R5 Auto Focus accuracy under household fluorescent lighting.

[Chris.Chapterten]

Hi All,

Just curious to gauge some user experiences with shooting under fluorescent lighting with the R5 (or any mirrorless camera for that matter).

My experience:

On a recent fashion shoot the modelling light on my flash head blew out and I was in a very dark room. Luckily (or so I thought) directly beside where I was shooting were two lamps with fluorescent bulbs. I turned those on and used them to increase light level to make focusing ‘easier’.

Upon reviewing the images taken with my R5 and RF 50mm f1.2L lens, I found the auto focus accuracy wasn’t as good as I was used to. Normally I achieve 95%-100% accuracy with that lens easily. In this case, with the fluorescent lights on, it was probably 10-20% lower than that, with the majority being back focused. Sometimes 5 in a row backfocused.

The fluorescent lights were the only thing that was different on this shoot and the only time the AF accuracy had been a bit of an issue.

I have read some reports that DSLR phase detect AF systems sometimes don’t get along well with household fluorescent bulbs, but wasn’t sure about mirrorless?

The sensor readout speed of the R5 is 1/60 correct? I would assume the AF calculation would be done just as fast if not slightly faster than that? Could the 60hz flicker of fluorescent bulbs be interacting with the R5 in a negative way when it comes to AF?

Has anyone else here shot under similar conditions? How were your results with the AF performance?

Thanks!

 

[YuengLinger]

Curious if you tried with anti-flicker on and off? Do you mean old straight tubes or CFL? Some of the older ones might be more problematic because of ballast going bad? Could it have been dimmer than you thought?

 

[Chris.Chapterten]

Curious if you tried with anti-flicker on and off? Do you mean old straight tubes or CFL? Some of the older ones might be more problematic because of ballast going bad? Could it have been dimmer than you thought?

It didn't occur to me to use anti-flciker as I wasn't noticing much flicker in the viewfinder while shooting. But that's a good idea, it might have helped.

The bulbs were CFL, just two bare bulbs in a lamp with no lamp shade.Not hugely bright, but they were relatively close to the subject (1-2m). I also had exposure simulation turned off, so the brightness levels may have been a bit lower than I remember.

This is the only time I have ever used CFL as a light source for focusing and the first time I have had focusing issues with that camera and lens combo. Maybe a coincidence?

 

[Del Paso]

I noticed very often AF hunting on less-structured surfaces with my EOS R, anti-flicker on!
Lighting was "room" neons.
No hunting at all, same surfaces, lighting, hour etc... with the 5 DIV...???
PS: in both cases, the lens was the EF 135 f2.
PPS: for me too, it was the first time I had such obvious focusing issues with the R.

 

[Mt Spokane Photography]

I still use CFL bulbs to light my photos in one of the rooms in my studio. I have 8 of them, 4 to a softbox. To make them fit, I use 2 X 85 watt and 2 X 65 watt in each softbox. Thats equivalent to some fairly powerful incandescent bulbs. I see no issue with autofocus. Autofocus is likely to be less accurate in low level lighting, I suspect that's the issue.

 

[Chris.Chapterten]

I noticed very often AF hunting on less-structured surfaces with my EOS R, anti-flicker on!
Lighting was "room" neons.
No hunting at all, same surfaces, lighting, hour etc... with the 5 DIV...???
PS: in both cases, the lens was the EF 135 f2.
PPS: for me too, it was the first time I had such obvious focusing issues with the R.

Very interesting, thanks for your reply. I’m surprised the 5D IV did better. DSLR AF is supposed to be quite sensitive to the IR spectrum so I had assumed they would do worse with this type of lighting given it has very little IR.

I still use CFL bulbs to light my photos in one of the rooms in my studio. I have 8 of them, 4 to a softbox. To make them fit, I use 2 X 85 watt and 2 X 65 watt in each softbox. Thats equivalent to some fairly powerful incandescent bulbs. I see no issue with autofocus. Autofocus is likely to be less accurate in low level lighting, I suspect that's the issue.

That’s good to know you’re having success with CFL bulbs and AF. Are these normal bulbs you can buy at the supermarket or specially designed for photography? 300 watt of CFL per softbox must be quite bright indeed! I also wonder if having multiple bulbs mounted together like that would make the ‘dim’ moments of the flicker cycle, somehow brighter?

It’s hard to remember the exact light levels I was working with as obviously the flash completely overpowered the ambient. Though I am sure I have focused in dimmer light levels in the past with fewer issues (non-cfl light sources). Hard to say 100%. I might do some more testing and see if I can figure out what’s going on.

