Canon develops amazing low-light sensor

Canon has just announced a new 35 mm full-frame sensor that is incredibly sensitive to low-light levels. The sensor does this in part by using large pixels (19x19 microns^2), which is about 7.5 times larger than the pixel size of their other cameras. Right now the sensor is optimized for full-HD video. Here is the video Canon has released of some footage from a sensor prototype. Most impressive are the shots of the Milky Way, illuminating a person using an incense stick, and being able to shoot in moonlight as if it were daylight.

So just how sensitive is this camera? According to the specs, the camera can form a useable image at ~0.03 lux. Doing a quick back-of-the-envelope calculation, 0.03 lux corresponds to a sensitivity of about 40,000 photons hitting each pixel each second (for green light). If we shoot at 24 frames/second, this is about 2000 photons/frame/pixel–or about 0.5 femtowatts.

Any sensor will have noise associated with it. In order to get a useable image, you should have a decent signal to noise ratio. Let's say that 5% of the pixels have noise on them (corresponding to a fairly noisy image), that means each pixel experiences about 100 noise counts/second.

What is remarkable is that this takes place at room temperature. There are other sensors that have been developed for scientific applications that are much more sensitive than this Canon camera, but they must be cooled first. For example, a camera I have worked with in the past is the Andor iDus. This camera is sensitive down to the single-photon level when cooled to -80 C. This is achieved in part by using some clever electronics to reduce the readout noise and using larger pixels (26x26 microns^2). At -80 C, the camera experiences a negligible amount of dark counts/s (much less than 1). Increase this to 20 C (room temperature), and that number goes up to a couple of hundred noise counts/pixel (as best as I can tell from the specs).

This analysis should be taken with a grain of salt. The take home message is that at room temperature, the new Canon CMOS sensor performs on par with the best EMCCD cameras out there. This is seen in Canon's own tests (where they measure against a three-EMCCD). It would be interesting to see how well this new sensor performs when cooled.

I can't wait for this technology to eventually make its way into consumer-level technology.

Screen shots from the Canon Video

Eleven months at the bottom of the Earth

Last week I had the chance to sit down with cosmologist Keith Vanderlinde, a CIFAR Junior Fellow at McGill, who spent eleven months straight living at the South Pole in Antartica. During the winter temperatures dip below -70 C and their is continual darkness for nearly six months straight. It gets so cold during the winter that planes cannot fly in–once the last plane takes off you are stranded there until the following sumer.

While in Antartica, Keith was in charge of keeping the South Pole Telescope running. Every day he had to walk 1 km to and from the telescope, often in white blizzard conditions. Keith took his camera with him and captured a series of incredible photos of the night sky and life in Antartica.

While at the South Pole, Keith maintained a fascinating blog about what life is like. My favourite entry is about the 300 club:

There's a tradition here at pole dating back decades, that whenever the temperature outside falls below -100F, the 300 club convenes & initiates new members. You gain entry into the club by first sitting in the sauna with the temperature turned up to 200F, then running outside (a 300F temperature differential, hence the name) and around the pole, all wearing nothing but boots and a smile.

Only once - in the half century for which we have records - has the temperature failed to hit -100F over the course of a winter. It's expected that the 300 club convenes at least once each winter, more likely twice or three times. Well, with the sun now up and temperatures already rising into summer, our low for the year is sitting at -99.9F, and there's no way that would count. Seriously.

,

Kids skating in the snow

Children skating at Uptown Waterloo during the first real snowfall this winter.

Shot with a Panasonic GH1 and Olympus 45mm f1.8 lens.

Merry Christmas

Merry Christmas!

Shot with a Panasonic GH1 with 20 mm lens @ f1.8

Markus Reugels's high speed liquid art

Markus Reugels is a talented photographer who captures incredible photos of liquid splashing. These kinds of high speed captures are typically accomplished using a flash that illuminates the scene for only a fraction of second, freezing the subject. Markus uses multiple water drops and precise timing to sculpt and capture shapes that exist for less than a ten-thousandth of a second.

Here are some of my favourite photos of his (from Flickr):

Square 2 Square Dance Time-lapse

This past weekend I shot my first time-lapse video of a street festival, called Square 2 Square, that has been running this summer in Waterloo. Every few Sundays the main street through town is designated a car free zone. Local business set up booths and fun activities take place. During the last Square 2 Square festival I took some footage of a street water fight near my house. This time around a good friend of mine, David Trinh of Hep Cat Hoppers fame, set up a Lindy Hop booth. I set my camera on top of his car and set my camera to take a picture every 3 seconds. Nearly three hours and 3000 pictures later this is what happened.

Another reason I love Waterloo: during the three hours I was out on the street I ran into seven other physicists.

