Timelapsing the Total Solar Eclipse

[geoffsims]

Hi all,

While there is already a general TSE photography thread, I thought it may be useful to begin another, specifically discussing time lapse. Considering time lapse is all the rage now, is anyone going to attempt this? Has anyone seen any *good* attempts at this in the past? I certainly haven't.

It is the "holy grail" on steroids; the light change happens over a much shorter time frame. I am hoping we may be able to determine the best way to tackle this, and I begin with a few thoughts:

1. It seems like a no-brainer to use specific eclipse camera control software (i.e., Orchestrator or Maestro). These can control the camera via USB/serial cables, and have the ability to scripted with events relative to the eclipse contact times. Effectively, you can predetermine your bulb ramp and script it exactly as you require.

2. How to tackle the timing? The totality portion of the eclipse is so short compared with the duration of the partial phases on either side. If you shot every second for instance, and played back at 30 FPS, partials would run for 2 minutes either side of the 3 seconds totality (VERY rough calculation assuming 1 hour partials either side of 90 second totality). Although not physically realistic, could it be worthwhile shooting at a lower cadence throughout the partials, in order to give a bit more time to totality during the final sequence?

3. I present some tests I ran during TSE2008 in Mongolia. Not intended as time lapse research, they may indeed help. I shot some images at various timess (relative to 3rd contact: mid eclipse, C3+10s, C3+20s, C3+30s, C3+1m, C3+2m, C3+3m, C4+4m) at FIXED exposures. For the curious, I was using a Canon 400D, 17mm lens at f/4 and ISO 100. The exposure times are 1/30s and are attached to this message. They may be useful in probing how the light changes at various times, and how to tackle this. All images were shot in RAW and converted to TIFF then JPEG using default settings. Of course the same happens prior to 2nd contact, in reverse order.

4. It is worth noting that using the above camera settings (17mm, f/4, ISO 100) an exposure of 1/4s to 1/8s is optimum for the totality eclipsed Sun & sky portion of the photo, while exposures of 1s to 4s light up the landscape sufficiently. Usually, I combine a number of exposures to produce a nice landscape shot during totality. Similar HDR techniques could be employed for better dynamic range during totality, but timing could be an issue.

Thoughts, anyone?

cheers
Geoff

[bartman]

Hi Geoff,
I certainly am no expert,,,,,,,but recently read a post (http://www.iceinspace.com.au/forum/showthread.php?t=95162&highlight=timelapse) on IIS about TimeLapses. Had a bit of a play with the free trial and love it!
I think that it might be useful in a situation like this.
You could adjust ( either manually or computer controlled) the iso, fstop or exposure. Then using the LrTimelapse program to smooth things out.....( see some of the tutorials on how they do this......it seems ....easyish)

As for the timing, I think you could use eye balling the pics and adjusting the various controls accordingly ( assuming you are using liveview or something similar). You have all the times of the eclipse, so use that in conjunction with the live views and make a guess, ......I guess?:confused2:
If you are after precision......ignore the above heheheheheh:rofl::thumbsup:
As for HDR, I think the same rule would apply.....except...as you said , time would be a factor.
It would be a great project......none the less
Maybe get a few( :eyepop: ) cameras doing the different options.:D
HTHs
Cheers
Bartman






http://www.iceinspace.com.au/forum/showthread.php?t=95162&highlight=timelapse

[geoffsims]

Hi Bartman,

Thanks for your comments. I'm familiar with standard time lapse techniques, including LRTimelapse.

However, I'm really interested in discussing specific challenges associated with a Total Solar Eclipse. In my opinion, trial & error or reviewing the Live View screen are not really acceptable solutions!

There is no room for error here.

Multiple cameras is of course a good option, if you can spare them. I am considering installing the CHDK firmware on to some point & shoots make them time lapse robots!

Geoff

[colinmlegg]

Hi Geoff,

Good thread...it's something I've been thinking about for awhile too. Still not sure if I'll go yet, but if I do I'll be driving from Perth.

You're right, the light drops very fast near the end. I think the best option for high quality (given unlimited finance) is a high resolution cinema camera with lots of dynamic range like the Red Epic or Scarlet. That would be my dream. !! http://www.youtube.com/watch?v=GZM5Fyw8iog&feature=player_embedded !!

