Autofocus problems in low light situation

[Sensographer]

Well, just coming back from a one day studio training session I would like to ask if it's a known phenomenon with some Canon cameras that autofocus has serious problems when working in low light situations (which a studio situation is).

I used the EOS 1000D (because my 5D Mark II still has not been delivered) with different lenses (but mostly 70-200 / f 2.8 @ f8 or f11) and was very frustrated because I had ways more problem with autofocus than the Nikon guys or everyone else. Several times the motor of the lens just was moving (or even not moving anymore) and when I pressed the button the camera simply refused to click. Then I had to focus somewhere else on a more light spot and it worked.

The Nikon guy had some kind of light coming from his camera which helped him for a stress free autofocus. I was wondering if such thing can be bought for a Canon camera also or if it's a built in thing.

Also I am wondering if that might be more a problem with the entry level Canon Cameras. Is that much better with a 5D Mark II ?

Thanks in advance

[apersson850]

You can pop up the built in flash and get a strob of flashes from that one, to support autofocus.
You can put on an external flash unit, like the Speedlite 580 EX II, and get AF assist light from that one. You can use that even if you don't use the flash itself.

Finally, the 1000D has the simples AF system in Canon's current setup of DSLR cameras. More expensive cameras will give better result, but all will have problems if it's very dark.

[Curtis N]

The Nikon guy had some kind of light coming from his camera which helped him for a stress free autofocus. I was wondering if such thing can be bought for a Canon camera also or if it's a built in thing.The least expensive version of an AF assist light for Canon is the ST-E2 transmitter, but it's over $200.

There are many factors involved. More expensive cameras have better AF systems and the 5D Mk II has a pretty good one. It also helps to use fast lenses (which you are).

AF systems need contrast. Aim at some part of your subject where the luminance changes (lapel of a dark coat over a white shirt, for example). This focus-recompose technique is not always acurate because of the geometry involved, but at f/8 you should have plenty of DOF so it won't be noticeable.

Sometimes, the best solution in a studio is to use bigger modeling lamps in your strobes.

[Sensographer]

The least expensive version of an AF assist light for Canon is the ST-E2 transmitter, but it's over $200.

As indeed I already own a ST-E2 I just checked the usage as an AF assist light and it worked great. Only open question now is how to plug in the Pocket Wizard to remote control the flash units ...

[Curtis N]

One option is to use a short cord to plug the PW into the PC socket on your camera. You'll need a way to keep the PW attached, gaffer's tape or velcro should work.

There is a way to tape the auxiliary contacts on the ST-E2 to use it as an infrared flash trigger, the strobes will fire via their slave sensors. I don't know if you can do this and get AF assist too.

[RDKirk]

when working in low light situations (which a studio situation is).

Modeling lights. I have never had a problem focusing with any Canon DSLR in the studio. If you think that's hard, try a view camera with an f/8 lens in the studio.

[Sensographer]

Modeling lights. I have never had a problem focusing with any Canon DSLR in the studio. If you think that's hard, try a view camera with an f/8 lens in the studio.

Hi Kirk
Yes, we had modelling lights, but the setup were more or less two strip lights coming from left and right, which made it more difficult to autofocus.
And by the way I was the only one (nobody else had a Canon) having these problems. And no, it's not because I wasn't able to master the camera ;)

[bobbyz]

I think the Nikon's have a AF assist lamp on the camera body.

[theveed]

Just to give you a tip, if you're using a relatively small aperture anyway, have the subject hold a cellphone with the lit LCD facing you while the phone positioned next to her eye, lock your focus on the phone, then turn off AF.

If the model does move back and forth by a couple of inches, just adjust your MF by a few mm turns.

Other than that, just set up a weak light that won't register in the exposure but bright enough for AF to lock. I usually just mount a small, weak flashlight (those cheapo bulbs not LED bulb) on a gorillapod or clamp near where the model stands if the studio's too dark.

[blssdwlf]

The Nikon guy probably has a built-in AF-assist lamp or was using his flash AF assist light. If you were not using an external flash for AF assist, you'd have to rely on your pop-up flash for your AF assist (and blind the subject.)

The 1000D has AF sensor points only good to f/5.6. It wouldn't have been able to take advantage of the f/2.8 lens.

The Nikon guy had some kind of light coming from his camera which helped him for a stress free autofocus. I was wondering if such thing can be bought for a Canon camera also or if it's a built in thing.

Also I am wondering if that might be more a problem with the entry level Canon Cameras. Is that much better with a 5D Mark II ?

