In two previous articles I covered the most common problems that face anyone doing long exposures. In part 1 of 2 I discussed: Poor Focus, Dim Stars (low contrast), Strange Colors and Pink or Purple glow. In part 2 of 2, I tackled gaps in star trails, and noise.
It seems like I have omitted a rather important element: how to choose your exposure settings in the first place. An astute Flickr user asked What is the Best ISO for Stacking Startrails. Good question and I realize I have not approached the question from that point of view… starting from the beginning, that is. So pay a little attention here and I will hopefully demystify that question for you.
First, the answer will always be “it depends”. Just as with daylight or indoor exposures the settings to pick depend on the conditions. Is there a bright moon? Is there a strong sky glow from light pollution? How well does your camera manage noise? How cold (or warm) is it? Are there any bright light sources that need to be managed? What is the intended result?
Most people attempt to approach a star trail the way they would expose any low light scene. But that approach is flawed. Let me illustrate the germaine elements with a story about my friend, the little star named “Drizzle”.
Drizzle Drizzle Little Star
The film in a film camera, or the sensor in a digital camera can be thought of as a piece of absorbent canvas. Imagine that the light (photons) of a single star are a stream of tiny little droplets of star juice. The brightest star – Sirius – can spurt 200 droplets per second. The faintest stars visible to the eye, like Drizzle, only manage a single drop per second. Moreover to be able to notice anything at all, we require a minimum of 20 droplets. Doing a tiny bit of math we realize that we have to collect droplets from Drizzle for at least 20 seconds. But for Drizzle to stand out in any noticeable way we need to collect Drizzle juice for about a minute – 20 was only enough for Drizzle to be discernable.
But wait. We have a collector in front of our canvas that grabs the incoming droplets and shrinks them. That collector is our lens and the size of the iris (aperture). Even when we make our collector opening as big as we can (f/2.8) it will take two collected droplets to equal up to one direct droplet. In fact, if we make our aperture even smaller (f/16) we shrink all the incoming droplets so that they are only 1/100th of their original size. If we set our collector on f/16 we will have to collect two hundred minutes worth of Drizzle’s meager output to be able to see Drizzle’s shine.
But wait… we forgot about something else much more important! Drizzle and his companions are continuously moving across the sky and each droplet of their juice passes through our collector and falls on an ever-changing place on our canvas. After perhaps 15 seconds the juice of all of our stars will fall noticeably further away – on the next pixel. Poor Drizzle stands no chance of making an appearance because he cannot spew enough juice in his brief time over any part of our canvas to leave a noticeable mark.
How then can we help little Drizzle make an appearance? Well, we can collect more of Drizzle’s juice by using a bigger collector (lens). With a bigger lens instead of collecting one droplet at a time, we can collect two or with a really, really big lens perhaps 4 droplets at a time. After all, a bathtub in the rain collects more water in a minute than a thimble will! Sadly we already have the biggest collector we could afford so how else can we help Drizzle? Answer: We will employ the last trick our camera can muster: we can change our canvas so that one droplet leaves a mark as noticeable as though it were 10 times its size. This is what is happening when we change our ISO from 100 to 1000.
Oh, I forgot to mention that while we are collecting juice from our starry friends, truly random events occur that cause droplets to appear out of nowhere and plop onto our canvas. We call this noise. The longer we allow our canvas to collect juice, the more random droplets there will be scattered all over our image – unwelcome intrusion by the sprinkler from hell. Unfortunately our sky is not completely black. Dust and moisture in the air grab light pollution from distant city lights and make the sky rain droplets everywhere. The stronger the light pollution is the harder it will be for Drizzle to stand out. In fact, when the sky becomes as bright as Drizzle we will not be able to see Drizzle at all.
What did we learn from Drizzle, besides sympathy for his plight?
- If we narrow our aperture we will get fewer noticeable stars. And the stars we do get will stand out less (lower contrast).
- If we use a bigger (larger diameter) lens we can collect more light and thus see more stars.
- If we increase our ISO we get more stars, and unfortunately more noticeable noise.
- The longer we expose the more noise there will be.
- As the background sky glow increases the dimmer stars will be overwhelmed and there will be insufficient contrast to see them.
