Please select Into the mobile phone version | Continue to access the computer ver.
Detailed study of my P3's image quality (and cold temps)
12Next >
5651 71 2016-1-12
Uploading and Loding Picture ...(0/1)
o(^-^)o
QBK
lvl.2

United States
Offline

There are several recent threads on the forum asking "Are the images from my P3 supposed to look like this?", e.g.:
Focus quality variation across image view,
Something wrong with my camera or is this the quality?,
and New Firmware - Blurry Photos and Video.

(Update:  see also Out of Focus Camera... Please help!)
(Update #2:  this post in the first thread linked above demonstrates an impressive (though tricky) technical fix to a similar problem).

So, even though we're all obsessed with geofencing now... my house isn't in a no-fly zone.  And I've been getting poor images.  So I did some tests to try and figure this out.  I hope this is useful to some of y'all, but really what I'm looking for is feedback.

Backstory:  I got my P3P in June, flew it around in summer weather in New Mexico and Alaska, and got very good images.  Then, in October, my Lightbridge died.  I finally RMAed it in December (had a great experience -- thanks, DJI!) and got it back right after Christmas.  DJI replaced the whole camera.  I updated the firmware, reduced gains to 80 across the board to get rid of vibration, and started flying (in Massachusetts, now -- so it's cold) and shooting.  And I was really disappointed with the image quality that I saw.  So I started Googling for data, found the threads above, and eventually started doing my own tests.

What I saw in the field:  First, a word on methodology.  All the images in this post are unprocessed frame captures from 4K video.  I'm perfectly aware that this does not maximize image quality!  My goal here was to compare the P3P camera against itself at different times.  Furthermore, I have a lot more video footage to work with (especially from the old camera).  Using unprocessed 4K frame grabs throughout ensures a consistent standard.  Finally, this is a more forgiving test than using DNG -- any lens defect that shows up in these tests would be even worse in DNG.  Oh, and yes, I did take the plastic cover off the lens (and inspected it minutely for haze, defects, etc).

Most of the images are ~2MB, and I can only attach <1MB files to a post.  Downsampling them would completely miss the point of this post!  So I'm posting Dropbox links instead.

With that said, here are two shots from my original camera in warm weather (55 - 90 Fahrenheit).  Note that I never used it in colder weather.

Original Camera 1
Original Camera 2

Here are two shots from my new camera in temperatures around freezing (30-40 Fahrenheit).  I haven't had an opportunity to use it in warm weather.

New Camera (cool temps) 1
New Camera (cool temps) 2

And here are two shots from my new camera in sub-freezing temperatures (15-20 Fahrenheit).

New Camera (sub-freezing) 1
New Camera (sub-freezing) 2

I'd really appreciate comments on your perceptions of the sharpness and detail in these images.  (And note that I'm not concerned with composition, color balance, etc!)  I'm looking at these at 100% (1:1) magnification, and I'm not comparing them to DSLR images -- I'm interested in the before-and-after comparison.  

Here's my take:
1.  The old camera is excellent for a small sensor.  In particular, the lens is more or less flawless.  I see pixel level detail across the frame.  Everything is sharp down to the single pixel level.  There's no visible CA (chromatic aberration) in these or any other shots.  There are strong JPEG artifacts, and it's a small sensor, but I have no complaints here.
2.  The new camera is consistently worse except in the lower right corner.  The center and left portions of the image range from "slightly blurry and low-contrast" (above freezing) to "dreadful" in the sub-freezing picture of the house and trees.  In other images, I see quite a lot of CA.  (Which is not a huge concern in itself, but is frequently a symptom of other lens problems... particularly given that it was absent in my original camera).
3.  Temperature seems to make a huge difference.  The images taken above freezing are okay, and if I'd never seen what my original camera could do, I'd be content with them.  But the ones below freezing are seriously blurred.  I estimate that their linear resolution is at least 2x worse, meaning these images would have to be downsized from 8MP to 2MP (1080p) to look as sharp as those taken with my original camera.

Resolution testing:  To try and get some objective data, I did some tests.  First, to see whether vibration was the issue, I shot a bunch of brick walls both (1) while flying and (2) while on the ground.  I also varied shutter speed.  I found no significant difference, but stationary shots seemed at least as good as flying shots, so all remaining shots were taken with the aircraft resting on a stool.

Then, at about 45 degrees, I set up in front of a brick wall and took a bunch of shots of a couple of resolution charts.  I moved them around the frame, captured frames, and then merged the frames to make the following two JPEGs.  One shows several shots of a large resolution chart held at 13' distance.  The second shows several shots of a small resolution chart held at 13' and 25' distances.