Thanks in advance

[bobbyz]

AF happens with lens wide open so even when you shooting at f8 with f2.8 glass, camera is keeping lens at f2.8 during the time it is trying to AF. Only when you press shutter, it stops down to your shooting aperture (f8 in this case).

[blssdwlf]

AF happens with lens wide open so even when you shooting at f8 with f2.8 glass, camera is keeping lens at f2.8 during the time it is trying to AF. Only when you press shutter, it stops down to your shooting aperture (f8 in this case).

Right - although in this case he is best AF point (center, cross-type) is limited to f/5.6. His glass is opening up to f/2.8, his AF point isn't sensitive enough to take advantage of it which would explain his difficulty focusing.

[Curtis N]

Right - although in this case he is best AF point (center, cross-type) is limited to f/5.6. His glass is opening up to f/2.8, his AF point isn't sensitive enough to take advantage of it which would explain his difficulty focusing.Not exactly.

Some Canon cameras have more advanced cross-type AF sensors that are activated only when an f/2.8 or faster lens is attached.

But any AF system will work better in low light with fast glass. An f/2.8 lens will work better than an f/5.6 lens. An f/1.4 lens will work better yet.

[apersson850]

Some Canon cameras have more advanced cross-type AF sensors that are activated only when an f/2.8 or faster lens is attached. That's true for the 40D/50D only, so it's not very many models that has that feature.

But any AF system will work better in low light with fast glass. An f/2.8 lens will work better than an f/5.6 lens. An f/1.4 lens will work better yet.If an f/2.8 lens works better, then it's just because that's often the limit for high-precision autofocus (which is cross-type on the 40D/50D, but linear on all other models). But an f/1.4 lens is not a bit better than an f/2.8 lens is for autofocus. Neither is an f/3.5 better than an f/5.6 lens, since both are on the normal precision side of the fence.
Center point on 1D models have the limits f/4 and f/8 instead of f/2.8 and f/5.6, so the applicable values shift one stop there.

[blssdwlf]

You guys are both right. I read it wrong a while back and carried the idea forward - D'oh!

I re-read where I got it from http://www.bobatkins.com/photography/eosfaq/eos3af.html

The difference between the f/5.6 sensor and f/2.8 sensor is precision. With the OP's f/2.8 70-200, his best f/5.6 AF point would get the focus to a point within the lens' max aperture DOF. If he had an f/2.8 sensitive sensor, he would've focused to within 1/3rd of that. But there was no mention of it affecting focus lock.

In that case, since the OP couldn't get a focus lock, its probably due to lack of an AF-assist light since he didn't mention out of focus shots...

[apersson850]

I've not investigated the 1000D too closely, as I consider it a bit too small for me. Does it not allow high-precision autofocus with any focus point? On my (now) old 400D, at least the center point does have linear high-precision focus, when using lenses with f/2.8 or better.

[Curtis N]

But an f/1.4 lens is not a bit better than an f/2.8 lens is for autofocus. Neither is an f/3.5 better than an f/5.6 lens, since both are on the normal precision side of the fence.Absolutely false.

Lenses are always wide open during autofocus. Fast lenses let in more light. Lack of light is what causes the AF problems to begin with. The more light the AF system has, the better it works. It's that simple.

[apersson850]

It's not at all that simple. You obviously don't know how the AF system works.

Inside the optics that make up the AF system, the AF effectively has its own apertures. That aperture is always smaller than the one that's required from the lens for a certain type of AF (normal or high precision) to work. Hence as soon as the lens is bright enough for high precision AF, making the lens brighter makes no difference at all. That light is of no use to the AF system.

The only difference is the threshold between normal and high precision, where the lens speed does make a difference.

These apertures don't look or work identical to the apertures in a normal lens, but they restrict light in the same way.

[vadim_c]

The least expensive version of an AF assist light for Canon is the ST-E2 transmitter, but it's over $200.....
I already mentioned that least, always, never etc.. are the words that should be avoided as you can get caught easily :-)
http://www.bhphotovideo.com/c/product/542934-REG/Vivitar_DF340ZC_DF340Z_Digital_TTL_Shoe.html

[vadim_c]

It's not at all that simple. You obviously don't know how the AF system works.

Inside the optics that make up the AF system, the AF effectively has its own apertures. That aperture is always smaller than the one that's required from the lens for a certain type of AF (normal or high precision) to work. Hence as soon as the lens is bright enough for high precision AF, making the lens brighter makes no difference at all. That light is of no use to the AF system.

The only difference is the threshold between normal and high precision, where the lens speed does make a difference.

