Soft corner issue on my Mavic 2 Pro too?

[globetrotterdrone]

We’re all probably nitpicking extremes and for 99% of average users, they probably won’t ever notice but I agree on all points that the sensor / camera needs to be sharp out of the box without blurry edges, etc.

Exactly. There are quite frankly some folks who overcomplicate some of the issues here. I mean, a wide angle lens not stopped down is often soft to the edges deriving from its construction. So some may cry 'wolf', where there is none but there are some problems with QMS as we can speculate with the decentralized optical elements, which are no way normal or build tolerance.

 

[CanadaDrone]

Try not to worry too much about getting a perfect lens (all the sensors are fine - the lens is the issue if you are seeing uneven sharpness - seems to be some confusion there). You'd know if you had a bad sensor - usually that manifests itself with entire lines of dead/stuck pixels or green/red lines, or it simply doesn't work.

I have $3,000 DSLR lenses that show more variation than some of these M2P tests and it is for the most part completely normal. A small amount of lens decentering is considered normal even in very high end lenses, which will create small sharpness variances in different parts of the frame when you shoot a test scene or chart and scrutinize at 100% (usually corners). If it's severe or very central, obviously that is a warranty issue that needs to be dealt with, but if it's only minor, you may very well find your replacement is worse.

Lenses will almost always be sharpest in the center, and that will diminish as you get to the edges/corners - this is true even of some of the best optics money can buy. Stopping down will reduce overall sharpenss after a certain point, but increase corner/edge sharpness because you are using a more central portion of the lens. In the case of the M2P though, after F4 diffraction starts to degrade the image and quite noticeably so beyond F5.6. F8-F11 are unusably bad outside of emergencies, in my opinion.

Most lens issues are also magnified at close subject distances one might use for testing, but are the opposite of how most people use their drones 50-300ft in the air or whatever. Make sure it actually impacts your field usage before opting to roll the dice on a replacement.

The reason you might see some areas of the image sharper than others when using different focus points with the camera otherwise stationary is due to field curvature and many lenses are affected by this to some degree, especially wide aperture wide angle lenses like the one on the M2P. Again shooting at relatively short subject distances at the widest aperture will exaggerate this phenomenon. Curved lenses naturally want to make a curved image, but the sensor in the drone is of course flat. When a lens exhibits field curvature it may be focused, say, 10 feet away in the center of the image, but the focus would only be 7 feet away at the mid part of the image and 6 feet away at the edges. Again, perfectly normal if you need to shoot test scenes and view at 100% to easily see it. Way up in the air this would be covered by the enormous DOF of the tiny 1" sensor and shouldn't be visible.

For general usage, the DOF on the M2P is so enormous (basically from about 3 feet to infinity once you're in the air), I haven't changed my manually set focus setting since I bought my drone. I keep focus peaking on as a confirmation but haven't had to adjust it yet. If your subject is very close you may need to adjust that, and I never recommend using autofocus as it can very easily ruin your footage if it hunts even a tiny bit, especially if you don't notice until you get home which is likely.

My advice is if you need to pixel peep test charts at relatively close distances to notice anything, your drone is fine. If you are performing a test that will coax out phenomenons like field curvature, your drone is fine. If you can see obvious and distracting softness in unnatural portions of the image (i.e. not near the edges/corners) that are visible during normal viewing/usage, then I would consider a replacement.

 

[Dakrisht]

Try not to worry too much about getting a perfect lens (all the sensors are fine - the lens is the issue if you are seeing uneven sharpness - seems to be some confusion there). You'd know if you had a bad sensor - usually that manifests itself with entire lines of dead/stuck pixels or green/red lines, or it simply doesn't work.

I have $3,000 DSLR lenses that show more variation than some of these M2P tests and it is for the most part completely normal. A small amount of lens decentering is considered normal even in very high end lenses, which will create small sharpness variances in different parts of the frame when you shoot a test scene or chart and scrutinize at 100% (usually corners). If it's severe or very central, obviously that is a warranty issue that needs to be dealt with, but if it's only minor, you may very well find your replacement is worse.

Lenses will almost always be sharpest in the center, and that will diminish as you get to the edges/corners - this is true even of some of the best optics money can buy. Stopping down will reduce overall sharpenss after a certain point, but increase corner/edge sharpness because you are using a more central portion of the lens. In the case of the M2P though, after F4 diffraction starts to degrade the image and quite noticeably so beyond F5.6. F8-F11 are unusably bad outside of emergencies, in my opinion.

Most lens issues are also magnified at close subject distances one might use for testing, but are the opposite of how most people use their drones 50-300ft in the air or whatever. Make sure it actually impacts your field usage before opting to roll the dice on a replacement.

