Exposure Simulation is by default enabled on RP, but can have dramatic effect on AF performance.

Recently I noticed my eos-RP sometimes hunting for focus when shooting for darker subjects whereas my much older 7D (through the viewfinder) focuses just fine. Both using the same 70-200 f/2.8L lens and settings.

My brief experiment:
In a brief test at home with a static subject, I traced the problem on the RP back to the exposure simulation setting. When it was turned on (= default setting) it sometimes had trouble focusing, whilst turning it off caused the AF to perform stellar. I performed my test with both cameras set to full manual exposure with the center spot-AF selected and one-shot mode enabled. A dark-gray coffee-machine with plenty of texture and contrast was my subject in a very well lit room.

EOS 7D with settings: [M f/2.8 1/800 ISO 100] -> Focused in an instant.

EOS RP with settings: [M f/2.8 1/800 ISO 100 expo. sim. ON] -> Kept hunting, could not focus

EOS RP with settings: [M f/2.8 1/800 ISO 250 expo. sim. ON] -> Focused in an instant

EOS RP with settings: [M f/2.8 1/800 ISO 100 expo. sim. OFF] -> Focused in an instant

Hypotheses:
It's weird how just increasing the ISO with 1-1.5 stops causes the camera to focus just fine. Because changing the ISO does not cause any more light to enter the camera, it only amplifies the existing signal. Therefore it should have no impact on the performance of phase-detection auto-focus, yet it could influence the contrast-detection auto-focus. Therefore my hypothesis is, that by enabling expo. simululation, contrast-AF can mess up and sees no contrast in the darker subject when the ISO is kept low. Slightly increasing the ISO will help the contrast-AF algorithm to 'see' contrast in the darker subject again. On the other hand, by disabling exposure simulation, the camera focuses mainly with PDAF (just like the 7D), and focuses just fine.

Poor implementation?
Though my hypothesis could explain my findings, I still find it a rather poor implementation of exposure simulation, since the PDAF could have easily taken over and locked focus, when the contrast-AF had trouble doing so. Though it does not happen too frequently, I had major issues the other day, locking focus on a black cat on a sunny day (there were plenty of subtle highlights in the fur), whereas my 7D did just fine. Increasing the ISO allowed me to focus onto the cat, but caused the background to be completely blown out. I guess for most people, at least those that do not shoot in full manual, this problem is not a major issue. Because in any other mode, the exposure will automatically be adjusted to expose for a darker subject which benefits the contrast-AF. (Though the final image will be slightly over-exposed this way).

Solution:
I noticed the other day, that a simple solution is already built into the camera, but it only kicks-in during flash photography. When doing flash photography, I used a relatively high shutterspeed and low ISO as to kill all ambient light. Now (without the flash firing) the exposure simulation on the RP caused the back-LCD to be completely blacked out, as expected. However, when I tried to focus, the exposure-simulation was briefly interrupted and the LCD brightened up allowing the camera to focus properly. After focus was acquired exposure-simulation turned on again, and the screen turned black as before. It would be cool when the camera always uses this trick when contrast-detection AF fails to acquire focus. Basically, permanently having a flash trigger mounted on top of your RP (but no flash connected), fixes the focus problem. This way you even have a functional exposure simulation still. Best of both worlds? There should be a better solution. For example, since the camera also meters the scene, it could just (internally) increase the brightness of the frame for contrast-detection AF to do it's job. Meanwhile the user sees only the frame with the exposure properly simulated.

In prior cameras with Live View AF (for example, my 7DII) with Exposure Simulation enabled, the lens aperture would stay WIDE OPEN while you partially pressed the shutter to AF, and until you fully depressed the shutter button and then that is when the diaphram would close to the shooting aperture and the exposure is made.

Set your mirrorless RP so the ISO is relatively slow (e.g. ISO 200), and the aperture chosen is f/16 or f/22. And while looking into the front of the lens, press the shutter button partially so that AF begins, and note if the diaphram of the lens is a small opening or a large opening. Note which occurs...

1. If the opening of the diaphram is small while AF, it focuses at the shooting aperture (f/16 or f/22)...and in lower light it can have a harder time of AF unless the ISO is increased to a more sensitive setting.
   
2. If the opening of the diaphram is large while AF, it focuses with the lens aperture wide open (e.g. f/2.8 or f/4), and closes down the aperture only for the actual exposure.
   

Thank you for your reply! I get what you say, my 7DI works the same way. With my RP option 2 occurs. No matter how small an aperture I chose, there is always a thin depth of field visible through the EVF/on the screen whilst focusing. The camera therefore fouses with a large aperture and should therefore have a lot of light to work with. Still, however, in some situations if the ISO is too low, it will not focus sometimes on darker subjects. Increasing the ISO with just 1 stop and focusing suddenly may work perfectly again. (I am pretty sure the aperture did not change in the meantime, since I would recognize the clicking sound this makes). This dependence on ISO for it's focus performance seems unnecessary to me, though I may overlook some technical details here.