These apertures don't look or work identical to the apertures in a normal lens, but they restrict light in the same way.
It is not the apertures on the focus sensors that restrict using of faster lenses but the distance between the intakes of the focusing rangefinder.

It is set at a fixed distnce and there are two of them: one that fits the 2.8 aperture and the one that fits 5.6 aperture. It would be good if that ditance were variable depending on the aperture of the lens but apparently it is very costly to implement.

[kenwood33]

No its not canon, its your camera. I used the same camera before, and it has a lot of trouble focusing even for studio shoot where the subject is lit by the modeling light. I switched over to my 40d and problem solved. I did more test with the 1000d/xs outdoor, in a shaded area, it keeps hunting for focus. I would say it's focusing speed and accuracy is even worse than my old rebel XT. When I was testing outdoor, I had a 430ex flash on, which emits a red light to assist focusing, but it is not helping much. Again when I switch to a 40d under the same situation, no focusing issues.

[Curtis N]

It's not at all that simple. You obviously don't know how the AF system works.Your insults are not well-taken and only serve to undermine your own credibility (beyond the fact that your argument makes no sense).

Regardless of how an AF system works, or what components it has, the light has to get through the lens aperture before it gets to the AF system. More light will get to the AF system if it goes through a bigger hole to get there. More light equals better AF system performance. That is not the only factor involved but it is always is a factor. If you don't understand how a lens works, then the intricacies of an AF system are probably beyond your grasp.

[apersson850]

Consider what you write yourself before complaining about insults. There's a big difference between stating that somebody is an idiot ("...beyond your grasp.") compared to that someone has not yet learned about something ("...don't know how...").

The reason I wrote "the AF effectively has its own apertures" was because I didn't want to get too detailed in that post. vadim_c gave you more details above.

To explain why the large aperture of some lenses doesn't help the AF system, we can use yourself in an example.

Imagine that you are standing in a ship's cabin. The cabin isn't equipped with any porthole, so it's pitch black inside.
Also imagine that we've moved your eyes together a bit. Your sense of distance is based on your stereoscopic sight, where the two slightly different images from your left and right eye are combined by your brain into a perspective view of your surroundings. Since you probably don't like the idea of having your face re-arranged, you can imagine wearing some funny glasses, where the optics is such that you effectively look out through lenses that are closer together than your own eyes are. Apart from that, they don't change your eyesight at all.

Now if you drill a small hole in the wall of that ship's cabin, you can suddenly see outside. But if the hole is too small for both of your eyes to look out at the same time, you'll have no sense of distance to the objects you can see. You can see them perfectly well, as the hole is larger than your single eye, so there's no limit to how much light can enter your eye. But you can't perform any ranging, since you can't use both eyes. This is equivalent to using a lens with f/8 on a camera that needs f/5.6 for autofocus to work.

Now find a bigger drill and enlarge the hole in the wall, so that it's big enough to cover both your eyes (or at least both lenses of your imaginary glasses). Suddenly you can tell the difference between near and far, as you can use both eyes. That's all that is important. You can make the hole bigger, but that will not help you in determining the distance to the objects you see. It will make the inside of the cabin brighter, but that's all that will happen. Still, you may feel that you have more difficulties than normal to judge if one thing is nearer than another, or vice versa. That's because your eyes are closer together than you are used to. Or at least your glasses make it seem like that.

Let's for the moment say that we keep the hole large enough for you to be able to tell distance, but not any larger than that.
Now you use the feature of your glasses that let you fold them in the other direction, so that the optics you look out through instead end up spaced further apart than what your own eyes are. According to the instructions, this should give you a better judgement of distance. But when you try looking out through the hole in the wall, you now again find that you can't tell distance at all. That's of course because one of your "eyes" is now always behind the wall again, since the distance between the optics in your glasses is longer than the hole is wide. You've lost stereoscopic eyesight when using the high-precision mode, because the hole is too small.

Now enlarge the hole again, so that it becomes large enough to allow you to once again look outside with both your eyes. The world outside still looks the same to your eyes, since they see it the same way through the optics in your funny glasses. The cabin where you are standing is brighter, though, as the hole in the wall will let more light com in.
The main difference to you is that you suddenly can tell whether object A is closer than object B with much higher precision, since your eyes are now further apart. The baseline you use for your measurement is longer, so your accuracy improves. But it does take that the hole in the wall is big enough, or your high precision mode is not only inferior to the standard mode, but it fails completely.