Large resolution chart (merged)
Small resolution chart (merged)

In case you're wondering about my methodology, I was pretty careful about splicing the chart crops into the right location, all editing was done at 100% JPEG quality, and I checked the final merged images to confirm that they look identical (at 3:1) to the originals.

I'll skip over most of the nerdy details of the calculations that I did, and go to the punchline.  While the left side is weak, that's mostly because there's a lot of CA (several pixels of red/green banding in high-contrast transitions).  The absolute resolution is still darn good.  I'm getting about 1.3 to 1.5 pixels/line.  In other words, I can resolve 10 alternating black/white lines in a region that's just 13-15 pixels high.  This is about as good as you can get (given aliasing, etc).

CA isn't usually a big problem, but when it's this bad it causes a dramatic reduction in micro-contrast.  In other words, you can still resolve fine black/white lines, but you see them as dark grey and light grey.  This manifests as "blurriness" on real objects (trees, grass, rocks).  You need very sharp high-contrast edges to even identify that it's being caused by CA, and you need multiple length scales to really diagnose it.  CA this bad can be a symptom of something misaligned in the lens, and I suspect that's the situation here.

Okay, one more test.  Above, I said resolution was "1.3 to 1.5 pixels/line".  Actually, it's 1.3 in the early photos, and 1.5 in the later ones.  That made me wonder if temperature was playing a role (even at 45 F) -- was the camera getting worse over time?

Temperature Test (70F down to 22F):  So I made the most boring 4K video ever.  I put my Phantom on a stool outside, facing a resolution chart at 13' away, and took a 21-minute video.  It was 22F outside, and I started the video inside at 70F before taking it outside, with everything already set up.  Afterward, I took 43 frame grabs at 30-second intervals (starting about 8 seconds after the P3P was exposed to cold air), cropped them to the chart, and flipped through them.

Unfortunately, I can't figure out how to make an MP4 out of them without spending a lot of time on it, so I'm just posting four sample times:  0:00, 5:30, 11:30, and 21:00.

(These photos are attached to the post).

Resolution clearly decreases over time.  I can't prove that this is due to temperature, but it's a pretty strong implication.  The lighting does get a bit flatter over time, but this is all at ISO 100, and I don't think the flat lighting should affect resolution this way.

I want to point out, in particular, how the resolution dies.  In the first frame, if you look at the top set of 17 lines, you'll see distinct lines on the right which vanish abruptly at the point marked "6".  This is due to sensor resolution.  There just aren't enough pixels to resolve more lines, and so we see an abrupt transition from lines to grey mush.  Now, look at the last frame.  The entire upper set of lines is mush, but we can look at the lower set of 9 lines (at the right edge, their spacing is identical to that of the upper set).  The transition from "lines" to "mush" is much, much more gradual.  This is due to lens (focus/centering/whatever) failure.  So, whereas before the lens outresolved the sensor, now it's not.

Conclusions:
1.  My new camera's lens (+alignment) is not as good as my old one.  This is causing variation over the frame, CA, and loss of resolution (at least at cold temps).  It's visible in 4K frame grabs as well as DNGs (not shown).
2.  Given the rash of very similar complaints and issues, I tend to suspect there's a batch of cameras with this feature.
3.  Cold temperatures make it much, much worse.  At 22F, I'm seeing not just CA, but actual defocusing/decentering (I can't tell the difference since the lens is fixed-focus).
4.  Letting the aircraft cool down (at least for up to 21 minutes) doesn't help -- it just equilibrates at "bad".
5.  Vibration has nothing to do with this.

Questions:
1.  Especially for the DJI folks:  is this a legitimate RMA issue, and will it get fixed if I do send it in?  Or is this "within spec" (meaning that I just got fantastically lucky with my first camera), and the best I can expect is a replacement camera that's just as bad?
2.  Anybody else feel like doing similar tests?
3.  Do y'all agree with my analysis of the images given here?






Resolution chart at 0:00

Resolution chart at 0:00

Resolution chart at 5:30

Resolution chart at 5:30

Resolution chart at 11:30

Resolution chart at 11:30

Resolution chart at 21:00

Resolution chart at 21:00
2016-1-12
Use props
BigRedBen
lvl.1

United Kingdom
Offline

QBK, Thanks for an extremely well thought out post.