Failing that, DSLR timelapse. The challenge then is firing off enough shots and changing the settings fast enough (as you say). Tests I've done with 5D2 + fast card (Sandisk), yields a maximum frame rate of about 1.2 fps. This is with 0.4 s continuous shutter and constant settings. The frame rate drops if triggering via the remote port...more like 1 fps. If triggering via USB frame rate is slower still.

Getting a nice smooth fall off in light over the last few minutes will be the biggest challenge, and may not be possible with a single camera due to DSLR limitations. Some things to consider-

Bulb ramping: If ISO and aperture are constant, you'll get about 4 stops of flicker free dynamic range. 1/30s -> 0.5s. I haven't done the calcs yet, but I think that 4 stops will be eaten up very quickly in the last 10 - 20 seconds before totality.

ISO, aperture stepping: This will expand your dynamic range, at the expense of frame rate. Best you can get if changing settings via USB is ~ 0.5 fps.

Fllicker: ISO and aperture stepping can be offset by adjusting bulb appropriately, otherwise 1/3 step adjustments will produce flicker. This may be difficult to adjust smoothly in post due to the big saturated blob in the frame. Exposure down adjustments are not possible once saturated. One way around this is to underexpose by more than 1/3 of a stop and always adust exposure UP.

Light Curve: Are any available for the last couple of minutes? I haven't looked yet, but I wonder if the drop is linear?

That's a quick summary of my thinking to date. Would love to hear others ideas on this.

Colin

ps. I have an automated exposure control system built around Eclipse Orchestrator + Netbook + custom microcontroller.

[bartman]

Fair enough Geoff! as i said im not an expert....good call though on the "trial and error " bit...... Eclipses dont come around that often.......


Could you corroborate with other users maybe?


Go Colin!!!! I'm sure you'll produce stunning pics:P

Would Bassnuts' ( Fred ?) doohicky work on the eclipse?
Maybe contact him :shrug:

Cheers
Bartman

[colinmlegg]

I think it'll have the same issues, Bart. The limitation is how quickly you can fire the shutter/update settings. Probably the best case scenario with a single DSLR camera is 6 or 7 seconds (24fps) of smooth footage at the centre line. As you said, 2-3 cameras synced together to do HDR is a good idea.

[bartman]

:P excuse my naiivety and spellink , but isn't that what Freds device does? Automatically adjust iso/fstop/shutterspeed/ colostomy bag drainage ( ;):lol:;):P just kidding on the last one ....oh i try to b funny)over a period of time?
Dooh I just realised ....its not a uniform linear drop off and ramp up of light. Its more exponential.

So in that case, is there maybe a bright spark out there that could code a sequence ( sorry I'm not savvy on the open source coding stuff) that would be in time with the eclipse and could control the DSLR via Freds doohicky? ( or similar device ( arduino) sorry... is Freds device based on Arduino? )
I'm guessing that the parameters would have to be from previous experiences.....so that would difficult.
I guess thats why I thought that eyeballing :eyepop:would be the go.......
As you said Colin, Dslr's aren't fast enough......Go NASA:D

Bartman....going to bed now.....
hope I didn't make to much sense:thumbsup:

[colinmlegg]

It does do that, but the issue is how quickly you can adjust the settings and fire the shutter to follow the light curve. Best case is around 0.5 fps if you're adjusting things like ISO and aperture (although Fred's using a novel approach to do this and might be able to do better than USB?). So if you use a base frame rate of 0.5 fps, you'll only get 60 frames of totality at the centreline. You could use 0.5 fps for partial and 1 fps for totality, but the step from one to the other maybe noticeable in the final footage. Maybe worth testing that assumption?

3 or 4 camera synced HDR is a good idea. That should be enough dynamic range to smoothly ramp through the final partial phase into totality and could be done in post without needing to worry about light curves and would get you back to 1fps.

[Bassnut]

Thanks for the rap Bart:thumbsup:, but no, our machine couldn't go that fast.

[Phil Hart]

Great post Geoff. I have been thinking about this a little, but at the moment am probably going to concentrate on getting some nice telescopic shots which i've never done before.