Now you can make the hole in the wall even bigger. It will make the inside of the cabin brighter and brighter, as you enlarge the hole. You can actually take the whole wall down, to make it as bright inside the remains of the cabin as it is outside. But as long as you can't move your eyes even further apart, it will not make your ability to judge the distance to objects you can see any better at all.

The same happens with the AF system in a camera. As soon as the opening in the lens is wide enough to allow the AF system to use a certain type of sensors (normal or high precision), it will do that. Making the aperture larger will not make that particular sensor type any better. The distance between the lenses in the funny glasses is equivalent to what vadim_c describes as the distance between the intakes of the focusing rangefinder.

[RDKirk]

Actually, the absolute light level does have an impact on focusing--specifically on the speed of calculation. With greater light, the system calculates the proper lens movement more quickly. Not any more accurately, but more quickly. Of course, if the light is too low, it does not complete the calculation at all.

[RDKirk]

Hi Kirk
Yes, we had modelling lights, but the setup were more or less two strip lights coming from left and right, which made it more difficult to autofocus.
And by the way I was the only one (nobody else had a Canon) having these problems. And no, it's not because I wasn't able to master the camera ;)

Frankly, I've never had a problem autofocusing in any condition that I could see the subject at all, except in situations that there was very little subject contrast. But if I can see a line, I can always autofocus on it.

[apersson850]

Actually, the absolute light level does have an impact on focusing--specifically on the speed of calculation. With greater light, the system calculates the proper lens movement more quickly. Not any more accurately, but more quickly. Of course, if the light is too low, it does not complete the calculation at all.
That's true. It's of course then easy to think that a faster lens always gives faster AF, but as soon as the lens is large enough to allow a certain type of AF sensor to work, then the AF optics can't take advantage of an even faster lens. Only a brighter subject will help.

[René Damkot]

IMO you are both talking about a different subject it seems.

Curtis is saying is that any AF system will work better if there's more light to work with. There is more light to work with if you provide more light, be it by using a faster lens, or by turning up the modelling light.

Anders is taking about the technicalities of the AF system (interesting by the way).

So let's keep it friendly :)

[Curtis N]

You can make the hole bigger, but that will not help you in determining the distance to the objects you see.An interesting analogy. Fundamentally unsound, but interesting. It helps explain why certain AF sytems need a certain aperture to work. That is not in dispute.

But a lens aperture does not limit light in the same way as a ship's porthole. If this is difficult to grasp, set your aperture at f/16, look through your viewfinder and press the DOF preview button.

When you do this you will not see a smaller hole. You will see the scene get darker, as if the ambient light level has been reduced. The AF system sees the same thing.

All AF systems need a certain amount of light to work. The ambient light level and the lens aperture both contribute to the amount of light that reaches the AF system.

The bulk of my photographic income is made shooting in theatres with low light levels. I have a bit of experience in that environment (maybe 100,000 shots from a dozen venues over the last 5 years). I have used a variety of cameras from consumer to professional grade, and a variety of lenses from f/1.4 to f/5.6. When the lighting designer decides to make things moody and drop the level to near darkness, I know which lens I'll need.

[apersson850]

It's not at all unsound, it's exactly how it works.

You are also right in that the AF system sees the light coming in from the lens, but it doesn't see it like the sensor does. The sensor will se the image being brighter or darker. The AF sensors, behind the special AF secondary image forming optics, will see light rays coming through the lens along a certain path. That particular path doesn't become brighter just because the total opening in the lens is wider.

If you just don't think I know how this works, then take a look at Canon's specification for the working range of EV-values for autofocus. If your theory was correct, they would list different values for a f/2.8 lens and a f/1.4 lens, right? Since we agree they both use the same sensor (high precision), the difference in light required for them to work ought to be two stops, right?
But it isn't. The specification for light levels is the same, regardless of which of these two lenses you use.

Actually, if you have a lens which allows you to manually change the aperture, even when the lens is off the camera, you can see something similar if you look through it with your eye. If you have any of Canon's older FD-lenses you can trick that into stopping down the aperture even with the lens off the camera. Look through the lens wide open. You'll notice that unlike when you let the lens project an image on a piece of paper, you'll not notice that the subject you see through the lens gets darker when you stop down. That's because you see it with the optics inside your eye, not just projected on a plane consisting of a paper or something.

If you aim the lens at an evenly lit white wall, you'll see a larger or smaller hole, through which you can see the equally bright wall. This is the same way the AF optics look through the lens. If the aperture is obscuring the subject, it's black. If the aperture is open wide enough so that the subject is visible at the point where the AF optics is looking, it's white.

No grey in between.