I had this exact problem with a Phantom 3 Advanced that I received as a christmas present , I ended up sending mine back to the distributor and asking to be sent another so we will see if I have the same problem again with the new craft. I hadn't been flying in such cold weather as you but I guess down around 40f and saw significant softening on the left hand side of the frame, with the right hand side remaining sharp-ish.

btw the SN of my camera was P79DCK19041116 if that helps you narrow down any possibilities of a defective batch of cameras.

Cheers
Ben
2016-1-12
Use props
nigelw
First Officer
Flight distance : 518084 ft
United Kingdom
Offline

You certainly have a decentred or badly ground lens & a comparison to prove it.  I'd say you have enough to request a replacement camera since the new one isn't as good as the old one.  I wouldn't bother showing them the below freezing ones though as that's not in the spec

I will see over time if my good copy degrades with temperature, but there isn't much variation here at the moment so it's difficult.  Ideally it needs to be the same subject & subject distance, so heading for the mountains isn't practical for me.
2016-1-12
Use props
terrylewis
Captain
Flight distance : 3517287 ft
  • >>>
United States
Offline

Excellent analysis, thanks for the work! Is it possible that the temperature is causing any significant sensor movement due to materials expansion/contraction with temperature increases/decreases?
2016-1-12
Use props
crbalch
lvl.3
Flight distance : 167057 ft
United States
Offline

Subbed, great post
2016-1-12
Use props
imagines.and.wo
lvl.3

Croatia
Offline


i just send back my p3s to dealer.  
2016-1-12
Use props
mikedaniels
lvl.3
Flight distance : 48986 ft
United States
Offline

temperature affects lenses. Plastic the most. That is why good lenses (glass) will cost more.
2016-1-12
Use props
BigRedBen
lvl.1

United Kingdom
Offline

mikedaniels Posted at 2016-1-13 00:41
temperature affects lenses. Plastic the most. That is why good lenses (glass) will cost more.

I agree...

But I still think this is an issue, there are people out there shooting snowy scenes with Phantoms and getting beautifully sharp images so this softness cannot be 'normal'. I had appallingly soft footage and stills from my P3A at around 5C/40F and I would hardly call that 'exceptional operating conditions' outside of what these aircraft are designed to fly in. If the camera did this at -40C/F then fair enough, I wouldn't be surprised. But I shouldn't have to wait until its 20C+ to get good footage out of my Phantom. I have used GoPros in much colder conditions and suffered none of these issues.

It would seem that a batch of lenses are miss-ground or mis-aligned with the sensor. Fingers crossed my replacement P3A that is being sent to me any day now also doesn't have cataracts. ;)
2016-1-13
Use props
nigelw
First Officer
Flight distance : 518084 ft
United Kingdom
Offline

mikedaniels Posted at 2016-1-13 00:41
temperature affects lenses. Plastic the most. That is why good lenses (glass) will cost more.

That's not why glass lenses cost more.  They cost more because it takes a lot longer to make them & time is money.

The temperature is much more likely to affect the material they're housed in.  Aluminium expands & contracts much more than the lenses, altering alignment & spacing.   If they're not right to start with, they'll get worse as things change.
2016-1-13
Use props
QBK
lvl.2

United States
Offline

BigRedBen Posted at 2016-1-12 13:11
QBK, Thanks for an extremely well thought out post.

I had this exact problem with a Phantom 3 Advan ...

Thanks for your kind reply, BigRedBen!  And good luck with your return process.  I wish I could send mine back to the distributor (Amazon says talk to DJI), but I'm keeping my fingers crossed for a good response from DJI.

The soft-left-side thing seems to be common in a lot of the complaints I've seen.  Of course, there's probably 1000 perfectly happy customers out there for each complaint, so this might be a rare defect... but that's cold comfort to those of us who have it.

Thanks, too, for the S/N info.  My old camera S/N was P78DCE26031632, and my new one starts with P78DCK.  So I'm guessing that:
1.  P78 indicates the P3P camera, while P79 is the P3A camera (no 4K).
2.  Maybe the DCK series is, or contains, the defective lenses?  I wonder if I can get DJI to scrounge up an old DCE-series...
2016-1-13
Use props
QBK
lvl.2

United States
Offline

nigelw Posted at 2016-1-12 13:24
You certainly have a decentred or badly ground lens & a comparison to prove it.  I'd say you have en ...

Thanks for the feedback and advice, nigelw!  That's what I thought, but it's awfully helpful to get another set of eyes on it.  I was hoping one of the DJI guys would drop in a response, but maybe this is the kind of issue they don't want to get involved with (or maybe they're just busy).

I think I'm going to request an RMA tonight.  Fingers crossed.