 

[Joules]

I still use CFL bulbs to light my photos in one of the rooms in my studio. I have 8 of them, 4 to a softbox. To make them fit, I use 2 X 85 watt and 2 X 65 watt in each softbox. Thats equivalent to some fairly powerful incandescent bulbs. I see no issue with autofocus. Autofocus is likely to be less accurate in low level lighting, I suspect that's the issue.

Which body do you have? If you take pictures without anti flicker on, do they show the light strobing?

I could see the AF being affected by the strobing. If the lights indeed flicker at 60 Hz, one period is exactly as long as the R5 takes to read the entire sensor. I don't think we know if that is actually what it does to read out its focus points, but if so, the points will not be uniformly lit which may explain issues. We also don't know how long the exposure time for the focus acquisition is, right?

 

[Chris.Chapterten]

Which body do you have? If you take pictures without anti flicker on, do they show the light strobing?

I could see the AF being affected by the strobing. If the lights indeed flicker at 60 Hz, one period is exactly as long as the R5 takes to read the entire sensor. I don't think we know if that is actually what it does to read out its focus points, but if so, the points will not be uniformly lit which may explain issues. We also don't know how long the exposure time for the focus acquisition is, right?

Trying to do a little more research into this. Apparently CFL bulbs flicker at, at least 1/2 power frequency. So 60hz power source would be 120hz flicker. Modern CFL bulbs often also have an electronic ballast which means the flicker can be much faster (thousands of times per second).

In my case it is more likely that the light level of the bulbs was low enough to negatively impact the AF performance. I would think that the dual pixel AF wouldn’t need the IR spectrum that CFL bulbs lack? It is quite different to a DSLR AF system after all

 

[Joules]

Trying to do a little more research into this. Apparently CFL bulbs flicker at, at least 1/2 power frequency. So 60hz power source would be 120hz flicker. Modern CFL bulbs often also have an electronic ballast which means the flicker can be much faster (thousands of times per second).

In my case it is more likely that the light level of the bulbs was low enough to negatively impact the AF performance. I would think that the dual pixel AF wouldn’t need the IR spectrum that CFL bulbs lack? It is quite different to a DSLR AF system after all

I don't think IR is relevant to the issue, as the IR filter in front of the sensor should virtually eliminate it anyway. That's why the IR AF beam that some flashes have is no longer working with RF bodies as far as I'm aware.

But as long as the readout duration is an even multiple of the bulb's flicker period (1/frequency), it should create uneven illumination across the frame if the exposure time used for acquiring AF is too short for multiple periods to balance the effect.

 

[YuengLinger]

Have you tried some tests at home yet? Just get some CFL's and use a target or friend to see if you can reproduce? If so, maybe you could try different settings to see if any improve performance.

Not trying to blame the user, but, as this seems to be, so far, a one off, besides low light, was there anything else unusual? Maybe your subject had just a bit of hair hanging in a certain way that tricked AF into missing eyes?

Were you using Servo AF or One-Shot? Was the subject moving a lot?

It might be hard to recreate your conditions--even if you knew the wattage of those bulbs. CFL's have a pretty big range of manufacturing standards and QC...

Here's one article about dimming the bulbs that also provides some insights. I don't believe they flicker at 120Hz, but much faster.

How compact fluorescent lamps work–and how to dim them | EE Times

Occam's Razor? Just dim? As for the idea of 60Hz throwing off the AF, I don't know the engineering, but Canon must have understood the need for AF to work under most lighting conditions, especially the most common ones!

 

[JPAZ]

......... I have 8 of them, 4 to a softbox. To make them fit, I use 2 X 85 watt and 2 X 65 watt in each softbox.......

If one uses 8 CFL bulbs vs 2 CFL bulbs, does that make a difference? Does anyone know, do the individual bulbs flicker in synch or do they flicker at random. If the flicker is random, the more bulbs one uses, the less likely any short time interval is "dark" which would affect the focus ability.

 

[Chris.Chapterten]

Have you tried some tests at home yet? Just get some CFL's and use a target or friend to see if you can reproduce? If so, maybe you could try different settings to see if any improve performance.

Not trying to blame the user, but, as this seems to be, so far, a one off, besides low light, was there anything else unusual? Maybe your subject had just a bit of hair hanging in a certain way that tricked AF into missing eyes?

Were you using Servo AF or One-Shot? Was the subject moving a lot?

It might be hard to recreate your conditions--even if you knew the wattage of those bulbs. CFL's have a pretty big range of manufacturing standards and QC...

Here's one article about dimming the bulbs that also provides some insights. I don't believe they flicker at 120Hz, but much faster.

How compact fluorescent lamps work–and how to dim them | EE Times

Occam's Razor? Just dim? As for the idea of 60Hz throwing off the AF, I don't know the engineering, but Canon must have understood the need for AF to work under most lighting conditions, especially the most common ones!

