Phantom 4A/P camera vs. Mavic 2 Pro camera...

[Dronefriend]

BTW, look at this one, quite shocking results from the P4P! Curious if the Mavic 2 is also soo much handicapped. It's nowhere near a proper 24mm fixed focal lens, such as the olympus 2/12...

I know, I’m interested for this reason. Have not had time to test yet

 

[Raptorman0909]

Nice footage. How far did you fly?

Probably a bit further than normal, but this is a fairly remote area in the desert. I did have the airspace in view at all times even if I didn't have the drone in view 100% of the time.


Brian

 

[Raptorman0909]

All lenses except a handful have to be corrected. No such thing as a perfect lens. Yes, the truly great glass (such as the Otus line of Zeiss, for example) achieve near perfection but then it comes at a cost. You don’t want to pay $5000 for a lens in the Phantom that will have all the aberrations corrected.

The lens in the P4P is actually very good for that price point and the loss in resolution post-correction is for practical purposes negligible.

Once again RPP, I agree 100%. The P4P does indeed have a pretty decent lens on it and that was one of the things that impressed me immediately.

As far as a next Phantom, if there is one and I think there will be, I suspect the next one will be the last as the body design is a bit dated so if there is a P5 it may well be the last of the line. Going forward I think a newer design that's perhaps a bit more compact and possibly fold-able will eventually replace the Phantom line. I don't know that the Mavic body is it as I suspect the successor will likely be a bit bigger than the Mavic line to permit a larger battery that could be needed to handle the greater energy budget to handle the image processing for, say, 150Mbps and true 4K60. My suspicion is that the current Mavic line is just not big enough to handle the increased image processing workload.

I could see a scaled up Mavic like drone that will be fold-able, bigger and with a larger battery sufficient to handle the larger drone and the increased image processing workload while maintaining a flight time upwards of 30 minutes. The current Phantom line has been around for a long time and the landing gear in particular is the biggest negative -- a next version should address this particularly to limit the props being in the FOV. The upside to the Phantom body design is that the shape of the body probably provides a decent amount of lift with modest forward motion so any new design should endeavor to provide additional lift while moving. I know that the flight times are higher while moving than standing still.


Brian

 

[flyguy95]

Here’s some image quality comparisons between the M2P and P4P. What are all of your thoughts?

 

[WetDog]

Correction = loss of detail and increase noise (vignetting correction = boosting ISO). The more you have to correct, the more your picture will suffer.

At the level of photography we are talking here, you have to expect significant on camera lens correction. It's what's for dinner.

If you wanted to spend $20,000 on a Zeiss video lens (and another $60,000 for the drone and support), you can expect better optics. On a $1500 product, not so much.

And no, correction does not imply degradation. That's silly. Modern digital / optical algorithms are pretty robust. Whether or not DJI gets it correct is another story, but in general, you can't say that because something is digitally manipulated it's degraded.

 

[WetDog]

Once again RPP, I agree 100%. The P4P does indeed have a pretty decent lens on it and that was one of the things that impressed me immediately.

As far as a next Phantom, if there is one and I think there will be, I suspect the next one will be the last as the body design is a bit dated so if there is a P5 it may well be the last of the line. Going forward I think a newer design that's perhaps a bit more compact and possibly fold-able will eventually replace the Phantom line. I don't know that the Mavic body is it as I suspect the successor will likely be a bit bigger than the Mavic line to permit a larger battery that could be needed to handle the greater energy budget to handle the image processing for, say, 150Mbps and true 4K60. My suspicion is that the current Mavic line is just not big enough to handle the increased image processing workload.

I could see a scaled up Mavic like drone that will be fold-able, bigger and with a larger battery sufficient to handle the larger drone and the increased image processing workload while maintaining a flight time upwards of 30 minutes. The current Phantom line has been around for a long time and the landing gear in particular is the biggest negative -- a next version should address this particularly to limit the props being in the FOV. The upside to the Phantom body design is that the shape of the body probably provides a decent amount of lift with modest forward motion so any new design should endeavor to provide additional lift while moving. I know that the flight times are higher while moving than standing still.


Brian

I'd love it if the Phantom body was deprecated in favor of a Mavic like structure. I love the Mavic for it's portability, but it's simply too small to be able to handle winds that I encounter all of the time. The Inspire is too big.... A bulked up Mavic might be just right.

 

[FlyingUser]

All lenses except a handful have to be corrected. ...

That's correct, question is how much correction is needed? 10% (barrel) distortion means you loose much more effective sensor area/field of view during processing. Same applies to vignetting.

The above mentioned neither too expensive nor too heavy Olympus 2/12 (24mm eqv) for example shows 6% distortion and 1.5EV vignetting. The phantom lens seems significantly worse to me...

I'll take a closer look on my raw files today, 'how much is too much".

 

[Raptorman0909]

I have pro grade DSLR's (Nikon D800E's) and some pretty decent glass that I sometimes correct for and other times not. With video I don't see this being hugely important and the processing effort to do so with video, 4K video, would not be trivial particularly coming from a lossy source. So, lens correction when using good lenses is less necessary. For me barrel is often useful and the negative effects in correcting for distortion like that are often as bad as the original problem in the first place. Now chromatic aberration is something I always correct for when present -- barrel, not so much.

As mentioned before I'm more than happy with the lens on the P4P and was duly impressed from the first flight nearly 2 years ago. Getting fussy with 36MP+ still images is one thing, but for video I see less of a need.