Will probably setup a timelapse but more to capture crowds and perhaps change in clouds before/after totality. Might not be able to do much more than that but will keep thinking. Definitely worth trying something.

Also worth looking at Magic Lantern software for the Canon DSLRs.. can auto expose timelapse sequences based on histogram in camera.

cheers
Phil

[geoffsims]

Colin - thanks for your insights. Do you really think it's feasible (possible?!) to use multiple cameras for the one sequence? Not only would you need identical optics, but surely with wide angle shots alignment would be a royal PITA?

I don't see any reason why the ISO or aperture would need to be modified (in general). For example, set the ISO to 400. As long as you can expose fast enough during the partial phases (?), this would make the exposures times quicker during totality. You could shoot, say, 1/16s and 1/2s exposures (with the intent of blending the landscape/sky portions) in quick succession not by changing the shutter speed (slow), but just sticking with bulb mode controlled via serial cable. From memory, bulb can be accurate to as fast as 1/30s (??? any ideas here).

It still doesn't combat the issue of the ~5-10 minutes pre/post totality. It could work out to be easiest to shoot three sequences:

1. Main partial
2. 5-10 minutes pre/post totality
3. Totality (incl +/- 1 minute either side)

And then cross fade them. Sequence #2 could be a fixed exposure, hopefully with enough latitude to allow smooth transitioning using LRTimelapse for example.

Phil - interesting idea and crowd stuff is definitely worth doing too. I am *relatively* new to timelapse, from what I have read (e.g. Timescapes forum) using ML to auto expose doesn't work that great? FYI, check a video I made of TSE 2010 in French Polynesia:

https://vimeo.com/13611368

Ignore the boring stuff (skip to about the 3:00 mark!), but it includes two time lapses taken by myself & Joe Cali. This is old school time lapse, however. Joe (the fisheye wide angle one) was using a video camera with an automatic timelapse function, whereas I was using a compact digicam (Olympus uTough) on HD video mode, and just sped it up. Interestingly the video cameras auto exposing works surprisingly well.

[colinmlegg]

Geoff, some comments below:



* I should say my thinking is based on the goal of tracking the light to yield an optimal exposure throughout. It's influenced by day-nights I've done and 3 lunar eclipses. May not be entirely relevant for a TSE.

[geoffsims]

Hi Colin,

That's impressive you have had success using 2 cameras for the same scene, and as I imagined only doable by effectively eliminating any parallax effects. I think in the context of a TSE, this additional requirement would severely limit not only the artistic composition of the shot, but also the locations at which you shoot from. Generally, it is the foreground objects (nearby, or otherwise) that make a wide angle totality shot really stunning. For me, I like to remain mobile as well (i.e., move at the last minute if the weather is dodgy) so trying to find a good shot with no near foreground objects, at the last minute, may be too difficult and probably a deal breaker. Still, it's an interesting idea if anyone is willing to try it!

OK, so 1/10s is the limit for bulb via a PC. This is probably OK, as you can choose the correct ISO to not have to go faster than this. This way, you could probably shoot every 1 second, alternating exposures (then combine each set of 2 shots using layer blends). Effective cadence would be 2 seconds during totality, yielding a 2.5 second sequence at 24 fps (for a 2 minute duration totality).

The only problem I see with my suggestion of three separate sequences which are cross-faded together, is that there is little room to cross fade, particularly between sequences 2 and 3. I would imagine you'd need at least a second or two of footage overlap to do a good cross fade.

By the way, I have not seen any light curves, but will have a look.

[gregbradley]

As an eclipse is not the same as night time darkness I think you would be able to do a time lapse.

My D800E's internal time lapse would work if set to aperture priority mode. It works down to fairly dark and then when dark it does not boost the exposure time nor the ISO enough to night time proper exposures but I am confident it would handle dim light from an eclipse.

A number of Nikon models have the time lapse function built in. You'd have to check their spec sheet.

There is a new D600 full frame Nikon coming out in 2 weeks. Its supposed to be cheap for full frame at $1500 and is 24.7mp. It will have that time lapse function.

Greg.

[colinmlegg]

Geoff:



Greg:

[OzEclipse]

Hello Colin. Long time no see... about 18 years 11 days and 8 hrs this Nov isn't it?