I had the same problem getting controlled conditions, except worse because I don't have my old camera any more to compare with!  If you want to go to the effort, you could print a resolution chart and take it up to the mountains with you (with a tape measure).  On the other hand, if your copy is good, cross your fingers and keep it that way!
2016-1-13
Use props
QBK
lvl.2

United States
Offline

terrylewis Posted at 2016-1-12 14:03
Excellent analysis, thanks for the work! Is it possible that the temperature is causing any signific ...

Where did you find that disassembly pic, terrylewis?  I spent a while wandering around Google Images and didn't find it.  Don't tell me you took your own apart...?!

I'd be inclined to suspect that the lens is moving relative to the sensor, but trying to say "the sensor is moving" or "the lens is moving" kinda violates the principle of relativity.  (Galilean, not Einsteinian).  But the idea that something is expanding and contracting seems compelling.

If I were a lot braver (dumber?) than I am, I'd start taking the camera apart and try to adjust it.  I suspect that would end just as badly as the lawn tractor that I took apart 12 years ago...
2016-1-13
Use props
QBK
lvl.2

United States
Offline

mikedaniels Posted at 2016-1-12 19:41
temperature affects lenses. Plastic the most. That is why good lenses (glass) will cost more.

Is there a reason to believe that the lens in the P3P camera is plastic?  I hadn't heard that.  All the camera lenses I'm familiar with (except lensbaby and similar soft-focus lenses) are glass, with the exception of a few resin-molded aspherical elements... which are generally more expensive than the glass elements they replace.  I've been under the impression that for spherical elements, glass is cheap and easy.
2016-1-13
Use props
pantera989
lvl.4

New Zealand
Offline

Hard to tell without side by side photos of the same subject from both cameras. I can say for sure that the new cold temp photos are definatly worse then the old camera, and new warm temp pics, but can't say for sure the same wouldn't have happened to your old camera under the same conditions.

Also I think shooting still DNG's would have been a better way to test.
2016-1-13
Use props
QBK
lvl.2

United States
Offline

pantera989@gmai Posted at 2016-1-13 22:31
Hard to tell without side by side photos of the same subject from both cameras. I can say for sure t ...

I definitely agree with your first point -- I really wish I could do a side-by-side test!  Unfortunately, it's impossible since I don't have the original camera any more, I'm living 2000 miles away, and it's a different season.  

And your point that the old camera might have gone wonky in similar conditions is well-taken.  It's possible that all P3 cameras go wonky in the cold.  But... I feel like I've seen high-quality stills and video taken in cold conditions.  But this is why I'm asking on the forum, so thanks for your input!

I don't necessarily disagree with your point about DNGs.  I did it this way partly for convenience, and partly because RAW photos are always processed (even if only by the default settings of the RAW converter you used to view them or convert to JPEG).  I've seen nasty arguments that only resolved when it turned out that the two parties were viewing the same RAW file in different programs (e.g. Lightroom and Canon DPP).

When I looked at frame grabs and minimally-processed DNGs taken at the same time, they looked very, very similar to me.  So I stuck with this to avoid any temptation to tweak exposure, color balance, etc.
2016-1-13
Use props
lalvar40
lvl.4
Flight distance : 168540 ft
United States
Offline

Great analysis QBK. Could you take a look at my pictures, I am wondering if my camera is defective too. I am comparing it to my old P2V+. Those look better than the ones on my P3P.Phantom Pilots
2016-1-13
Use props
LordNeo
lvl.4
Flight distance : 548212 ft
Chile
Offline

Don't try to compare with the snowy footage around, it's already processed and most of the CA and sharpness can be fixed in post.
2016-1-13
Use props
omegas
lvl.3
Flight distance : 23757 ft
Russia
Offline

QBK , I have the same problem!

Normal (not perfect) image at home (+24C) and blurry outside (-25C - -5C)...

Any idea how to fix this?
2016-1-13
Use props
nigelw
First Officer
Flight distance : 518084 ft
United Kingdom
Offline

QBK Posted at 2016-1-14 03:10
Where did you find that disassembly pic, terrylewis?  I spent a while wandering around Google Imag ...

http://www.toptoyspace.com/2015/ ... tom-3-video-camera/

Google image search ;-)
2016-1-14
Use props
terrylewis
Captain
Flight distance : 3517287 ft
  • >>>
United States
Offline



Thanks Nigel,
   As linked in the article you referenced, the original disassembly was performed by
5imx, back in May 2015.