Amazing Australian Astro and Time Lapse Photography

Alex Cherney is an Australian photographer who has taken some incredible images of the night sky. His photos have won numerous awards and are routinely featured on NASA's astronomy photo of the day site. Below are two of my favourite videos from Alex. They remind me of the superb time lapse video taken at the Very Large Telescope Array. [vimeo]http://vimeo.com/15583376[/vimeo] I love the reflections in the lake and the music from the didgeridoo in the video. The next video was shot of one-and-a-half years and won first place in the STARMUS astrophotography competition. For more incredible footage check out his Vimeo page. 

I just got my intervalmometer (a remote control that can be programmed to take photos at regular intervals) and am excited to try out some time-lapse photography. When it comes to photography, I am always amazed and inspired by people like Alex.

You can purchase some of his prints on Redbubble. I am buying a copy of this astounding panorama to hang in my living room.

Light Craft Workshop Fader ND Mark II Review [Updated]

I recently bought a Light Craft Workshop Fader ND Mark II variable ND filter to use for shooting videos, and was curious about how well it performs. I used the Fader ND to film the water fight that broke out near my house last week, and was pleased with the results (Vimeo version embedded after the jump). But to be more thorough I decided to conduct a series of tests of the Fader ND. In short, the optical quality of the filter is very good with little degradation of the image. The filter is also neutral throughout its useable range of attenuation introducing almost no colour shift. I strongly recommend it to those looking for a budget variable ND filter. Read on for my full review of the Fader ND Mark II.

[Update] I have added some resolution tests of the Fader ND at longer focal lengths at the end of the post.

Introduction

Shooting video in daylight is challenging; it is often so bright out that a high shutter speed on the camera is required. This can lead to jerky and unnatural motion in moving subject, and force the use of high apertures. A solution is to use a neutral density (ND) filter in front of the lens. An ND filter blocks out some of the incoming light allowing a slower shutter speed to be used. Good ND filters attenuate every colour equally (hence the term neutral) while bad ND filters can introduce an unwanted colour cast.

The two main solutions are to use stack of fixed ND filters to control the light level or to use a single variable ND filter whose strength can be adjusted. For my purposes, I decided to go with a variable ND filter. Most variable ND filters are made up of two polarizers that can be rotated with respect to one another. Light travels through space as a wave, and its electric field oscillates back and forth similarly to how a water waves undulates up and down. While water waves can only move up and down, light can oscillated in all directions (for example, side to side, up and down, diagonally, and even in a helical spiral). The direction that light oscillates is known as its polarization.

A single polarizers will cut out 50% (1 stop) of the light from an unpolarized light source like a light bulb. After passing through the polarizer, the light has a definite polarization. If a second polarizer, aligned with the first, is inserted next then all of the polarized light will pass through. If instead the second polarizer is rotated by 90 degrees all of the polarized light is blocked. By rotating the second polarizer in between 0 degrees and 90 degrees the amount of light passing through can be controlled.

The problem with using polarizers to create a variable ND filter is that the polarizers are not always of a high quality. There can be a serious loss of resolution if the optical quality of the polarizer is not good. Poorly manufactured polarizers, with extra materials between them, can also lead to a variable colour cast being introduced into the image–as the polarizers are rotated the colour cast changes.

There are several options on the market for variable ND filters. The deluxe higher end models are the Singh-Ray variable ND filters that can cost several hundred dollars. There is also a newer product out by a company called Light Craft Workshop that makes significantly cheaper (under $100) variable ND filters. I also came across these filters on eBay sold by Rainbowimaging that are incredibly cheap (around $30). I have ordered a number of things from Rainbowimaging before, and have been pleased with the quality and service received. Because they are so cheap, I nearly bought one of these filters just to try out. I had a hard time finding any reviews on them, but did stumble upon some Youtube videos that made it look like these filters introduced a significant colour cast. It was hard to tell if this is the case, but I decided not to risk it. Perhaps one day I'll order one and do a test.

In the end, I went for the Light Craft Workshop Fader ND Mark II (second generation version of the filter). Philip Bloom, a well known blogger and film maker, has good things to say about the Fader ND, so I ordered a 52mm filter from the Canadian distributor on eBay. Again service was excellent and shipping fast. While the filter I ordered will thread onto a 52mm front element, the polarizers are 55mm. This is to help limit vignetting, but means that any other filters/lens hoods that go on in front must have a larger thread size. Every size of the Fader ND uses polarizers that are one step larger.

The Fader ND comes with a convenient carrying case as well as a lens cover (55mm in my case) that works well. The polarizers rotate smoothly and are well constructed. There are marks on the filter to help serve as a rough guide, but in practice I have not found them useful. They do not correspond to the number of stops of light blocked. This is not a fault of the Fader ND, but is a consequence of the way polarizers work. As the angle of the second polarizer changes, the amount of attenuation will not vary linearly. If the Fader ND were to introduce marks that correspond exactly to the number of stops of light blocked, the marks would be distributed in a more complicated fashion. It would be nice to have accurate markings, but for the price I can not complain.

