Timescapes - Digital Timelapse Discussion

So, for shooting sunset time-lapses or anything where you transition from day->night, do you guys manually ride exposure (say, adding one more stop of exposure for every five minutes of sun drop once it nears the horizon) or do you use auto-exposure with a spot-meter on a given target?

I was thinking last night as I was shooting near sunset (will post the video up later this week, still needs some work - otherwise I'm happy with it), I stopped shooting before the light dropped too much, because I was set in manual exposure, and the 2s interval didn't give me enough time to bump the exposure and look at the in-camera meter... Anyhow, I was thinking that if I spot'ed on a neutral building that was pretty much dead-center, and then did a +1 or 2 EV adjust, I could auto-all the way into night.

Does anyone do it that way?

tia!

!c

Shooting on Manual give you more consistency in the final exposures so you spend less time tweeking each exposure in post. When I shoot Manual I just keep an eye on the histogram and gently alter the exposure as required. Shooting on Auto works fine you you often need to dial in some exposure compensation along the way as the metering will be fooled by the bright lights at night so you often need to dial in some overexposure, but use the histogram if you can. Using Auto exposure generally results in more exposure variation so needs more tweeking in post. I shoot RAW and find it rlatively easy to tweek the shots but it still takes time to do.





JJ

Basically any time you introduce the 'camera's brain' you risk variations... Some basic measures that can help are matrix metering (or spot metering on a position that won't change 'that drastically' like you mentioned), covering the eye piece so the meter does not catch stray light, and crossing your fingers! If your willing to babysit 'Riding the shutter' with light meter in hand as you mentioned is another good method and easy to take the 'steps' out in post. IMHO the ultimate solution is to take the change in light in manual chunks and fade them together in post like this

Beautiful shots, JJ.

IMO, what is really needed is some type of "ratcheting" mechanism for the shutter. What it would do is increase or decrease the shutter time based on in-camera or other metering, but would never move in the opposite direction, which is what causes flickering. In other words, it's very much like the ratchets on a roller coaster when you are going up. You can go up, but never down.

How to accomplish this, though, is the question. If a camera's firmware could be hacked, that's one way. Another possibility would be to use something like the Time Machine, but have the exposure times controlled by a stand-alone light meter. The problem with that method is the light meter being thrown off by someone walking past it, etc.

Anyone else have thoughts on this? I'd really like to figure this out.

Even if I need to use one of those compacted cameras with the CHDK hack or whatever. I want to be able to shoot these night-to-day or day-to-night shots.

I read an article about the timelapses done in the movie 'Nixon'. They were those smooth transitions from dusk to dark. Very beautiful.

The article said that the filmmaker was doing some tests, showed them to Oliver Stone and Stone wanted them to be smoother exposure transitions. The filmmaker got ahold of a prototype Arriflex (I think???) and had a custom metering system made. The meter would read off of a grey card that was out of the frame and in some consistent light (I'm thinking it was in the shade or something..I will find a link). Anyway, the meter coming off of a neutrally lit object that did change illumination as the sun went down gave a much smoother transition of exposure than just using a TTL meter.

The science/technology sounded amazing...lemme look for the link.

Here's the article

I was considering something like that for my automaton, well, after I get all the motion stuff dealt with first *grin* - I was figuring I could use a photocell or the like to meter overall light, and then simply adjust the timing for 'bulb' shooting mode. That would be the fairly easy part, the harder part is applying the effect of aperture and any filters one may have on.... It suddenly become more like trying to implement what the camera already has built-in *laugh*

!c

That's really cool. Wouldn't be terribly hard, either, to build something if you've got your own intervalometer (i.e. home-made) to meter off of something not visible to the lens. Heck, don't some of these new spot-meters hook straight up to a computer?

!c

Ummm...an intervalometer doesn't do any meter reading...it's only a timer.

IMO, I don't know anything that's made to meter off of something that's NOT in the frame and give that exposure to the camera to use. My knowledge is limited to Nikon DSLR's and some other brands. Dunno about the film industry.

I think you mis-read my statement - I'll re-state: if you have a home-made intervalometer (e.g., one you've made yourself), it's not a big deal to add a photo-sensitive device to it and then adjust your exposure time (in bulb mode) according to a light reading.

!c

Thanks

Yeah, I definitely mis-read your statement. If you can figure out something like is described in the article, it would be commercially viable I would think.

I'm not smart enough for such shennanigans.

Totally facinating... How on earth could you hack an external meter reading from an into a DSLR? Now we are talking...

I know guys at Geo. that daisy-chain cameras (also at S.I.) for shooting sequences of up to 60fps with DLSR's. I wonder if you could daisy-chain two cameras, one getting the meter reading and setting it for the other camera. All you'd have to do is have the metering unit pointing at a grey card, then dial in the apprpriate EV compensation for the shooting unit. As the light dimmed, your cameras meter would adjust as per the ambient light. Might require some testing...some camera dissasembly...some tinkering...too bad I just started a teaching gig and barely have any time shooting for my own project!

I have been thinking about how to do such a hack for about a year since I found that article. I wonder if that Mitchell unit could be adapted for a DSLR?

I did make this timelapse right after reading that article about the Mitchell unit. I shot it on Apereture Priority, and noticed that I saw the same frantic 'speeding up' effect on the clouds that the author saw in the cars in front of the capital.