- Eventually an exposure that is too long will wash out the sky and stars.
And the most important take away: The number of stars we can capture in an image is unrelated to the length of the exposure because the stars are moving.
That last one surprises most people. That is why I highlighted it. I see many folks trying hard to work out the right exposure based on the ambient light. But that is not the correct starting point.
And Now… The RIGHT Exposure
I hear you: “We just want to know what the correct exposure is! We really did not need to hear about your constipated little star.” True, perhaps I spilled more than you wanted to know but my little drizzle buddy hopefully made it clear that only the aperture and the ISO settings have a significant effect on how many stars will be observable in the image. Now that we know that a really small aperture will eliminate most of the stars, that a long exposure will invite more noise, and that a higher ISO will both allow dimmer stars to be seen AND increase the effect of any noise we hopefully can use those parameters to narrow in on what we need to control. Here are our goals:
- Use shorter exposures for less noise, better contrast, and less interference from background glow
- Select moderately open apertures for more light and better contrast
- If possible shoot in cooler ambient temperatures.
- Where possible select locations with darker skies.
I hear you oh impatient one. You wanted to know WHAT EXACT SETTINGS you should use. Try this:
f/4, ISO 200, 4 minutes.
That should be about right on most nights where there is half or less moon and not too much sky glow. But remember that exposure triplet is for the sky not the foreground. We can refine those settings by answering the following questions and adjusting the exposure, ISO and aperture as indicated. Adjustments to larger ISO are optional. Adjustments to a lower ISO are not… except to try and see that failing to reduce ISO or exposure time produces lots o’ noise.
- What is the air temperature (Fahrenheit)?
- Greater than 80 degrees (ISO ÷ 4)
- Between 60 and 80 (ISO ÷ 3)
- Between between 38 and 60 (no change)
- Below 32 (ISO × 2)
- Below Zero (ISO × 4)
- How bright is my sky (background glow)?
- About what you would expect less than 20 miles from a large city: ISO ÷ 2 and exposure ÷ 4, f-stop +1 (1 minute at ISO 100, f/5.6 or f/7.1).
- It is quite dark but the moon is full (ISO ÷ 2) – exposure ÷ 2
- It is dark, but the moon is 1/2 full (no change)
- Some noticeable sky glow, but I can see the Milky Way (ISO x 2)
- I sat in the dark for 30 minutes and I literally cannot see my hand in front of my face. I can only tell it is there because it blocks out the star light and I still have sensation in my fingers. (ISO × 3) or f/stop + 1.
- How many stars do we want in the image?
- As many as I see with my eyes – aperture at f/4 or f/5.6 or even f/7.1
- Plenty (ISO × 2) – an aperture at f/4 to f/5.6
- A sky full (ISO × 4) – aperture at f/2.8
- How good is my camera at managing noise?
- Really Terrible: ISO ÷ 4 and exposure ÷ 2.
- Terrible (ISO ÷ 2)
- OK (no change)
- Good to Great (ISO × 2) and increase exposure by double
Suppose we end up computing a value of f/5.6, ISO 25 and exposure 2 minutes. What does that mean? It means something is far less than ideal and throwing in the towel is in order – especially if the minimum ISO on the camera is 200. If on the other hand the series of multiplications and divisions nets an ISO greater 1600 it is better to dial the ISO down a bit. The above is pseudo scientific and adapted from observations and exposures – both successes and failures. Actual results may vary.
Ok, now I am sure an explanation is in order. Let me start with the one thing I think causes the most variability – the ambient temperature. The hotter it is, the stronger the noise. That is simply a fact of the electronic world. Film shooters have an advantage here. Film does not become noisier as it is exposed longer – it does have a different problem though. The longer film is exposed, the less sensitive it becomes. Scientists and astronomers are well aware of the temperature problem. In fact, the really serious image makers super-cool their sensors to remove as much noise as possible. How much difference does temperature make? Well, a guy named Gary Honis built a miniature refrigerator to cool his camera. He is very happy if the mini-fridge gets the temperature of his camera down to 5 degrees! Why? Well one can take a look at his charts. But I can summarize: a 5 minute exposure at 77 degrees fahrenheit had more than 2,500 noisy pixels while at 5 degrees, less than 100 pixels displayed noise! Twenty five times LESS noise! When it is exceptionally cold outside I smile because I know that if my battery survives my image is going to be that much better!