Thanks again QBK for the OP!
2016-1-14
Use props
PerlanSwed
Second Officer
Flight distance : 615942 ft
Sweden
Offline

Thanks for the details!
I have the same problem, my old P3P took crisp detailed photo and video.
But it was warranty replaced due crack in the shell.
The new one takes "blurry" picture and videos.
I have sent it in now for warranty repair.

2016-1-14
Use props
QBK
lvl.2

United States
Offline

lalvar40@gmail. Posted at 2016-1-13 23:22
Great analysis QBK. Could you take a look at my pictures, I am wondering if my camera is defective t ...

Thanks, lalvar40!

I looked at the two photos you posted.  Neither of them is really idea for resolution/quality testing, but that's typical (as people have pointed out on this thread, neither are mine).  It's hard to get ideal test photos from an airborne wide-angle camera.

(That said, if you really care about this, you might want to put some effort into finding an appropriate scene, with good lighting, and shooting it with your Phantom sitting on the ground with the motors off to eliminate vibration.)

Anyway, back to your pics.  The first one is hard to get anything from, because I couldn't find directly comparable elements in different parts of the frame.  If you compare the right 20% to the left 20%, they don't have any elements in common at the same distance.  And looking at these in my browser, there's no way to evaluate objectively, so I gotta compare apples to apples.

The second one has the grass in common from left to right, at the same distance (foreground).  Here, I'd say you've got a point.  The lawn on the left is more smeary than the lawn on the right.  However...

...I'm not sure this is out of spec.  No optical system is perfect.  It's a subjective call, but while I think your images are less than perfect, they're as good as a lot that I've seen out there (from people who are happy with their images).  If I have to draw a line, I think I'd decide to be satisfied with your pics.  I think this degree of unsharpness might be par for the course.  But feel free to take my opinion with a grain of salt!

(And, if you plan on flying in cold weather, obviously things could change... but there's no way to know without trying it).
2016-1-14
Use props
QBK
lvl.2

United States
Offline

omegas Posted at 2016-1-14 01:57
QBK , I have the same problem!

Normal (not perfect) image at home (+24C) and blurry outside (-25C - ...

omegas:  Unfortunately, I don't have any great ideas, except RMAing it or (if out of warranty) replacing the camera.  And there's no guarantee that'll work.  I'm rather afraid of finding out that the first batch of cameras was just remarkable, and what we're seeing now is par for the course.  It's possible that they're all like this, and that 99% of users are just content with it.  I looked at DJI-Tim's color demonstration footage yesterday, and (even though he's shooting in 1080p), it looks like his left side is even blurrier than mine -- in warm weather.  And he's not complaining about it.

Also, as other posters have noted, DJI doesn't seem to officially recommend flying below 0C, so you may get limited sympathy for problems at -25C.  Which stinks, but whatcha gonna do?
2016-1-14
Use props
QBK
lvl.2

United States
Offline

LordNeo Posted at 2016-1-14 00:22
Don't try to compare with the snowy footage around, it's already processed and most of the CA and sh ...

LordNeo:  A fair point.  In my experience, though, genuine unsharpness (or blur, or defocus, or whatever) isn't really fixable in post.  Softness is -- e.g. video shot with a Canon DSLR.  This is what unsharp mask or the "sharpening" control in Lightroom fixes.  But applying those techniques to images (even RAW) with bad focus or a poor lens gives, at best, a superficial crispiness.  I spent a while fiddling with the DNGs that I shot at the same time as the "sub-freezing" shots above, and even after almost 10 years of experience with Lightroom, I can't pull much out of them.

nigelw and terrylewis:  Thanks for the links!  5imx seems to be impressed by the engineering, and it is a cool little device.  But it also looks pretty clearly like the optics system consists of a lens, a sensor, and a barrel.  No fine adjustments that I can see.  So if something is off, it's off.

PerlandSwed:  Good luck!  In Sweden, you must have even colder temps than here.  Did you get any cold-weather footage with your old P3P, and was it noticeably better?
2016-1-14
Use props
lalvar40
lvl.4
Flight distance : 168540 ft
United States
Offline

QBK Posted at 2016-1-14 09:54
Thanks, lalvar40!

I looked at the two photos you posted.  Neither of them is really idea for reso ...

QBK..thank you very much!  I feel a little bit better now. If I do post processing ( lightroom, GIMP, or PS), yes, I can get much more sharpness.  My doubt came because when I had my P2V+, just the compressed jpegs were great w/o any additional processing.
2016-1-14
Use props
LordNeo
lvl.4
Flight distance : 548212 ft
Chile
Offline

QBK Posted at 2016-1-14 12:14
LordNeo:  A fair point.  In my experience, though, genuine unsharpness (or blur, or defocus, or wh ...