To test the resolution and check for the presence of a colour cast, I carried out two different tests. In the first, I set up a tripod outdoors and shot a sequence of photos of a picture from the book "A New Kind of Science" by Stephen Wolfram. I do not have an Airforce test chart, but this book has very high resolution photos that contain fine patterns suitable for the test I am interested in. I am not able to calculate the resolving power of the lens with and without the Fader ND, but any major loss in resolution should be readily apparent.

Resolution Test

The resolution test was conducted on my Panasonic GH1 using the Panasonic 20 mm f1.7 pancake lens. In all the photos the aperture is fixed to f 5.0. First a picture was taken without the Fader ND to serve as a reference. The Fader ND was then screwed into place and a sequence of pictures taken with different attenuation settings. Each picture was taken with the Fader ND set to attenuate one extra stop with respect to the previous photo. The exposure was set by first halving the shutter speed and then adjusting the Fader ND until the correct exposure was reached (as I had to rely on the metering in the camera, the exposures are close, but not perfect).

When the Fader ND is set to 0 degrees there is a loss of 2 stops of light (maximum transmission). This matches Light Craft Workshop's claim of 2 stops of attenuation. Each subsequent picture represents an extra stop of attenuation. The central portion of the image is shown at 100% crop to provide an idea of the resolution with and without the Fader ND.

I am surprised at how well the Fader ND performs; there is barely any discernible degradation in the resolution of the image. The Fader ND also performs well over a large attenuation range (meaning any degradation in resolution is due solely to the quality of the polarizers and not the rotation of them). I get over 11 useable stops of attenuation which is more than Light Craft Workshop specs the filter for. Setting the polarizer close to 90 degrees (most of the light blocked) leads to appearance of a "cross" pattern that is typical for these types of filters.

Colour Cast Test

To test for the presence of a colour cast, I shot a series of pictures of white sheet of paper. Without the Fader ND present I manually set the white balance of the camera. This is the white balance setting used in all shots. Next the Fader ND was screwed on and pictures of the paper taken for different settings of the Fader ND. Again, the exposure was compensated for by adjusting the shutter speed and ISO. These photos were taken in the passengers seat of a moving car, so they are blurry at lower shutter speeds. I also had some problems keeping the lighting/exposure as consistent as I would have liked, but they are still instructive. The camera settings along with the RGB values for the white in the image are shown (the images were underexposed, but that does not change the results of the test).

Again, the performance is excellent. From 2 to 10 stops there is a slight but consistent colour cast as blue suffers a higher attenuation. After 10 stops a small bluish tint appears. In both cases the colour cast is minor and can be easily corrected by white balancing the camera with the Fader ND on. What is most impressive is how neutral the Fader ND is as the attenuation is changed. Cheaper crossed polarizers are notorious for introducing dramatic bluish, greenish, and reddish colour shifts as the polarizers are rotated. This is not the case with the Fader ND Mark II.

Resolution Test at Longer Focal Lengths

There has been reports that the Fader Mark II performs poorly at longer focal lengths. To see if this is the case, I repeated the resolution test using my Panasonic 45-200mm zoom lens. I placed the camera on a tripod and set the aperture to F8 and the shutter speed to 1/320 of a second. First I tested the camera with out the Fader ND (ISO set to 100) taking photos at six different focal lengths. Next I screwed on the Fader ND and repeated the test. To compensate for the two stops in lost light I upped the ISO to 400. The results can be seen below.

There is a clear loss of resolution as longer focal lengths are used. There also seems to be a loss of image contrast as well. By the time 200mm is reached, the image is noticeably blurred. Shorter focal lengths are not nearly as bad. I definitely would not use the Fader ND on a longer reach zoom lens for photography purposes. For video it is probably fine where the high resolution image out of the sensor is down sampled. The significant drop in resolution is likely due to the optical quality of the polarizers used. Any roughness or tilt in the surface of the glass can lead to this decrease in quality. I have heard reports that the much more expensive Singh Ray filters as well as the Fader ND HD do not suffer this loss of resolution at longer focal lengths. Buying two high quality linear polarizers (or a linear polarizer and a reversed circular polarizer) and building your own variable filter may be another way to avoid these problems.

Conclusions

The Fader ND provides excellent value for the money. There is not a noticeable degradation in image quality at shorter focal lengths and the filter is neutral from 2-10 stops. My only regret is not buying a larger filter. Right now my largest lens has a front filter size of 52mm, so the Fader ND works on all of them. In the future if I buy lenses with a larger front filter I will have to purchase a larger Fader ND. If I had bought a larger Fader ND from the start I would have "future proofed" myself. Given the quality and value of the Fader ND Mark II, I will have no qualms buying a second one when the time comes. If, however, you are looking for a variable filter to use with long reach zoom lenses for photography, you are better off looking for another solution.