Interesting thread Geoff.

Colin & Geoff wrote :
"By the way, I have not seen any light curves, but will have a look.

Thanks. I had a quick look but couldn't find anything in terms of EV."

I have an eV light curve here:
http://joe-cali.com/eclipses/PAST/TSE_ASE1999/tse990811.html

It's about 3/4 way down the page and measured in eV. I took the dome off a Sekonic incident light meter and pointed it at the sky not straight at the eclipse so it's not a measure of exposure but of ambient light levels from the sky and eclipse falling on the sensor. If you want the dome on incident light values, subtract about 2eV from the readings.


Auto exposure works quite well with a very wide angle lens. The sky in umbra is pretty evenly lit. Dial in a bit of -ve compensation (0.5ev - 1eV) or else the sky will come out 18% grey.

You could also try to dial in -3eV then use an rfc to pull up the shadow detail. The -3eV will give you some mid-outer coronal detail.

Foreground detail can be brought up by stacking and masking in the landscape from numerous "under exposures."

When we were on the beach at Tatakoto, I ran the timelapse camcorder at 1 frame per 30s for about 2 hrs which gave about 10 s of footage covering that period. I started it as soon as I could get the camera set up on your tripod after we arrived. About 20 mins before the eclipse I put in a fresh battery and changed it to 1 fps for 40 mins. gives about 80s clip of the time around totality. I probably could have slowed down the first 2 hrs to 1 frame per 10s to give 30s of footage but then the evaporating rain off the lens wouldn't have been as interesting.

I think to be kind to a viewer and keep their attention you need to time it so that the ingress partial and total are about the same length of time and the egress partial is about half the length of the other two.

I am very definitely trying to record umbral passages and I'm not so concerned with coronal detail which I capture in close up in other ways. I know that Geoff is trying to capture inner, outer corona, sky and land.

The camcorder timelapse of the 2010 eclipse and another captured on a film SLR can be viewed on this page :
http://joe-cali.com/eclipses/PAST/TSE2010/joe.html

and another of the 2008 also with a film SLR can be viewed on this page :
http://tinyurl.com/9zm637o

Even though the landscape is dark in these cheap machine jpg scans off colour neg film, the landscape can be easily brought up by stacking two landscapes and adding the pixels. Joe

[geoffsims]

Colin: For the landscape, based on my previous results, landscapes typically need 1-4 seconds (depending on how much detail you want) at ISO 100, f/4. Obviously can convert this to whatever f/ and ISO you are interested in.

Joe: I had never seen your light curve. Very nice. What was the altitude of the Sun during that eclipse?

Before thinking too much more about this, I am going to run some of my own quantitative tests using my past sequences. I will produce some light curves of my own, in ADU, of the sky and landscape.

[OzEclipse]

I'd have to look it up but at a guess - 40-50 degrees. I was in Bucharesti for totality.

[colinmlegg]

Hey Joe, yes it's been awhile since that Puno -> Cusco train ride. I see you've done a few eclipses since ;). I've been back to the altiplano a few times myself ...had a lasting impact on me!

Thanks for the light curve. I need to spend a few hours reading your site. Lots of interesting stuff there.

I'd love to see the eclipse in Cairns, but I've left my run a bit late so my only option now is to drive up and rough it in my troopie. Will decide this month. Like Geoff, my focus is wide angle and video/timelapse. Ideally I'd like to shoot from an elevated spot overlooking the ocean and preferably free of people. I have a fulldome fisheye and multi-camera system that can cover the entire sky. Any thoughts on where I could locate myself?

Colin

[geoffsims]

OK, I produced a light curve from my data. It's not a conclusive scientific test, but it's not bad considering I had no intention of producing such results. I used the original set of images I posted, plus a similar set from pre-totality (the original set were post-totality). Recall these were fixed exposure so the relative intensities are correct. In this instance they were 1/30s, f/4, ISO 100, 17 mm lens on Canon 400D.

I used DCRAW to convert the raw images with no white balance, scaling or Bayer matrix interpolation (essentially outputting the raw data points in ADU at each pixel position). I then converted the images to FITS format, and collapsed them into the various colour channels (R, G, G, B) using Python routines (NB: there are two green channels as a result of the Bayer structure). I collapsed the colours because the response from each colour filter will be different. Then, I picked a few regions of interest and examined them.

