p.3

To make an accurate test you need to be aware of a number of various details. This chart allows you to get a lot more than just the camera resolution. But because the resolution is the main subject of this page, let's discuss a little bit more on that subject. Below you see a photo of the Main Chart and the 2x magnified its center circle. The size of the original photo is 2032x1352 pixels.

Then, the number of vertical pixels is 1352. To be more accurate, you really need to measure the number of pixels between external sides of the horizontal lines that determine borders of the frame in the chart with an appropriate software. I've used Photoshop v.7 for this purpose. In this case the number of pixels is 1280. So, you should see the line #13 on the circle (1280 is the closest to 1300). This is actually the case in this example (the second blue line from the center is #15, line #14 - the first inward from this blue line - is visible only in parts). From this example (Celestron C5 scope attached to Canon D1s body, photo made at the lowest camera resolution) we can conclude that this lens is a good match for the sensor.

Circle Test examples. On the left there are three other examples. Photo 1 is the crop from a picture made by Nikon Coolpix 3100, a 2048x1536 snapshot camera (chart vert. size 1374), photo 2 & 3 by Canon 1Ds with 100mm macro lens with chart vert. size of 1316 and 2630, respectively (its low & hi resolution modes). Although circle 1 is slightly larger than 2, the thinnest line visible is the line #11 (expected #14) in the circle 1 and #14 in the circle 2 (expected #13). The larger format produced visible line #25, with expected #26. This is equivalent of 52 line pairs per mm.

Insert 1. Results of this simple test reveal that Coolpix 3100 is marginally acceptable (11/14=79%), and although it's still above 75%, I would consider looking for another one (the error in this particular test is 1/13=7%). Of course, if you are planning to use just a lower resolution mode of this camera, this test should give good lens efficiency, and probably you won't lose anything on picture detail, compared to its full resolution mode. The 4M pix Nikon E4300 lines of up to line #12 were visible (not shown), making it even worse.

Insert 2. What happens when the lens resolution significantly exceeds sensor resolution is presented here. Sometimes you don't need a high resolution photo, or have a low memory problem and must make shots on the lowest resolution mode. In this case you can see more lines than the theoretical maximum. This is because the lines are not continuous and break in several places. Line #15 and even #16 is visible this way on a matrix that should allow only line #13 to be seen.

Insert 3. This is the highest resolution camera & lens combination that I've tested. Actually, most Canon L lenses that I've tested give a very similar image, an indication that higher, but not by a lot, resolution sensors would be needed to use full power of those lenses. I should mention here, that the circle resolution chart results give worse resolution numbers than the line charts. This is because the resolution is tested in all angles at once, as opposed to only 1 angle tested with a single line chart. Circle chart also tests the pixel shape. Square pixels will give worse results than octagonal.

Inserts 4a and 4b: shots made with SONY DSC V3 at f:8 (4a) and f:2.8 (4b) on wide angle setting (7mm) at 3072 x 2304 (7.1Mpix). You can see clearly that the lens is the sharpest at f:2.8, and the optimum for 35mm cameras, f:8 is a poor setting in this case. This is because of the small sensor size (7.18 x 5.32 mm), that is used in most of snap-shot cameras. The Carl Zeiss lens is nearly perfect and approaches theoretical resolution on the wide angle mode, but light diffraction prevents from achieving a higher resolution using this chip size. On the telephoto mode V3 performs not that good (not shown).

Film cameras (24x36mm) have the max sharpness at about f:8. The same is true for digital SLRs that have the same sensor size. Here is the same circle photographed with Canon 1Ds Mark II with Canon EF 180 mm macro lens. In this case you'd need to make photos of the chart from a larger distance and use a multiplication factor, m/x, as shown on the right (m is the height of the sensor in pixels, and x is the height of the chart). L is the corrected circle number.

A real life example how lens quality is important I've made on one of the highest recommended cameras in 2003: Canon PowerShot A70. It is a 3.2M pix camera with several resolution modes.

The A70 chart analysis shown above was made on its two highest resolution modes: 2048x1536 (called L) and 1600x1200 (named M1). The main circle chart is on the left. In the center there are their 3x magnified fragments, and a tiny real photo is shown on the right. The photo from the 1600x1200 pix image has been digitally enlarged by Photoshop to fit the size of the full resolution photo. Since the chart heights were only 1500 and 1172 pixels for those two resolution modes, at 100% chip use one would expect to see line #15 and #12, respectively. This is the case only in the M1 mode, and indead, the line #12 is the thinnest visible. In the highest resolution only the same line #12 is visible, with line #13 partially visible. Assuming 12.5 lines visible for this resolution, the chip (linear) use is 83%, which is equivalent of a 2.2M pix camera. This means that it is still worth to use the highest resolution mode for best performance. The difference between the L and M1 modes however, is quite small, as one can judge from the photos on the right. An owner of A70 should seriously consider making photos on the 1600x1200 mode since in a not perfect lighting conditions those two modes will produce very similar resolutions, but you can make twice as many photos on M1.

At the bottom of the center field of the main chart there are circles of different colors. Their diameter is marked in the same units, the number of lines per picture height. The minimum number of pixels that may resemble circle from a distance is 9 (it's a cross 3x3 pixels wide where the center one is darker than those on the side, and corner pixels are the lightest). This is a very restrictive test. A small insert from the same photo as the Circle Test, example 2, is shown below (top part). The same pixel resolution (#13) but only #11 in line resolution lens (1100 lines per pict. hgt.) is shown below it. The photo is enlarged 4x, the numbers were added afterwards.

As you can see, the small crosses are clearly visible around the line #5 with a sharp lens and not visible with a not so sharp lens. Please note that for this 'not so sharp' lens the sensor uses 11/13=85% of its power, and is in fact acceptable, according to the standard depicted above. This lens is a better one than the Nikon Coolpix 3100 at maximum resolution. The green channel gives the least noise on the sensor used. Surprisingly, green crosses are the hardest to distinguish, however they look most round this way :-)

The Main Chart shown above in its entirety was made with a Canon 16-35mm L f/2.8 lens at 16mm. It shows a small barrel distortion (the horizontal lines are not straight), larger than the 24-70mm L f/2.8 gives at 24mm. On the other hand, the 16-35mm lens gives no barrel distortion at 24mm and a very slight pincushion at 35mm (lines bent to the other direction, not shown and discussed here).

For more detailed info about the image sharpness, please see Norman Koren Photography tutorial.

The second most important feature for the quality of picture is the color noise or pixel color variation. The noise (equivalent of the grain in negatives) can degrade picture quality so much that resolution charts could be quite irrelevant. The next page deals with this subject.

 

____________ ___________