At the beginning of the clip the clouds slow down (0:11), and at the end of the clip the clouds speed up (1:03) as the effective ratio of shutter time / cloud speed changed. I was mesmerized by it for a long time. Haven't tried to replicate it again.

http://www.youtube.com/watch?v=rWz5WMUkPdk

Ok, so here's how I'm thinking of doing it, considering I've already got the code and the controller and such to do bulb-mode time-lapses, it's not a big stretch for me. However, I _must_ get the motion control done first (just got the 2d CAD done for a new pan unit, that'll mount on the 55" linear slide I just ordered, and move as little as 0.01 degrees vs. my current 0.07 design) -- and I'm going to the desert for a week in about 10 days. *sigh* so little time, and so much of it spent on a bad prototype.. Live and learn!

Anyhow, so here's the gist:

Take a small tube, probably PVC, paint the inside with the ultra-flat black camo paint (we used this a lot with home-made pinhole cameras, we don't want any bounce inside of it), place a small lens at one end, and find a decent focal point - and put a photodiode there. I have a bunch of plastic polaroid lenses, you can salvage these out of re-sale store cameras for a couple of bucks. Make a little mount for it, Then, I'll wire it off to my motor control unit, and have the UI configure an aperture, an ISO setting, and one of four values: don't measure light, no lock (increase or decrease exposure time, based on light reading), lock low (only increase exp. time), and lock high (only decrease exp. time).

Then, the photodiode would be wired straight into an analog input (via a cable that could be disconnected) on the arduino controlling the motors. If the flag is set to measure exposure time, immediately before triggering the optocoupler to connect the pins on the camera cable (and open the shutter), the analog value is read off of the photodiode. If it's changed from the previous value, the exposure time is adjusted according to f/stop and ISO. This would be additive/subtractive, permanently modifying the exp. time for the current program, so it wouldn't have to remember what changes happened before.

The pro's to this technique:

* it's easy, and costs only a few bucks (well, assuming a whole lot about having the arduinos, controls, etc. =)
* it requires no mucking with the camera
* it wouldn't add much weight to a program, reading an analog levels takes only an ms or two

The con's:

* arduino (and the AVR line of chips) have an ADC limited to 1024 possible readings - not a lot of granularity on light levels
* well, it requires writing code and making circuits, and having the code to do everything else, and a user interface, and...
* there's no guarantee it'll be any better than using matrix metering =)

I could probably find a device that would increase the resolution, as long as it doesn't use i2c, as that won't work with my design, but I really want to get the motion control down first =)

As to it being a market-able product, I totally agree, and I know of one person that is about a year ahead of me doing anything like that. I can tell you how to make it, but it's one of those things where I'm not sure it'd be able to/would want to productize it.

BTW, if you have the interest, I'll be happy the share my current motor/camera control code for the arduino -- it's rather stable and neat. The UI code is a sloppy mess, and I doubt you'd need that anyhow. (Unless you're interested in dealing with keypads and lcds =) If you'd like to do something like this, and just don't know where to start on the programming front, it could probably help you a lot.

!c

BTW, Michael, love the video. Where were you when you took that?

!c

That is looking out my front door here in Salt Lake City. I had the computer plugged into the wall, camera powered by 12v power source, tethered to the Mac. Makes it easy to shoot a LOT of frames.

Thanks for the compliments. I need to re-process that sequence and get rid of some dust, do some cropping, etc... That sequence has some potential if I use it the right way.

BTW, I read your post about making the external meter...man I hope you make these, because I know I couldn't. I can change the oil and shocks on my truck, but making that would be a stretch.

Man, I wish I had a view like that outside of my front door! I just see heat rising off the concrete, heh.

A truck - that answers my earlier question (in the lighting thread) about what kind of vehicle the JKS links were for. I love their stuff. Going to have to get the bar pin eliminators for my rear shocks soon. Stupid skyjacker shocks keep eating up bushings!

I'll look into what it'll take to get something together for that. I figure Mumford would be all over it once he reads this, it could be integrated into his system very easily. But, just doing a napkin calculation, I could make one that had to be programmed off a computer (set up the programs in the computer, and upload them to the unit to be used in the field) serial port for less than $40 in parts. If I added an LCD for field-programming, the price starts going up rapidly. So, I could put together a kit that would require just a drill and some assembly of the parts for cheap, but selling pre-made units would drive the cost up quite a bit. (Mind you, for that price, it would _never_ be as accurate as the in-camera metering and it wouldn't hold a candle to the pro system they talk about in that article.)

It'll probably be late November/December before I get to that stage of my project, assuming my day job stays on-schedule, if I start getting behind on my projects there, I'll be pulling 80+ hours in the office. I'll keep ya updated as I go.

!c

So how about harvesting the data off this unit and running it through an intelligent filter of sorts that determines the exposure on the camera? We're talking laptop here but if the USB SDK is available for nikon/canon bodies this 'should' be in the range of possibilities... Hmmmmm Most bodies have a 1/3 eV granularity so that should produce a smooth transition from day to night.

I'm thinking that maybe even doing it prior to the exposure is a waste of time/energy.

Now don't get me wrong, I'm all for getting it correct in the camera, but it seems to me that any type of de-flickering software ought to be able to make a smooth transition from day to night. Shooting the camera on an auto setting, shutter priority for example, would let your camera adjust the exposure as the sun went down, and then a piece of post production software can smooth out the jaggies. Seems that I've actually seen a piece of software that will do that...must do a search.