Results are attached, I used only the green filter and units are in raw ADU. Some explanation:

- Whole image (median) is just a global median pixel value of the entire image.
- Sky (median) is a median of a 50x50 pixel region of sky, approximate 50 degrees from the Sun
- Land (median) is a median of 50x50 pixel region of the landscape.
- Around Sun (mean) is the mean value of a 60x60 pixel region centered on the Sun

As evident from the initial set of images, they are extremely underexposed near totality. From the ADU counts, you can see it flattens out just above 250 ADU which is probably an indication of the bias level of the camera. So data here is not really valid. It is also worth pointing out that the "Around Sun" curve data is probably skewed too as many of the included pixels would have been fully saturated, particularly a few minutes before/after totality.


Now to spend a bit of time examining this data, along with Joe's data, and the implications for how to tackle this project!

[OzEclipse]

After Qosco, I spent another 3.5 months hobbling around the continent in that leg brace. Had a great time. After the knee reconstruction I went back to central and south America for another 4 months in 1998. As you say, easy to get hooked. Anyway - off topic!

Re-eclipse and fish-eye. Most beach sites have trees lining the beach. Because of the incoming tides and equipment type project like yours needs to be high up the beach where the trees will be covering half the sky and your FOV so you need a site off beach where you can see most of the sky. There is a lot of open ground by the road side in the areas 1km around the skyrail base station.

If you want something more secure where your gear won't be disturbed, I am setting up an observing site on a 40 acre farm near Port Douglas. Cost will be $50-60 per person and you can move in to set up anytime in the day or two before.

This site will have a lot of very experienced eclipse chasers, Astro club members and a Sydney Observatory Group and their web cast coming off it.

http://joe-cali.com/eclipses/PLANNING/TSE2012/Port_Douglas_Obs_Site/


You can camp there on Tuesday night at no extra charge however there are no facilities except portable toilets. However, while camping out the night of the eclipse or at least being out most of that night is unavoidable, I'd recommend against camping the whole time. It's oppressively hot and humid in November. When I was there mid-Nov 2010 it frequently rained at night and of course lots of insects. Really unpleasant!

I might be able to offer you some various options for better accommodation depending on your budget. Call me on 0429 233 616 if you are interested.

cheers

Joe

[OzEclipse]

Sorry, I left out one character in the URL in the previous message.
It's : -
http://joe-cali.com/eclipses/PLANNING/TSE2012p/Port_Douglas_Obs_Site/

or

http://tinyurl.com/9lzmvqq

And I should say that anybody not just Colin who wants access to this type of field site arrangement is welcome to contact me. So far it's mainly been bigger tour groups and people with more complex equipment set ups who have requested access.

Individuals with just a camera and tripod or a pair of binoculars should easily find somewhere for free (see the site survey articles on my web site) and should probably save their money.

Cheers

Joe

[gregbradley]

Hi Colin,

Minimum EV steps are 1/3rd looking through the menus. I have used it attempting a day/night transition and its reasonably good up to a point but then it does not do long enough night shots at high enough ISO to be useable for that.

Perhaps with auto ISO on that may help it as well. As I recall that is what I also had going.

Greg.

[geoffsims]

I thought I should add that one interesting possibility is using compact Canon digicams for time lapse. Using the CHDK firmware hack, which runs off your SD card, allows access all sorts of features that are otherwise not available, including intervalometers (via custom scripts). There is the possibility to write some quite intricate scripts, although at least a few already exist that dynamically change the exposure time to get the optimum exposure, while keeping the aperture/ISO constant. I haven't run a test yet, but considering the simplicity, it appears to give good results.

The main benefit I see here is that you don't really have much to lose (since they are presumably not your primary camera[s]). Additionally, you could do things like point them away from the Sun (to capture people operating equipment), or any other interesting angle you can think of, while reserving the primary DSLR camera for the full frame landscape sequence.

Incidentally, I have two quite nice newish compacts from the last two David Mailin competitions, so I think I may as well put them to good use.

