Low light focusing tips.
Last updated May 2, 2001, version #10
New, test your camera against mine! (click here)
There are many reports of difficulties with focusing the CP990 in low light. I personally don't feel that the 990 is any worse than most of the digital cameras without IR assist out there as far as low light focusing goes, but you can still have plenty of problems focusing in low light, and the rapidly flashing green light is a common sight in low light.
The key point
The key point to realize here is that the camera settings can effect how readily the camera can focus in low light. As a gross example of this, if you stop the camera down to F11, then ask it to focus, it will have trouble; there just isn't; enough light coming in. As we will see, the camera can have wildly different ability to focus under the same lighting conditions depending on how you set it.
How focusing works on the CP990
Understanding how focusing works on the CP990 will help us understand the focusing difficulties. The CP990 uses contrast detection focusing. As such, it reads out the CCD, and examines the central portion of the image for contrast. It moves the lens, attempting to maximize contrast. Because the CP990 has over 4000 possible focus positions for the lens, it has to use a very fast routine to read the CCD out determine focus. It probably has to be able to do 20-100 measurements per second to be able to focus within a second. Obviously, the camera can't make 4000 measurements to find the best focus, it probably uses a binary search of lens positions so that only a few 10s of measurements need to be made.
Because of the need for speed, the camera can't examine the entire image for contrast. Most contrast detection systems used in video and other equipment today achieve speed by reading out just a few (or even one) horizontal lines of the CCD, making them sensitive only to vertical edges. Evidence is that the 990 also uses this technique which you can prove for yourself by drawing a black line on a white paper and focusing on it holding the camera horizonally and vertically and placing the line different distances from the center. The 990 only uses a small active focusing area, and only uses vertical features to focus; probably because of speed considerations.
Factors that influence focusing on the CP990 in low light.
Several factors influence focusing on the CP990. Because the camera is looking for contrast, these factors fall into two categories; those that influence the location or size of of the focus area, or those that influence the exposure (aperture, ISO).
Effect of ISO
The easiest to understand factor is ISO. As ISO is increased, the ability to focus in low light is increased because the image is "brighter". You can see this effect with your own eyes if you watch the LCD while you increase the ISO; it becomes brighter and easier to see in low light as you increase the ISO, and it is also easier for the camera to focus. It is easy to set up an experiment to test this; for example I have the camera set up right now on a tripod so that it will not focus at ISO 100, but will focus at ISO 400. You can see that this could cause the camera to not be able to focus in a situation at ISO 100 in M-Rec, but to be able to focus in A-Rec where AUTO ISO would boost the ISO. This alone could account for many of the reports of difficulty focusing, especially if the shooter is taking the "lock at ISO 100 to avoid noise" common advice. The available ISO range is only a factor of 4, so there is only a small range of exposures in which increasing the ISO will help focusing
The Effects of focusing area and subject
The actual area used by the 990 to focus is very small. The area appears to be the size of the actual focus braces that show up in AF area mode. You need to have a contrasting subject in this area for the camera to be able to obtain focus in low light. If you point the camera at a diffuse object, such as a white wall, it may have trouble focusing in even bright light. This effect is worse in low light. My tendency is to point at the brightest object in low light, which isn't always the best choice. You have to get used to trying to find a contrasting edge to focus on in very low light.
Effect of focusing mode
For some reason, the 990 has much more trouble focusing in low light in "manual AF area mode" vs. "Auto AF area mode". This may be the biggest cause of reports of low light focus problems. Originally, I had left the camera in manual area mode with the center area selected, which is poor for low light focusing. I can't see any difference between Auto AF area mode, and AF area off mode. Interestingly, in Auto AF area mode, in low light, usually only the center area is used, even if the only contrasting target is placed in one of the outer areas. This may be because the center area is larger, so that in low light you either get focus in the center or not at all. Anyway, in low light, only the center area appears to work
The difference between manual AF area mode and auto or no AF area probably accounts for the reported low light focusing differences that have been reported between A-Rec and M-Rec also.
Effect of LCD monitor on/off
There was a report that this had an effect, but I could not reproduce it except under the following conditions, and it proves to be an extension of the above case with AF area mode. If you have manual AF area mode selected (which has trouble focusing in low light), and you turn off the LCD, the camera appears to revert back to Auto AF area mode which is better in low light, and suddenly it can focus. This makes sense I guess since you can't see the monitor to pick the AF area, so it reverts back to auto mode. You can prove that the camera reverts by selecting an outer area in manual mode, the turning the LCD off, getting a focus lock, and taking a picture. Examine the picture in Play mode and you will see that (generally) the center area was used, not the one that you picked.
Effect of metering and other modes
I could not find any effect of the metering mode (I had thought that spot metering might be a problem, but since the camera only seems to use the center area (larger?) in low light, there was little effect. Other settings, such as EV, have no effect. Contrast + and - has no effect either, but I had thought that it might; the processing must occur on the final image, not during the focusing process
How low can you go? Comparing to other cameras.