The reason you might see some areas of the image sharper than others when using different focus points with the camera otherwise stationary is due to field curvature and many lenses are affected by this to some degree, especially wide aperture wide angle lenses like the one on the M2P. Again shooting at relatively short subject distances at the widest aperture will exaggerate this phenomenon. Curved lenses naturally want to make a curved image, but the sensor in the drone is of course flat. When a lens exhibits field curvature it may be focused, say, 10 feet away in the center of the image, but the focus would only be 7 feet away at the mid part of the image and 6 feet away at the edges. Again, perfectly normal if you need to shoot test scenes and view at 100% to easily see it. Way up in the air this would be covered by the enormous DOF of the tiny 1" sensor and shouldn't be visible.

For general usage, the DOF on the M2P is so enormous (basically from about 3 feet to infinity once you're in the air), I haven't changed my manually set focus setting since I bought my drone. I keep focus peaking on as a confirmation but haven't had to adjust it yet. If your subject is very close you may need to adjust that, and I never recommend using autofocus as it can very easily ruin your footage if it hunts even a tiny bit, especially if you don't notice until you get home which is likely.

My advice is if you need to pixel peep test charts are relatively close distances to notice anything, your drone is fine. If you are performing a test that will coax out phenomenons like field curvature, your drone is fine. If you can see obvious and distracting softness in unnatural portions of the image (i.e. not near the edges/corners) that are visible during normal viewing/usage, then I would consider a replacement.

Agree on all points here as well as some A+ technical knowledge. I remember my Leica lenses 10 years ago distorting like crazy but it was normal at least photographically.

The DOF on the M2P is completely insane in a good way.

Meaning to ask you based on this post:

I haven't changed my manually set focus setting since I bought my drone. I keep focus peaking on as a confirmation but haven't had to adjust it yet.

Can you talk a little about getting the best and sharpest manual focus and how to retain that setting?

It’s difficult to spot focus using my iPhone 8 as the display so have been doing a lot of auto focus with a soft touch to focus before shooting the still.

What exactly is focus peaking on this drone and how does it work? Need to look into this.

I’ve had a few issues when shooting wide high up shots, 200+ feet where DOF is key but certain objects are out of focus. Ultimately, infinity focus is the key here but not sure how to achieve optimum focus via manual on this bird yet. Simply haven’t played around enough.

Adding: best practice for setting infinity focus?

 

[CanadaDrone]

Can you talk a little about getting the best and sharpest manual focus and how to retain that setting?

It’s difficult to spot focus using my iPhone 8 as the display so have been doing a lot of auto focus with a soft touch to focus before shooting the still.

What exactly is focus peaking on this drone and how does it work? Need to look into this.

I’ve had a few issues when shooting wide high up shots, 200+ feet where DOF is key but certain objects are out of focus. Ultimately, infinity focus is the key here but not sure how to achieve optimum focus via manual on this bird yet. Simply haven’t played around enough.

Focus peaking is a feature of mirrorless cameras (like the one on the M2P) and very simply what it does is highlight the edges of things in the frame that are in focus. How it actually works is by detecting the edges in the scene that have the highest contrast, and therefore the best focus. It requires real-time processing of the image, which is possible with the live-view coming off the M2P sensor (or any other mirrorless camera pretty much). It will draw a red line around all the edges of everything that is in focus, and you can define the threshold of what it will report as "in focus". I suggest setting it in the middle so you don't have too much distracting red on your screen, since if focus is set properly it will pretty well always be reporting that everything is in focus.

I use the Smart controller but I don't imagine it would be much harder to see on an iPhone 8.

While it may seem intuitive, don't just manually set infinity focus on the M2P. The sweet spot on every one I have seen so far is 2 or 3 clicks back from infinity, and you *should* pretty much just be able to leave it there, especially if you leave focus peaking on for constant confirmation. To set this, just send the drone up in the air, 20-50 feet or whatever (just not too close to anything) and starting at infinity, dial the focus back until the entire scene is blanketed in the red focus peaking lines (it should be obvious where this peak is). Keep an eye on it, but you shouldn't really have to ever change it after that for most normal flying.

I just took this from Google Images, but it is a good illustration of what focus peaking looks like on-screen. In this example, the rider, his motorcycle, and the green posts to the right are all being reported as in focus (basically everything on the same focal plane). The threshold you set will determine how thick the red lines are, and the discretion level at which the drone will report something as in focus. I leave mine in the middle which I think is just called Normal. Also for those that insist on using autofocus, at least with focus peaking you will get a visual queue when the drone hunts or drifts in/out of focus mid-shot.

