Wednesday, 28 November 2012

Fisheye sharpness

When I reviewed the Samyang 7.5mm f/3.5 fisheye, it was natural to compare it against the native Lumix G 8mm f/3.5 fisheye lens. Since the Lumix lens costs about 2-3 times the price of the Samyang lens, I would expect that the Lumix lens comes out as the winner.

However, what I found was that the Samyang lens was sharper, even in the extreme corners, and without the in camera CA correction which is done with the Lumix lens when using it on a Panasonic camera.

Some have doubted this result, and I have also been a bit unsure now and then. So I decided to do another test.

This time, I photographed a couple of trees against the bright sky, a useful test for lens sharpness. The focus was set on the centre of the image. With the Lumix lens, I used autofocus, and set a centre spot. With the Samyang lens, I used the 10x focus assist magnification available with the Panasonic GH2 camera. Here are the two images, both at f/3.5:

Samyang 7.5mm f/3.5 Lumix G 8mm f/3.5

To better evaluate the sharpness performance, I have compiled some 100% crops. From the centre of the image:

And from the top right corner:

Based on this study, it is perhaps not correct to say that the Samyang lens is much better. I think the Samyang lens appears to have slightly better sharpness in the corner, but the differences are fairly subtle.

Keep in mind that the Lumix lensis automatically corrected for CA artifacts, though. By looking at the RAW image files, and converting them to JPEG with third party software, we can see how the images were before the software CA corrections. These 100% crops are from the top left corner:

I used the free program UFRaw to convert the RAW images, but any program could have been used here, as long as it allows for doing the conversion without applying CA correction.

Here we see that without the in camera software CA correction, there are some significant red/green fringing artifacts. These are largely removed in the out of camera JPEG image. In the light of this, the Samyang lens's performance is even more impressive, as no in camera CA correction was done.


I wouldn't say that the Samyang 7.5mm f/3.5 fisheye lens is significantly better than the Lumix G 8mm f/3.5 fisheye lens, even if it appears to perform better in this test. So the image quality should not be an important factor when deciding between the two.

You could also consider the fact that they have slightly different projections. The Samyang lens gives images that are slightly less barrel distorted, and may be easier to defish.

The Samyang lens also handles flare better, see a direct comparison between the two in my review. Flare handling is very important for a fisheye lens, since you are quite likely to find the sun or a strong light source inside the image frame, due to the wide field of view.

When focusing manually with the Samyang lens, it is important to keep in mind that you shouldn't blindly trust the focus distance scale. On my lens, it is slightly off, and I reach infinity focus slightly before the infinity mark. This is not uncommon with manual focus lenses. Exact calibration would be very expensive, and the producers usually leave some slack for themselves by allowing the lens to focus beyond infinity.

The autofocus of the Lumix G 8mm f/3.5 fisheye isn't really needed for landscape pictures, as you easily get infinity into focus manually. But when taking closeup images, the autofocus can come rather handy.

Sunday, 25 November 2012

Prime lens overview

Prime lenses are lenses without a zoom function. They have a constant focal length, and a constant angle of view. When the Micro Four Thirds system was introduced, the very first prime lenses were the Olympus 17mm f/2.8 and the Lumix G 20mm f/1.7. While the Olympus prime lens was generally disliked for the limited maximum aperture, the Lumix lens was an instant classic with the brighter aperture and the perceived very good optical qualities.

Since that time, a lot of prime lenses for the M4/3 system have been introduced. In this article, I present an overview of these lenses.

First, here is an illustration which presents the lenses in terms of the focal length (the X axis) and the aperture (the Y axis). In the picture, the lenses are plotted as dots, where the size illustrate their relative price, and the colour the brand:

In the rest of the article, I illustrate what the focal length and aperture means, before going into details about the individual lenses.

Sunday, 11 November 2012

IR filters

When taking pictures, we normally want the resulting image to accurately depict the reality. However, there are times when we rather want to depict a scene in a way that is not consistent with the reality as we see it.

