 f/stops between a new moon and typical hazy day light. Now that is probably asking too much, but I would love to end a moonlit time lapse with a sunrise instead of overexposure. Does anyone here have the answer and can it be done with a typical DSLR?


A.M

Somewhere around here there is are two videos showing a timelapse in which a device slowly changed the aperture over time. So it holds a correct exposure for quite a long time because it can increase/decrease the aperture at the same rate as it gets dark. But that isn't a very elegant solution in my opinion. And I can' find the post right now.

I'm stumpted!

Well, we had a discussion on here a while back about metering ambient light (i.e., not scene light, but somewhere off-scene, and shaded from harsh changes) to control exposure time during shooting.

The idea is certainly do-able, I just don't know how effective. I've built a prototype (light meter integrated with intervalometer) and am just waiting for both my schedule and the weather to cooperate for a sunset test.

When I've got a chance to test it and make proper documentation, I'll release everything open source - software, etc. It will take about $70 in parts and an afternoon of work - minimal or no soldering, depending on how you do it.

!c

@matt
Dang, that sounds really cool! Although I would think changes in DOF, lens distortion and sharpness over time might look odd on film. Also I would be curious about how it gets past the natural 'stepping" a typical f/stop ring goes through. I also can't imagine this being more then a partial solution because most lenses only have 7-8 f/stop ranges. That being said I would still love to get my hands on one.

In my minds eye the perfect solution would include the ability to program the change of f/stop by very small fractions. Why do I need to be confined to 1/3 f/stop jumps in shutter speed? Yes I could imagine it being confusing having 50 settings between 1/250 and 1/200 but why not while tethered? I wouldn't mind lugging around a laptop to get that kind of flexibility and if I could program the settings for each shot, allowing smooth transition I would be in heaven. A man can dream right?

@ shutterdrone: Sorry if you've already explained this, but how are you planning to control the exposure time? Via bulb mode/cable release?

I believe the thread Matt was referring to is this:
http://timescapes.org/phpBB3/viewtopic.php?f=4&t=476
Apparently there is a ball bearing responsible for stepping which he removed to vary the aperture smoothly.

I looked into what steps my DSLRs (Canon) use as compared to my compacts (also Canon) because the compacts do much smoother variations. The lengthy thread is here:
http://timescapes.org/phpBB3/viewtopic.php?f=4&t=354&st=0&sk=t&sd=a
Results were that the DSLRs do 1/8 ev steppings in both aperture and exposure time which results in noticable flicker. The compacts do 1/32 ev steps which may still give slight flicker, but much smoother than the DSLRs and easily fixed with a deflicker filter. The thread contains examples. The problem is that the compacts are "disabled" from working at night on automatic settings, implemented by a 1 second maximum limit on exposure time. Of course, compacts also don't offer the same IQ as DSLRs. Most likely, the reason why compacts can do smoother steppings than DSLRs is that they are designed to do video as well. Hopefully we will soon see some high end compacts with larger sensors and hence better IQ from manufacturers who make time lapse capable cameras - I believe Canon is the prime candidate here. Even though DSLRs are starting to do video as well it's unfortunately not implemented in such a way that the camera can do smoother variations for time lapses. We can hope that smoother exposure time stepping will become available in the future and that lenses which can vary the aperture smoothly will also be developed in the future to improve the video capabilities. I have little idea how far out in the future that would be though whereas I believe some capable high end compacts should hit the market soon - hopefully in 2009.

Yup - exactly, it hooks up to the remote connector on the camera and you use it in bulb mode. The minimum exposure time depends on the camera (mine is around 70ms) but the max is unlimited. The device works such that you define how many steps per EV (1 = full EV changes, 10 = 1/10th EV changes), so, effectively, you could get 1ms changes in exposure. It also latches high or low, so that it will only increase or decrease exposure - so, if you're shooting a sunset, it won't accidentally jump back up. Also does some change monitoring, to keep the exposure from bouncing from minor changes in light.

It's entirely based on the conversation on here a while back about doing off-camera metering in the shade to get around the fluctuations in the in-camera meter. I'm getting a little frustrated though, the complete prototype has been sitting on my bench for weeks now, without a single chance to test it out in "real world" conditions. Almost makes ya hate the holidays =)

Here's the original thread: viewtopic.php?f=4&t=326

Mind you, I've learned a lot about measuring light and all that jazz since that post, let me say, the software was a bit harder than I had intended, but it is very accurate *grin* I've written some tutorials on the light meter chip I'm using, if anyone's interested, or you can wait until I publish the entire plans. (I'm to give a talk on it on the 12th of January, so I would make the plans public after then.)

!c

@shutterdrone, flyvholm,
thanks for those links and updates on HDR.

My background is in film capture and you maybe aware of a technique/gear that is used by filmmakers to achieve this result smoothly.
(Even shooting with a dSLR , I still set exposure with an external light meter, - either a spot-meter or incident meter or both!)

Have a look at this link, the actual 'what it does' is in the second half. http://www.clairmont.com/tech_tips/lpc_90.html

In addition to that device to extend the range even more, very careful use of adding ND filters and also removing them if necessary, during the interval times, with appropriate exposure compensation is an art in itself. (spare a thought for the camera assistant who may have to set/pull filters. adjust shutter or aperture manually all day from pre-dawn to late evening)

I owned a Norris LPC-90 for a while, that I used with an Arri-S. Very effective machine, but you had to be very careful in choosing the right subject for your spotmeter reading. With a movie camera's rotating shutter, the LPC-90 could vary exposure time continuously, unlike a DSLR shutter, which uses discrete steps. I read somewhere (maybe on Kinsmann website?) that a way to produce results similar to the LPC-90 (but with a much lower resolution...) is with a Coolpix 4500. Unlike the Canon point and shoot cameras (like the G6 that I use sometimes), the electronic shutter of the 4500 can vary in such small steps (I don't have the exact figure) that it seems to be continuous. In aperture priority mode, you just have to be careful in the way you frame and choose the metering mattrix. I will have to get an intervalometer that can work with my 4500 and try it...