 

[globetrotterdrone]

Well spoken and excellent summary @CanadaDrone
I am more than pleased with my new one and don't want to change anything again. The first one was faulty in optical terms, but gladly that turned out to be only a short time problem.
Happy and safe flying everyone!

 

[Not A Speck Of Cereal]

Focus peaking is a feature of mirrorless cameras (like the one on the M2P) and very simply what it does is highlight the edges of things in the frame that are in focus. How it actually works is by detecting the edges in the scene that have the highest contrast, and therefore the best focus. It requires real-time processing of the image, which is possible with the live-view coming off the M2P sensor (or any other mirrorless camera pretty much).

Just a nit: I have a DSLR with focus peaking (Nikon D850). Granted, it only works in Live View (when the mirror is up), but it's the same thing ... and it's not a mirrorless camera.

Either way, it's nice to have.

Chris

 

[WithTheBirds]

Have you tried the Lightroom “enhance details” conversion? Kinda a pain just to get what should already work but I’d be interested to see if that fixes it.

Also the brand new Camera Raw for Lightroom and Photoshop has the new “texture” slider which can be applied globally or with a local adjustment brush. It basically sharpens fine details only so you can jack it up to get texture without making the larger details look over sharp.

Again a pain to have to do just to get back to normal but I’d like to see if it solves this issue.

Why? An apparent improvement in how the demosaic algorithm is applied in processing the DNG's can't recover the details that were never recorded. The solution if to have the unit replaced/repaired. Why suffer and compromise?

 

[CanadaDrone]

Just a nit: I have a DSLR with focus peaking (Nikon D850). Granted, it only works in Live View (when the mirror is up), but it's the same thing ... and it's not a mirrorless camera.

Either way, it's nice to have.

Chris

I also have a Nikon D850 - when the mirror is not in play (in live view), it works because in that state it is functioning as a mirrorless camera and the camera is able to process the live image from the sensor. You are also limited to contrast-detect AF in that mode. If the mirror is down, it won't work because you are of course looking at the scene through a set of mirrors and a pentaprism.

Virtually every interchangeable lens mirrorless camera or higher-end mirrorless camera these days has focus peaking. Where it's most useful is with wide aperture manual focus primes, as you can quickly and easily determine focus in the EVF. You generally won't find it on cheaper mirrorless cameras like smartphones or your typical "point & shoot" camera, even though it would be very easy to implement it. The DSLRs that have it can only do so in live view where it operates exactly like a mirrorless camera. I personally can't stand looking through an EVF

Fuji actually has a hybrid viewfinder that allows focus peaking in an optical viewfinder with an overlay, which is pretty neat.

 

[globetrotterdrone]

MagicLantern on Canon 5D Mark II also featured such nifty little helpers like Focus peaking and zebras when in LiveView ...
I was astonished that the Mavic has them while the expensive Canons still don't (I think the new R and RP have it).

Well, it's still a bit unbelievable to have such a cool device like the Mavic 2 Pro in such compact size with such quality.
What times to be alive.

 

[Not A Speck Of Cereal]

I also have a Nikon D850 - when the mirror is not in play (in live view), it works because in that state it is functioning as a mirrorless camera and the camera is able to process the live image from the sensor. You are also limited to contrast-detect AF in that mode. If the mirror is down, it won't work because [...]

Yes, yes -- like I said, just a nit (saying that focus peaking is limited to mirrorless cameras is not really true).

Btw: I mostly only use the D850 Live View for manual focus tripod work, with occasional AF when speed doesn't matter (to help lock focus before I lose the light, for instance).

Chris

 

[Dakrisht]

Focus peaking is a feature of mirrorless cameras (like the one on the M2P) and very simply what it does is highlight the edges of things in the frame that are in focus. How it actually works is by detecting the edges in the scene that have the highest contrast, and therefore the best focus. It requires real-time processing of the image, which is possible with the live-view coming off the M2P sensor (or any other mirrorless camera pretty much). It will draw a red line around all the edges of everything that is in focus, and you can define the threshold of what it will report as "in focus". I suggest setting it in the middle so you don't have too much distracting red on your screen, since if focus is set properly it will pretty well always be reporting that everything is in focus.

I use the Smart controller but I don't imagine it would be much harder to see on an iPhone 8.

While it may seem intuitive, don't just manually set infinity focus on the M2P. The sweet spot on every one I have seen so far is 2 or 3 clicks back from infinity, and you *should* pretty much just be able to leave it there, especially if you leave focus peaking on for constant confirmation. To set this, just send the drone up in the air, 20-50 feet or whatever (just not too close to anything) and starting at infinity, dial the focus back until the entire scene is blanketed in the red focus peaking lines (it should be obvious where this peak is). Keep an eye on it, but you shouldn't really have to ever change it after that for most normal flying.