This may sound strange. As the camera industry has spent an enormous effort to realistically depict the scene, why would we want to distort it?

However, keep in mind that one of the current crazes within popular photography is instagram, hipstamatic, and so on, which is all about distorting the images, and making them look unnatural.

Another way in which photos can be distorted is selective focus. The combination of a large sensor and a large aperture can give you an out of focus background when taking a portrait photo, for example. The rendering of out of focus regions is called bokeh, and the effect can be much more pronounced than what you would see with your own eye. Hence, one could argue that this is a distortion of the reality.

Another type of distortion is of course to use a lens with an unrealistic projection, for example a fisheye lens.

In this article, I am looking at another type of distortion: To use an IR filter to keep off the visible light, and hence, to photograph light outside of the visible spectrum.

A number of different types of IR filters are available. Usually, they are categorized by the cut-off frequency. Visible light roughly goes from 340nm to 740nm, with ultra violet (UV) to the left of the visible spectrum, and infrared (IR) to the right, see the illustration below.

The idea of the IR filter is to remove light which is to the left of the cut-off frequency, only leaving the infrared parts of the spectrum.

I tried two filters, both where bought online at a low price:

The filter to the left is labelled with "IR950", meaning that it cuts off all light to the left of 950nm. Hence, it blocks all visible light completely, and even some of the infrared light. The filter to the right is less extreme, labelled "IR680": It retains some visible red light, but cuts off the larger part of the visible spectrum. I start with the one rated the lowest:


This filter looks completely black, however, when holding it up towards the sun, you can see that it lets some red light through:

To see how it affects the images, I have taken the same image without and with the filter. Here they are:

Without filter, GH2 with Lumix 20mm f/1.7, ISO 160, f/5, 1/160s:

With IR680 filter, GH2 with Lumix 20mm f/1.7, ISO 160, f/2, 1/2.5s:

Based on the exposure settings, we see that the IR680 filter steals about eight stops of exposure, which is a lot. Normally, you need to use a tripod and a slow shutter speed when using an IR filter.

So how is this interesting? It is just a very red tinted photography. However, when opening the RAW image, one can adjust the white balance, which gives a result like this, leaving some false colours:

Another possibility is to desaturate the images, to create black/white photos. Here are the two images desaturated:

Without filter With IR680

The desaturated images are not that different.

Here is another example image, taken with the IR680 filter, and no desaturation in post processing:


This filter looks even blacker than the IR680 filter. When holding it towards the sun, it still looks completely black. I tried it on the GF3, GH1 and GH2 cameras, and they could all just see black when using the filter. So my conclusion is that this filter is too extreme to be used with most digital cameras.

Different sensors can have different types of filters in front of the photosites, and hence, react differently to infrared light. That is why I tested three cameras before giving up. It is well known that some cameras are better than others for infrared photography.

I also tested two different lenses, as I have heard that the coating on some lenses can block infrared light.


I also tried a third IR filter, with a cut-off frequency of 850nm, not pictured above. This filter looks completely black to the human eye, but the camera can see a little bit through it.

I tested it with the Lumix G 14mm f/2.5 lens. Here are pictures of the same scene without and with the filter:

Without filter, ISO 160, 1/4000s, f/2.5 With IR850, ISO 1250, 1/3.2s, f/2.5

The IR850 filter appears to remove too much light for the Panasonic GH2 camera sensor, and is not very useful.

Note that there is a lot of flare in the image taken with the filter. This is not due to the infrared light, but rather due to the strong light source in the centre of the image, and the poor quality of the filter glass. I have written about this effect of using filters here.


I think the IR680 filter lets through too much visible light, and hence, it doesn't create that special effects. On the other hand, the IR950 filter is too extreme. It removes all the light that the camera can see, and hence, is useless.

I would guess that something in between would give more interesting results. Perhaps a filter with a cut-off at around 720nm would be better. Such a filter removes virtually all visible light, and allows you to photograph a band of infrared light, given that the camera sensor is sensitive enough in this spectral range.