I'll do your tests when the climate changes here (I'm in Pucon, Chile, so we have a wide range of weather, from 30°C to -3°C).
I'm a bit concerned about something else "autoadjusting" on the drone (like temp, EV or saturation) depending on the light and the situation, as that could mess with the testing.
I tried to improve the trees' picture with lightroom, but a lot of "mush" still remains, like a lens completly out of focus.
Have you tried checking the filter cover is ok? (the front part you unscrew to put the ND filters)
2016-1-14
Use props
nigelw
First Officer
Flight distance : 518084 ft
United Kingdom
Offline

QBK Posted at 2016-1-14 03:07
Thanks for the feedback and advice, nigelw!  That's what I thought, but it's awfully helpful to ge ...

"If you want to go to the effort, you could print a resolution chart and take it up to the mountains with you (with a tape measure)."

Sorry, but it's so much hard work carrying the P3 up there with all my other gear, I'll only be taking photos/videos of the scenery to make the most of my time.  It's usually a 700-900m ascent by foot after a 200 mile drive, & I only get up there every couple of weeks or so.
2016-1-14
Use props
nigelw
First Officer
Flight distance : 518084 ft
United Kingdom
Offline

QBK Posted at 2016-1-14 15:14
LordNeo:  A fair point.  In my experience, though, genuine unsharpness (or blur, or defocus, or wh ...

Fine adjustments are possible.  I think you'll find this very interesting

2016-1-14
Use props
QBK
lvl.2

United States
Offline

nigelw Posted at 2016-1-14 16:54
Fine adjustments are possible.  I think you'll find this very interesting

https://youtu.be/4sXtwr ...

Wow.  Wow, wow, WOW.  That's amazing.  Thanks for sharing it!

The focusing instructions at the end, though, are "Use needle nose pliers to carefully rotate the lens until properly focused on the screen."  Which... ouch.  If your goal is to achieve pixel-level sharpness at 4K (which mine is -- and it is actually possible), focusing with the 720P live feed as a "ground glass" isn't going to get there.

Still, it's amazing how easily the whole thing comes apart.  I'm not going to try it while it's under warranty, though!
2016-1-14
Use props
nigelw
First Officer
Flight distance : 518084 ft
United Kingdom
Offline

QBK Posted at 2016-1-14 22:18
Wow.  Wow, wow, WOW.  That's amazing.  Thanks for sharing it!

The focusing instructions at the en ...

I'd be inclined to find/make a better tool to grip the lens though.  Final focusing would be best done on a full-sized image on the screen & making notes as you adjust it a bit at a lime.  If my focus wasn't as good as it is now, I'd probably be in there now, rather than the trial & error of returning it for replacement/repair under warranty.
2016-1-14
Use props
QBK
lvl.2

United States
Offline

LordNeo Posted at 2016-1-14 16:22
I'll do your tests when the climate changes here (I'm in Pucon, Chile, so we have a wide range of  ...

I had the same experience with Lightroom and that picture.  The DNG is (of course) more malleable... but there's nothing I can do about the mush.  I mean, I can crank up clarity and sharpening to +100, and drop the blacks... but then it just looks like a crappy picture that was run through somebody's "Auto Awesome!!!" filter.

I haven't done anything with the filter cover except examine it very carefully with an LED light, and haven't found anything on it or in it (and it seems to be on straight).  I guess as a last resort I could try taking it off entirely for a test, but it looks fine.
2016-1-14
Use props
QBK
lvl.2

United States
Offline

nigelw Posted at 2016-1-14 18:07
I'd be inclined to find/make a better tool to grip the lens though.  Final focusing would be best  ...

Yeah, that would work.  It would be tedious -- take photo, download, look at it, tweak lens, take another photo... -- but you could make it work.

I might resort to that down the road if all else fails, but I just submitted an RMA request and I'm going to see what happens.  If it gets rejected -- or if it comes back equally bad -- I'll take the gloves (and the main board!) off.

Problem is, what I'm seeing looks like misalignment or a decentered lens, rather than just a focus adjustment.  If it was out of focus across the frame, I'd be more hopeful that the manual adjustment would work.
2016-1-14
Use props
omegas
lvl.3
Flight distance : 23757 ft
Russia
Offline

It is no problem to make manual focus adjustment like in this video. Problem is that focus degrades proportional temperature outside. You can get perfect focus at +20C and totally blurry at -10C. Or you can get perfect at -10C and blurry at +20C... And how to fix this correlation i don't know. Any idea?