So far I have only taken a handful of test shots at home with a CFL bulb and they were with the RF 35mm f1.8 set to f3.2. So I wouldn’t trust those results as that is right within focus-shift territory for that lens.

I’ll give it another go with my other L lenses and try to replicate the conditions. Though like you said it will be hard as different CFL bulbs are all made differently.

In terms of anything else unusual, I did notice about 3 shots where the focus point wasn’t perfectly over the eye, the edge of the focus box was hitting a small section of hair near the ear. Needless to say those particular shots were backfocused from my intended point of the eye. In other instances however the focus point was centred over the nose or the eyebrow, yet the result was still backfocused slightly. Subject was holding a static pose for the vast majority of photos and I was in One-shot mode, with Spot AF.

I still got enough shots in focus that I didn’t miss any critical shots luckily but I just noted a significant drop in focus accuracy overall. In the first look shooting in natural light (exposure of ISO 100, 1/200, f2.2) with all of the same camera settings and shooting technique, I missed focus in about 5 shots out of 500.

I would also be surprised if Canon hadn’t thought about the effects of flicker on AF, but it may also be something that is impossible to ‘filter out’ with a fast readout sensor and very quick AF acquisition?. The AF certainly wasn’t broken either, just not as good.

It’s interesting that a couple of people have also reported focus inaccuracies with neon lights as the main source of illumination. I also seemed to to struggle a little with AF when shooting with a colour LED panel set to a fully saturated red. (Saturated red light certainly reduces contrast in an image, so that may have played a part).

In the case of these CFL bulbs, lower light level may be a contributor... but it certainly wasn’t the dimmest light I have ever tried to focus in so I thought something else might be at play?

 

[YuengLinger]

Chris, every impression I get from your posts suggest you are a very skilled photographer. So it is with the most humble of attitudes that I recommend, even when subjects are static, if you are hand holding and have a fairly wide aperture, you use Servo AF. I discovered with the EOS R, Servo AF produces incredible handheld results that One-Shot cannot match. I photograph mostly people. Servo AF seems in the newer mirrorless Canon cameras to make micro adjustments while focusing, preventing focus problems caused by either me or my subject swaying slightly backwards and forwards.

With DSLRs I almost always used One-Shot with static subjects. My keeper rate goes down if I do the same with the EOS R or R5. I can tell you that's Servo AF for portraits was the most compelling improvement which led me to go all-in on mirrorless. (A close second would be exposure showing so well through the EVF.)

 

[Chris.Chapterten]

Chris, every impression I get from your posts suggest you are a very skilled photographer. So it is with the most humble of attitudes that I recommend, even when subjects are static, if you are hand holding and have a fairly wide aperture, you use Servo AF. I discovered with the EOS R, Servo AF produces incredible handheld results that One-Shot cannot match. I photograph mostly people. Servo AF seems in the newer mirrorless Canon cameras to make micro adjustments while focusing, preventing focus problems caused by either me or my subject swaying slightly backwards and forwards.

With DSLRs I almost always used One-Shot with static subjects. My keeper rate goes down if I do the same with the EOS R or R5. I can tell you that's Servo AF for portraits was the most compelling improvement which led me to go all-in on mirrorless. (A close second would be exposure showing so well through the EVF.)

Thanks for being so polite with your suggestion. I must say, using One-shot AF is a habit left over from shooting with the 5Dii, 5Diii and then 5Div over many years. With those DSLRs I found Servo AF entirely useless as the limited number of reliable ‘cross-type’ AF points meant I often had to lock focus and then re-compose slightly to get the desired framing. For sure, this is not an issue with mirrorless cameras anymore with their full AF coverage (I have never really had to focus and recompose with the R and R5). I’ve simply continued to use ‘One-Shot’ on my mirrorless cameras as AF accuracy has essentially been perfect with that mode/ technique up to now. I have gone entire fashion shoots of around 1500 images and missed focus only 5 or so times. Having said that, I always shoot with pretty good light levels: multiple 650watt modelling lights for my flashes (all at full power) or 350w LED panels set to full power when working with constant light.

I could certainly see Servo AF helping (with it’s constant adjustments) when conditions are less than ideal lighting wise and if the subject likes to move around a little more mid-shot. I have actually been meaning to try Servo AF with the Eye-AF enabled. My initial tests of Eye-AF were in One-Shot mode and I wasn’t very impressed with the accuracy/consistency. Selecting the AF point myself in One-Shot was always more accurate for me and I just stuck with that ever since.

When I get around to doing more tests under CFL bulb lighting, I will be sure to switch to Servo AF as a point of comparison with One-Shot. Perhaps the constant adjustments of Servo AF would help overcome any interference from ‘flicker’ (if that is even an issue at all?)

Cheers!