I'm waiting patiently Church.... still extremely intrigued!

It's an ancient film technique called riding the aperture ring. Incorporated in countless films for various effects. Nevertheless DOF changes are negligible focused to a lenses true infinity. The resolving power changes slightly and vig may creep as well but it's not that glaring with the light changes so radically.

@ shutterdrone
Wow, that sounds like a near perfect solution, very exiting! I had never though of using bulb before but makes perfect sense. I seem to be salivating over the very idea of this device, keep me posted! For now I think ill go and read that old thread.

@milapse
That sounds fair enough. I always wondered how they did it in films. What would one call this f/stop ring rotating device? Is it the sort of thing you can order or are they usually purpose built?

The type I am familiar with used to called an "aperture-driver'. It was controlled by a motion control system especially if shooting a motion T/L sequence.
Basically it is/was a stepper motor with a gear wheel engaging the aperture ring of a cine lens which could be 'driven' by the motor.

Depending on the time of year, you would 'block' (rehearse) the shot a day or so before, especially if there was moving shot with a sunrise/sunset involved.
Part of that process would be to log the exposure values of the scene minute by minute during the required shoot duration. Next those values would be used to program the moco computer to 'pull aperture' via keyframes over the duration of the shot.

As exposure changes rapidly during the early part of a dawn/sunrise then slows down later, the programming would plot these values and compensate accordingly. As micro-step drivers were used. a regular 200 steps/revolution motor could be made to have 12,800 steps/revolution for extreme accuracy when filming the shot.

The whole procedure is definitly overkill in this day and age but it worked!
Contemporary high end cine cameras (Arri 435's etc) have these functions available as small remote plug-ins controlling small servo motors.

Cool, I can hardly wait to see it! What do you use as a light meter?

Michael, I'm using the TSL230R light sensor IC from Taos that costs about $5. It basically reports a frequency using TTL that can be converted into a measure of light power falling on the sensor, from there I convert into luminance and use that as a driver into an EV calculation. A micro-controller interfaces with that, and uses an opto-coupler to connect to the camera, and an LCD and some buttons are used for the user interface.

I had talked in that previous thread about using a more simple analog light sensor, but using this chip let me get around all the analog-to-digital converter limitations I have.

A test the other night showed it was as good, if not better than, the in-camera meter for "big" changes in light - that is, I shot about 80 frames turning on and off the big overhead fluorescents while shooting a fixed subject. Tried with the in-camera meter set for full-scene, and my device only had one 'flicker' that I could tell, whereas the camera had about 15-20 flickered frames before getting stable. Looks like I have a chance this afternoon to do a sunset - hope it works for that *grin*

!c

Ah I see, a photodiode on a IC, cool!
By the way, I appreciate it very much how open source - like you tell us your secrets!

The iris ring tied to a light meter might work. it's gotta be very smooth and gradual, though.

Is there a way you could have a program adjust the exposure of RAW images to smooth the transition between shutter speeds of images that were captured on Aperture-Priority? (metered on an stationary object that won't be affected by passing clouds?)

You can adjust exposure of RAWs in very tiny increments, so it seems like you could make the transition from day to night seamlessly as long as the camera is smart about it.

Even if you have to do it manually it doesn't sound too hard.
1)Watch the movie
2)Write down frame #s during movie that have noticeable transitions from one shutter speed to another.
3)Reprocess x frames in front and behind of each bad transition from the original RAWs. You can calculate how much exposure compensation is needed by looking at the EXIF data to see the shutter speed of each image, calculate what fraction of a stop difference exists between them, and then set +/-half of that into your RAW editor.


...or am I just full of it?

Matt, that's the path I'd like to go as well. I've been wanting to make a "deflicker" that does the job automatically, taking an image sequence as input and delivering an output sequence with steps in exposure smoothed out. The output would be either tiff or jpeg depending on preference. Hopefully I'll get around to doing some work on it soon. I'd expect it to work well generally, but I've seen examples of significant flicker (~1/5 ev) where there was no change in exif at all, and the flicker appeared to be triggered by large changes in dynamic range (footage was smooth until the drastic decrease in lighting at sunset). Those examples were from a Nikon D80, I have yet to test if my Canons do it as well. If it's a general issue across brands/models this exif-based deflicker method will still leave some flicker behind that needs to be removed "manually" (the steps you described).

Fly - I see the difference between a solution such as this and the existing de-flicker apps being that you would script modifications to the RAW processing, rather than the jpeg frames of a video. Am I correct there?

I don't think that you would need the exif at all - just an analysis of the image (like a histogram) and a target weighting for the histogram. That's the same way I monitor the hardware solution when testing it - watch that the histograms stay as steady as possible through the ranges.

!c

I have always wanted a RAW processing action or script that could allow me to apply some curves across a sequence. So, for example, with a sunset sequence, the RAW processing would start 2 stops under (to tame the overexposure) and end two tops over (to pump up the waning light). I would love to be able to do that.