So the RP appears to be behaving like dSLR live view...wide open aperture until the moment of exposure, even when Exposure Simulation is enabled.

So, about focus issue in low light...it is happening in spite of wide open aperture, but is improved by ISO setting...apparently the AF detection makes use of amplified signal (higher ISO merely = higher gain (amplification) of analog sensor signal. Difficulty lock focus on black cat in daylight is apparently a reflection of 'AF difficulty' = comparing low signal level differences.

Many pros who tried wedding coverage with mirrorless have found they pick up their dSLR in low light circumstances, as AF slowness or difficulty locking focus in low light was an issue with mirrorless designs (including Sony A9...so it is not merely a bad implementation in the lower-market RP!). Combined with an inability to use IR AF beam from flash seems to point to an apparent Achilles Heel of mirroless, and it will be interesting to see if Canon figures out a solution in the R5.

Thanks for your reply! It's interesting to hear that some pro's still choose their dSLR over their mirrorless in low light situations. I would not have thought!

Now, I truely am sorry to sound repetitive here. But to illustrate the case I'm trying to make, I'd love to ask you a rhetorical question:
How would you feel about a camera if it's AF performance is directly correlated to the brightness setting of your back-LCD?

That would be nuts, wouldn't it be?

But... the AF performance of the EOS RP with exposure simululation turned on... is just that:
Imagine a scene and subject bathing in the midday sun with tons of light to focus. Yet you decide to dial the shutterspeed to 1/16000 (just pretend the RP supports such shutterspeeds), such that your SIMULATED exposure becomes somewhat dark on the back-LCD of your camera. And now... suddenly your camera won't focus because the SIMULATED exposure has too little contrast. Now when I double the ISO or halve my shutterspeed suddenly the AF becomes very snappy. This seems quite odd to me, and I can't imagine the A9 suffering from this.

I will try and upload a short video tomorrow that illustrates my point.

The mirrorless bodies are simply Live View all the time, so yes that is how it works. Exposure sim just simulates what would might happen should the aperture be shut down. As to the results found above, I am not sure what to think but it doesn't make sense to me. I may have to try the same with my M50, but what you call out isn't logical to me. The sensor is the only thing used to AF, it doesn't use the rear screen for anything, it is an output device.

If you don't have exact control over the lighting in the room nor put the AF point on the exact point each time, that could be part of it. Just the change of light glinting off an edge of an object could have a huge impact on the AF performance in low light situations.

Canon mirrorless bodies aren't quite up the task of replacing their DSLRs. The R5 might be close to changing that, but its release keeps getting pushed so we don't know.

Also the RP is an entry level camera and cannot be compared to the A9 by any stretch of the imagination.

RensBerg wrote in post #19083275
Now, I truely am sorry to sound repetitive here. But to illustrate the case I'm trying to make, I'd love to ask you a rhetorical question:
How would you feel about a camera if it's AF performance is directly correlated to the brightness setting of your back-LCD?

That would be nuts, wouldn't it be?

Given that AF is determined by contrasting areas, it seems intuitively obvious that it is a VISUAL task...not one determined by sound (which works even in the dark). So the fact that mirrorless AF does not do so well in dark conditions seems to me no different than using my own EYES to focus!
I also find it disappointing that the optically based IR AF aid is not usable, but I think I understand WHY...there is an IR filter in front of the sensor, so any IR-based AF simply cannot be detected by the sensor!

I guess I fail to see how that response answers any part of the question posed?

The alleged issue raised is that when you turn on exposure simulation (which is a software hack to give you an indication of what your settings may produce for exposurel on the rear screen), it somehow also affects the AF performance of the overall system.

This is HIGHLY unlikely and there are almost assuredly other conditions at work here to create the illusion that the rear screen display is somehow affecting the AF system. The only possible situation here would be that the exposure simulation and rear display driver is sucking so much processor power away from the task of AF, that it somehow affects it. That is the only idea I could have to explain how the two correlate to the effects observed, but again highly unlikely. Given the RP is such an entry level body, they could have gone with the least expensive processor configuration, and now one process could affect another.

But what if the RP executes AF at the shooting aperture when you are in SERVO mode? That is what my Fuji X-H1 does, and if you stop down enough, it will be unable to obtain an exposure lock. Switch to AF-S (One Shot) and this does not occur.

Not guaranteed the MILCs AF the same way the DSLRs do; Canon's DPAF is not sensitive at all to horizontal line detail, you can demonstrate that to yourself. Will be interesting to see what the R5/R6 do.

How about enabling red beam focus assistance on the flash?