I just took this from Google Images, but it is a good illustration of what focus peaking looks like on-screen. In this example, the rider, his motorcycle, and the green posts to the right are all being reported as in focus (basically everything on the same focal plane). The threshold you set will determine how thick the red lines are, and the discretion level at which the drone will report something as in focus. I leave mine in the middle which I think is just called Normal. Also for those that insist on using autofocus, at least with focus peaking you will get a visual queue when the drone hunts or drifts in/out of focus mid-shot.

Fantastic summary. Thanks. Playing with this now.

 

[brett8883]

Why? An apparent improvement in how the demosaic algorithm is applied in processing the DNG's can't recover the details that were never recorded. The solution if to have the unit replaced/repaired. Why suffer and compromise?

First of all I was the first person to suggest it was a defect that needed replacement.

Secondly that was in response to Paul’s note about basically doing a sort of pano/super resolution to correct distortion after the point was already made that it was a manufacturing defect that required replacement.

He was talking about how he was unhappy with the focus ability of the camera. It wasn’t meant to be a suggested alternative to resolve the OPs issue.

 

[WithTheBirds]

First of all I was the first person to suggest it was a defect that needed replacement.

Secondly that was in response to Paul’s note about basically doing a sort of pano/super resolution to correct distortion after the point was already made that it was a manufacturing defect that required replacement.

He was talking about how he was unhappy with the focus ability of the camera. It wasn’t meant to be a suggested alternative to resolve the OPs issue.

Gotcha- likely helpful advice to Paul yes. Btw- does capture One support the M2 DNG? It might be a recent upgrade if it does. It didn’t when I last checked.

The issue is unlikely to be a sensor defect. They are a pass/fail proposition off the production line. Poor alignment of sensor and/or optical elements in assembly is almost certainly the culprit.

 

[brett8883]

Gotcha- likely helpful advice to Paul yes. Btw- does capture One support the M2 DNG? It might be a recent upgrade if it does. It didn’t when I last checked.

The issue is unlikely to be a sensor defect. They are a pass/fail proposition off the production line. Poor alignment of sensor and/or optical elements in assembly is almost certainly the culprit.

Sensor issue... sensor alignment issue ?‍ lol this is a tough crowd

 

[Paul2660]

C1 vr 12 supports the MP2 dng. No dedicated color profile but essay enough to find one. Lens details are picked up so the distortion and curved horizon issues are taken care of. Curved horizon occurs when camera is pointed up.

Paul C

 

[Dakrisht]

Did some focus testing on my second M2P today (first one slammed into a building due to mag interference... oops) and just totally blown away by the DOF, sharpness, derail, tonality, color - everything about this camera is mind blowing.

The focus peaking is fun but certainly annoying and gets in the way of composition. Playing around with the different modes but in general, once focused, I disable it and shoot away.

Interestingly enough, infinity is not as sharp as 2-3 clicks back - weird and wonder why?

Photo attached for fun, minimal edits, DNG RAW to JPEG export via LR CC. Seems the forum attachment algorithm has compressed it a touch more but oh well.

 

[globetrotterdrone]

Interestingly enough, infinity is not as sharp as 2-3 clicks back - weird and wonder why?

In short: it is said that AF systems needed that change, where in the old analogue MF days infinity ∞ was "correctly" marked.
See: Who Killed Infinity Focus?

Next time I get to fly, I will focus to something in distance and then set to MF ... guess I will never need to refocus (except if something very close in the foreground needs it).
But even with my TS-E manual lenses from Canon, a tad below infinity is always sharper.

 

[FormerMember07122019_2]

Did some focus testing on my second M2P today (first one slammed into a building due to mag interference... oops) and just totally blown away by the DOF, sharpness, derail, tonality, color - everything about this camera is mind blowing.

The focus peaking is fun but certainly annoying and gets in the way of composition. Playing around with the different modes but in general, once focused, I disable it and shoot away.

Interestingly enough, infinity is not as sharp as 2-3 clicks back - weird and wonder why?

Photo attached for fun, minimal edits, DNG RAW to JPEG export via LR CC. Seems the forum attachment algorithm has compressed it a touch more but oh well.

I really like the composition of this photo. Symmetrical with a hint of asymmetry. Good job.

 

[Dakrisht]

I really like the composition of this photo. Symmetrical with a hint of asymmetry. Good job.

Thanks ??