The next issue is the camera itself. The smaller and denser the sensor (the higher the megapixels) the more prone it is to noise. Sure some cameras have sophisticated processing to reduce the noise but most of the algorithms also reduce the contrast and thus the sharpness as well. A big sensor with big pixels is better for lower noise performance. Another issue has to do with the design of the sensor – the method used to collect and read out the count of photons in each of its buckets (pixels). Some methods are just better than others.
Look at the histogram. If the shots are even close to being over exposed reduce the exposure time. If things are too dark, and it is miserably cold out increase the ISO or open up the aperture – or both. If it is warm only open the aperture because a higher ISO or longer exposures will result in more noise.
To get a well lit foreground requires one of the following:
- Shooting when there is more moon (and get fewer stars)
- Painting the foreground with artificial light
- Shooting for a much longer time (and live with the noise).
- Shoot some shots at twilight to get a nice foreground and layer in the star trails.
Can I Reduce Noise by Averaging?
In a word, no. The normal method of stacking selects the brightest pixel from each shot… and hopefully those are the stars. But those bright pixels can also be noise. If there is noticeable random noise in 10 images then there will be 10 times as much noise in the finished image! It is important to keep the noise as low as possible in the first place! Averaging can be done. Something undesirable then happens: the foreground image and the overall sky will look much smoother but the star trails will lose contrast. Why is that? What is the average of 100 + 1 + 1 + 1 + 1? Answer: twenty one. What did I just do… I took 5 shots with nearly black pixels (1′s) and one shot with a bright star (100) and effectively I reduced the brightness of the star from 100 to 21. Here are examples. First is stack which selects the brighest pixel from each of the 11 images. Next the same 11 shots are averaged. The red light on the foreground is from one of the frames. Averaging the star trails clearly reduce the contrast of the stars (and the Saguaro) very significantly.
11 Images stacked using "Brighten" (lighten) mode. One of the images has the Saguaro lit by a brake light of a nearby car.
11 images averaged. Notice the big difference in the contrast of the startrails and the muted light on the Saguaro. On the other hand the sky is "smooth" since averaging also averages out random noise.
If it is not possible to reduce noise by averaging, then what is LENR (Long Exposure Noise Reduction)? Many DSLRs can be configured to enable or disable LENR. Usually I leave it off because using LENR makes your shot take from 50% to 100% longer. Why? Well the camera is effectively doing what the photographer can do by putting a lens cap on. It is taking a dark frame. It closes the shutter and lets the pixel counts accumulate. It then uses those accumulated pixels to subtract them from the image it just captured. If you have amp glow (multi-pixel hot spots that appear pink or purple) the subtraction will remove the glow. If you have hot pixels – areas that always read out as bright, the subtraction will eliminate those as well since they will also be in the dark frame. Since the performance of the sensor changes rather dramatically with increased or decreased heat taking the dark frame immediately after the exposure works best. Of course the camera imaging chips may also do things like sharpen or blur pixels that it thinks are out of range with surrounding pixels. Blurring may improve the appearance overall but it also has an effect very similar to the average stack we looked at earlier. It is much better to do what we can to keep the noise low in the first place.
What about simply increasing the exposure with the exposure (or lightness) control? There is no magic in that slider. If you do nothing to “increase the exposure” the pixel values are left alone. If, however you increase the exposure by 1 stop, the effect is to double every value. A 10 becomes a 20, a 50 becomes a 100. This, of course, makes those pixels brighter – but it also makes every other pixel brighter, too, including the mild noise. A slightly noisy unnoticeable “3″ value becomes a quite noticeable “24″ when you increase the exposure by 3 f-stops – an eight-fold increase.
**NOTE: In the earlier Drizzle discussion the word droplet and photon appear to be used interchangeably. But photons are very tiny things and in reality Drizzle’s one droplet per second is really about 2,000 photons per second in case more scientific numbers are desired.