I have -25C today. I put my warm (+24C) p3p outside and make photos till cam has ~ -20C temperature. It is blurry very fast...

Take a look at 100% https://www.dropbox.com/sh/7ndf8 ... 7IoWgjOIfuTQSa?dl=0

sorry for my english
2016-1-15
Use props
QBK
lvl.2

United States
Offline

omegas Posted at 2016-1-15 05:27
It is no problem to make manual focus adjustment like in this video. Problem is that focus degrades  ...

omegas:  Thanks for doing this test!  Seeing the results from a few other controlled tests is very interesting and useful.

I only looked carefully at your 1st and last images, but I agree -- the image quality degrades noticeably.  Overall, I think I'd rather have your camera than mine:  over a much larger temperature range (44 C, whereas I only tested over about 20 C), I think you're seeing less variation.  And it's a bit more uniform across the frame.  But it's still not good.

Interestingly, your camera is the first I've seen that seems better on the left.  In the last photo, the building on the left still looks good at 100% (although the trees in the background are defocused).  It looks like a clear case of front-focusing at cold temps (whereas I can't tell if my problem is focus or alignment).

Part of what's driving me nuts right now is not knowing what to expect.  If we get a few more test results like yours, where a camera that's perfectly good at 20 C gets blurry at 0 or -20 C, then I'll start to believe that this is just the best that can be accomplished with this camera (and maybe, if it's out of warranty, I'll try opening it up and adjusting focus).  But if there's a chance of making it better, I want to pursue that.

Just to be clear:  have you actually opened up your camera and changed its focus?
2016-1-15
Use props
nigelw
First Officer
Flight distance : 518084 ft
United Kingdom
Offline

QBK Posted at 2016-1-15 13:59
omegas:  Thanks for doing this test!  Seeing the results from a few other controlled tests is very ...

I thought the same.  The lens needs screwing into the body slightly as it's focused too close.  If you can get good infinity focus in the cold, as it warms the focus will come in towards the camera.  However, you should still have good infinity focus due to the depth of field.

The ideal focus distance is 0.85m, where everything from 0.45m to infinity will be sharp, so as long as it never focuses at less the that, everything will be sharp.  However, as it's an aerial camera, focusing a little further out shouldn't be a problem.
2016-1-15
Use props
QBK
lvl.2

United States
Offline

nigelw Posted at 2016-1-15 09:18
I thought the same.  The lens needs screwing into the body slightly as it's focused too close.  If  ...

This reminds me to throw in an extra bit of analysis that I didn't include in my original post.  (Warning: nerding ahead!).

When I searched the forums for relevant discussions, I found some partial and potentially confusing information about depth of field (DOF) and hyperfocal distance for the P3P's camera.  This is a good time to clear that up.

There are two main points that produce confusion.
1.  The P3P camera's lens the same field of view as a 20.7mm lens on 35mm, format... but that doesn't mean it produces identical pictures!  The relationship between DOF and aperture changes with sensor size (crop factor).
2.  DOF is not an absolute -- it's subjective in at least one important way.  An image's DOF is the distance range within which objects appear "sufficiently" in focus.  More precisely, the image of any point source of light within the DOF is "sufficiently" small (similar to a point) on the imaging sensor.  And there is no objective definition of what "sufficiently" means here -- just a conventional one.

The best way to explain Point #1 is via Point #2, so I'll just note in passing that the Phantom 3's f/2.8 lens actually makes an image that's identical to a 20.7mm f/16 lens on full frame.

The root of all this is that, for an ideal lens at any instant, there's exactly one distance that is perfectly in focus.  Point sources at this distance get mapped to point images.  At any other distance, they get mapped to circles (or, for a real lens, blobs or onions or whatever your bokeh looks like).  But any sensor has a finite resolution, so it can't distinguish a sufficiently small circle from a point.

So, modulo some turn-the-crank math, all you have to do is define how big a circle (on the sensor) you're willing to accept as "pointlike".  This diameter (on the sensor, mind you) is the circle of confusion (CoC).  Then:
1.  For any focal distance D, you can define the DOF as the range [Dmin, Dmax] of distances at which point sources generate circles smaller than the CoC you've chosen to accept.
2.  The hyperfocal distance, Dhyper, is the closest distance such that, if you focus at Dhyper, then infinity will be "in focus" in the sense given above.