I am still thinking about the best way to do the primary sequence, but having multiple angles at least means you can edit and cut in various sequences throughout the partial phases, for example, to make an otherwise "boring" and long ingress/egress sequences, seem more interesting.

[colinmlegg]

Geoff, re. Canon point & shoots, I believe the minimum exposure step is 1/32 stop, compared with 1/8 for DSLRs. So setting the exposure to auto should produce fairly clean footage with only minimal flicker. Certainly worth a shot.

[colinmlegg]

Geoff, would you mind converting those charts to EV? I'd like to compare with Joe's curve. As you say, with a month to go I better get cracking on this!

[colinmlegg]

For those interested in Eclipse light curves, there's another from Fred Bruenjes, Eclipse Orchestrator author. Scroll 2/3 down the page.

http://www.moonglow.net/eclipse/2005apr08/index.shtml

[geoffsims]

Attached - I just took the log2 of the ADU values. By the looks of things, in Joe's plots, there is about 5-6 EV drop in light within 5 minutes of totality. Mine do not show as great a change. I'm not sure why - possibly because my exposures were too short and is below the noise level of the camera.

[colinmlegg]

Thanks Geoff. I've asked Fred for a higher time resolution version of his curve for comparison. Just eyeballing it though, it looks similar to Joe's.

I've gone back and forth on the bracketing question during totality. Leaning towards no atm, based on your images and using a 1/4s,iso 100,f/4 exposure. At iso 100 RAW (5D2), I should be able to pull 2 stops (or more) out of the shadows, which gets me to your recommended 1 sec for landscape. My concern is the highlights + horizon. Any chance I'll get blowouts along the horizon (where the sun still shines?). Should I pull back to f/5.6 to be safe, and live with darker shadows?

If I can avoid bracketing or complicated multi-cam hdr setups, I reckon 20 sec (24fps) of smooth(*ish) footage is feasible via bulb ramping. Assuming the light drops 3 stops in the last 3 minutes and totality is captured at 1/4 s shutter, 1fps.

1. Ingress Partial phase 3 mins -> C2. 1/30 -> 1/4 sec (bulb ramp). 180 frames

2. Totality. 1/4 sec (constant shutter). 120 frames

3. Egress Partial phase C3 -> 3 mins. 1/4 -> 1/30 sec (bulb ramp). 180 frames

* It's likely the bulb ramp will suffer some flicker near the start and end. Need to test this in the next day or so.

I'll run the full test soon and report back.

[geoffsims]

Colin, some good ideas! Linear or logarithmic bulp ramp?

Post or PM me your email and I will send a link (tomorrow) so you can have a play with my raw images from 2 eclipses (2008, 2010), and see which exposures give the best results. Keep in mind they were taken with a 400D. I will do the same.

From memory I could not squeeze out enough landscape light from 1/4s, but dynamic range on the 5DII is obviously better.

BTW, will serial port bulb control through EO be accurate enough for bulb ramp?

[colinmlegg]

Thanks Geoff. Email colinleggphotography @ gmail.blah

Unfortunately, PC timing is an issue for bulb ramping below 0.6 s. You need 0.001 accuracy between frames to smoothly follow the ramp and the best you can get from a PC is ~0.01. When I said earlier that bulb is only good for >= 0.1 sec, it's because +-0.01 is 10% inaccuracy. Just good enough for a single shot. 10% variation would still be noticeable as flicker though, if you strung a few of those together.

So, the way out is a microcontroller that does the timing. You can go 2 ways. 99% automatic with something like Little Bramper, or 100% auto with a custom controller like I have.

If you don't have Little Bramper or equiv, I have 2 spares from the first batch that I can lend out. It'll mean 2-3 seconds manually de-activating and re-activating the ramp within 10 sec of C2 and C3 and your ramp will be linear. I can send one over before I head off if you're interested.

The 3rd alternative is Promote Controller. It has bramping in the latest firmware. Unsure how sophisticated it is compared with LB?

And having said all that, it's likely you'll still see some flicker between 1/15 and 1/30 s. This is where even 0.001 is not fine enough. Better than the 1/3 stop alternative with AV or 1/8 in Full Auto though.

[Moon]

I've been thinking about this too. The best ones I've seen on youtube are the ones that show the crowd.
James