There has been a lot of talk that the CP990 is "worse" or "better" than other cameras at low light focusing. Personally I think that it compares well with the other cameras that I have used. One way to test "how low" the minimum light needed to focus to compare cameras is to use a very contrasting target (I used a black and white test chart), and reduce the light until the camera can't focus. Set the CP990 for ISO 100 and aperture priority and open the lens up all the way. (this is so that the camera will automatically select exposures over 1 second) Reduce the light until the camera can just focus. Take a shot and look at the shutter and aperture. Use the formula
EV=log(F*F/t)/log(2) Where F is F stop, t is time in seconds
to convert the lighting used to the EV system. For example, the lowest light that my CP990 could focus on the target resulted in an exposure that was took such a shot at 4.93 seconds at F3.7.
This corresponds to EV 1.47
Now use the formula for ISO 100 to convert this lighting level to footcandles...
The corresponds to about 1.4 footcandles; quite low.
Next I measured another camera, using my Sony D770, I adjusted the light until the D770 could just barely focus on the same target. I used the CP990 to take a shot (so that the same camera was used to measure the exposure), and it resulted in 5.65 seconds at F3.7
This is EV 1.28 or 1.21 footcandles. The D770 was only slightly better than the CP990 at low light focus in this test. There was some variation each time I did this, because the light levels are so low, so these results are not strictly accurate. This is also a best case example because the target has very high contrast. In the "field" I have seen the CP990 have trouble below EV 4. For reference, a 1/4 second exposure at F3 would be EV 3.3.
I'd like to hear reports of how other cameras operate, I'll make a table when I get enough entries, e-mail me!
Getting the best low light focusing action.
* Don't use Manual AF area mode at all, even if you only select the center area.
* Pick a sharp contrasting vertical area and put it in the center. The camera will only focus on the center area in low light. Don't pick a diffuse area. Keep in mind that the actual focus spot is very small. Tip! Turn the camera 90 degrees to get focus if you have to, then turn it back to take the shot.
* Boosting the ISO to 400, or leaving it on AUTO instead of using fixed ISO 100 helps a little bit.
Test your camera against mine!
There are so many reports of low light focus problems that I am trying to come up with a low light focus test so that people can compare cameras and see if theirs really has a problem. Here is the first attempt at such a test, I'd love to hear other people's results.
How to test.
(Photodisk has moved the location of the target on their web site, I have updated my link)
You need the PhotoDisk test chart from the PhotoDisk site. Click here
The PhotoDisk chart is a .sit "Stuffit" file. sorry, and if you don't have a decompressor for that already, you can download the free Aladdin Expander software from Aladdin Systems here. (It does .zip also, and is a good free alternative to WinZip)
The 4MB JPG version of the chart is good enough.
Print the chart on glossy photo paper at 10 inches tall, I think that it will be about 6.7 inches wide. Use a good quality photo printer, although we will be using the chart so far away that any quality difference will not matter much.
Set the chart up in dim light at right angles to the camera, with the camera perfectly centered on a tripod.
Set the chart exactly 6 feet from the lens.
Be sure that there are no reflections of room lights off the paper
Set the camera to ISO 100 locked, M-Rec mode, full zoom, Aperture priority at F4.0. Turn off the flash. Auto AF area mode (not manual).
Now the subjective part. With the camera centered on the chart, start reducing the lighting (I tilted some room lamps) until the camera can just barely focus most of the time. Jiggle the camera aim point around a little on center of the image in case the exact focus point matters. You will be able to find a lighting intensity that works most of the time in most spots. You don't need to take pictures while doing this, just half press to try and focus over and over.
Now take a picture. It will probably be a few seconds long.
Examine the picture in Play mode to see how long it was exactly. Do the procedure a few times to convince yourself that the times are consistant. My camera was always about 4.4 seconds. This is the number we are going to compare. I may convert to EV if needed later, I'd like to hear some reports and I'll put them up here.
Comments about why the test is set up this way...
* Fixed ISO is chosen because ISO can affect focusing slightly
* Full zoom is chosen since the 990 does not have a readout of zoom level; everyone can set the same zoom level this way.
* I've suggested moving the exact focus point around on the chart a little in case that matters.
* I've placed the cart far enough away so that slight quality differences in the chart should not matter.
* I picked this chart instead of a line chart because it simulates a real world subject. There are some drawbacks to using a chart at all, but at least this is something that everyone can reproduce.
* I don't know if the color of the light used matters, I tried both fluorescent and incandescent with about the same results.
* Using the camera this way uses the camera as sort of a light meter, to measure it's own lowest light level. There might be calibration differences between cameras that will matter; just have to see.