You don't have to turn the crank on the math; the internet will do it for you at http://dofmaster.com/dofjs.html.  (This is the second best calculator I've found -- I lost the URL for the best one).  However, you need to choose a CoC.   If you go to the URL above and select a given camera, format, or sensor size, it will look up the sensor size and apply what's known (for no good reason) as the Zeiss Formula.  It sets the CoC at 1/1730 of the sensor's diagonal measurement.  This is 0.025mm for full frame 35mm (36mm x 24mm), and 0.005mm for the Phantom 3 camera (which is a 1/2.3" sensor measuring 6.17mm x 4.62mm).
This is where it gets subjective, because that rule dates from the old film days.  35mm film didn't have very good resolution.  On our 4000 x 3000 pixel sensors, the Zeiss formula deems a circle with a 3-pixel diameter to be "acceptable".  Now, you may also find that totally fine!  I'm not telling you what to want!  But...
1.  I'm a pixel peeper.  And I can definitely tell the difference between 1080p and 4K when I watch it on my 4K TV (really!), so I care about resolution.
2.  I've examined images from the Phantom camera at 100% and 300%, and it's capable of resolving single-pixel details.
3.  Most importantly, when I did my tests, I wanted to avoid any blur that was my fault.  So I did the calculation for a 0.002mm CoC, which corresponds to 1/3000 of the horizontal dimension, or about 1.3 pixels per CoC.  This yields the DOF within which the lens is basically outresolving the sensor.

TL;DR Conclusions/summary:  I hope if you read through that, it helped clarify some things.  Here are the actual hyperfocal distances you get (and the associated DOF ranges):

0.005 mm CoC:  This is the old (Zeiss formula) standard, and guarantees a resolution of about 1250 horizontal lines (less than 1080p).  The hyperfocal distance is 3.02 feet, or 0.92 meters.  At that focal distance, DOF ranges from 1.51 feet (0.46 m) to infinity.
0.002 mm CoC:  This guarantees a resolution of about 3000 horizontal lines, comparable to the sensor.   The hyperfocal distance is 7.53 feet, or 2.29 meters.  At that focal distance, DOF ranges from 3.76 feet (1.15 m) to infinity.

I don't know where the camera manufacturer [intended to] set the focal plane, but I'll bet it's beyond 2.3 meters -- if only because infinity sharpness is far more important than the [0.46m , 1.15m] range.  Anyway, when I did my tests, I ignored anything closer than 2.5m (and put my targets at least 4m away).
2016-1-15
Use props
omegas
lvl.3
Flight distance : 23757 ft
Russia
Offline

QBK Posted at 2016-1-15 20:59
omegas:  Thanks for doing this test!  Seeing the results from a few other controlled tests is very ...

I didn't open my camera, because i contact with dji support and waiting for answer about it.

But i did this focus adjustment with other FPV-, action-, mobile phone cameras, it is easy. But in that cases i haven't temperature correlation. I'm just fine tuned focus and have perfect images at all temperatures.

Now i thinking about lost of warranty (if i decide to do this). And chance to find one perfect lens position to have sharp image from -10C to +30C at least. What if it is impossible?! What if it just hardware lens bug?

Anybody saw perfect sharp pictures from p3p/p3a in cold conditions? Can you show it?

I'm not happy even with warm my p3p images. I saw much better examples here...
2016-1-15
Use props
omegas
lvl.3
Flight distance : 23757 ft
Russia
Offline

I'm found this video, that shows p3p lens construction. May be someone understand something in optics and lens design and can see anything, that can help solve our focus problem...



2016-1-15
Use props
al.bratukhin
lvl.1

Russia
Offline

Hello. Have same issue with blurry/defocuse center of image and partly left side after 3-5minutes in -15C flight.
I am ready to try to refocus the lens on my P3P but I dont really think that it will help! because it looks like defocused part of image is just blurry NOT front/back focused.
I've read somebody told that leaving P3 in cold temperature(without battery) for long time before flight helps minimize blur... Any confirmation of this theory?
2016-1-16
Use props
omegas
lvl.3
Flight distance : 23757 ft
Russia
Offline

al.bratukhin Posted at 2016-1-17 02:30
Hello. Have same issue with blurry/defocuse center of image and partly left side after 3-5minutes in ...

Can you show your before/after focusing images?
I'm tried put my p3p outside for ~30 minutes - no result, just totally degradation of focus.

How do you refocusing it? Open and disassemble all camera? or just front glass with black mask?
2016-1-16
Use props
12Next >
Advanced
You need to log in before you can reply Login | Register now

Credit Rules