Brian

 

[RPP]

I have pro grade DSLR's (Nikon D800E's) and some pretty decent glass that I sometimes correct for and other times not. With video I don't see this being hugely important and the processing effort to do so with video, 4K video, would not be trivial particularly coming from a lossy source. So, lens correction when using good lenses is less necessary. For me barrel is often useful and the negative effects in correcting for distortion like that are often as bad as the original problem in the first place. Now chromatic aberration is something I always correct for when present -- barrel, not so much.

As mentioned before I'm more than happy with the lens on the P4P and was duly impressed from the first flight nearly 2 years ago. Getting fussy with 36MP+ still images is one thing, but for video I see less of a need.
Brian

I have been able to get compositions and make images with both the P4 and P4P that I would not otherwise be possible. I feel confident I could easily make 16" x 24" prints.

 

[FlyingUser]

I have been able to get compositions and make images with both the P4 and P4P that I would not otherwise be possible. I feel confident I could easily make 16" x 24" prints.

At ISO less than 800, f3.2 no doubt! Even if the correction story hits it harder than the 1/2.3" models, it is still an effectively ~1/1.2" sensor area with a (remaining) good lens performance.

If you wanted to spend $20,000 on a Zeiss video lens (and another $60,000 for the drone and support), you can expect better optics. On a $1500 product, not so much.

Inspire + better than average m4/3 lenses. Not perfect but significantly better.

And no, correction does not imply degradation. That's silly. Modern digital / optical algorithms are pretty robust. Whether or not DJI gets it correct is another story, but in general, you can't say that because something is digitally manipulated it's degraded.

Vignetting correction is practically about raising the iso in the corners. I agree if you start with iso100, you won't notice much if the extreme corners are ISO400. But what about starting ISO3200 and havint iso6400 in the not-soo-much extreme borders?

 

[Raptorman0909]

I have been able to get compositions and make images with both the P4 and P4P that I would not otherwise be possible. I feel confident I could easily make 16" x 24" prints.

And in about 3 years of drone flying I have done very little still photography with them. I keep telling myself to give it a go but so far video is my main application.


Brian

 

[With The Birds]

At ISO less than 800, f3.2 no doubt! Even if the correction story hits it harder than the 1/2.3" models, it is still an effectively ~1/1.2" sensor area with a (remaining) good lens performance.


Inspire + better than average m4/3 lenses. Not perfect but significantly better.

Vignetting correction is practically about raising the iso in the corners. I agree if you start with iso100, you won't notice much if the extreme corners are ISO400. But what about starting ISO3200 and havint iso6400 in the not-soo-much extreme borders?

In camera vignette compensation won’t be about raising ISO in the corners. Digital imaging systems mimic film sensitivity by adjusting the analog gain of the photo sites prior to AD conversion. Not something that might be implemented in practice without adding ridiculous complexity and cost. You will fine the compensation is employed by recalculating individual sRGB pixel values post sensor readout. I would appreciate the luxury of being able to perform that task manually rather than having it forced on me by the manufacturer.

 

[GadgetGuy]

And in about 3 years of drone flying I have done very little still photography with them. I keep telling myself to give it a go but so far video is my main application.


Brian

Spherical panoramas created at different elevations in interesting locations might pique your interest in taking more stills. I need to do more of them myself, as I, too, have been limiting myself to video lately!

 

[FlyingUser]

In camera vignette compensation won’t be about raising ISO in the corners. Digital imaging systems mimic film sensitivity by adjusting the analog gain of the photo sites prior to AD conversion. Not something that might be implemented in practice without adding ridiculous complexity and cost. You will fine the compensation is employed by recalculating individual sRGB pixel values post sensor readout. I would appreciate the luxury of being able to perform that task manually rather than having it forced on me by the manufacturer.

In "civilized" cases ISO setting also happens before the AD conversion, it's not just a simple digital push. I don't know how the Phantom sets the ISO but most prosumer cameras I know do it already on the analogue level at least until ISO1000. Vignetting correction happens on the same level.

 

[With The Birds]

In "civilized" cases ISO setting also happens before the AD conversion, it's not just a simple digital push. I don't know how the Phantom sets the ISO but most prosumer cameras I know do it already on the analogue level at least until ISO1000. Vignetting correction happens on the same level.

How might you independently adjust gain by varying amounts across the sensor?

 

[FlyingUser]

How might you independently adjust gain by varying amounts across the sensor?

Happens on the SoC, just as the general pre-ADC ISO setting. "I" can't adjust it, the manufacturer/sensor manufacturer defines the gain map in their firmware, based on the lens profiles provided by the lens manufacturer.

 

[With The Birds]

Happens on the SoC, just as the general pre-ADC ISO setting. "I" can't adjust it, the manufacturer/sensor manufacturer defines the gain map in their firmware, based on the lens profiles provided by the lens manufacturer.

Agreed- digital domain signal processing. Analog gain is global, can only be adjusted across the whole sensor. Vignetting correction and ISO setting are independent features and implemented by different means.

 

[FlyingUser]

Agreed- digital domain signal processing. Analog gain is global, can only be adjusted across the whole sensor. Vignetting correction and ISO setting are independent features and implemented by different means.

An example how it works. A similar test could be done with the P4P (and P4/mavic), check the edge and center noise at f/2.8 and f/5.6 under the same conditions. Best to use ISO800 or more.

Anyway, DXO results of the P4P over the P4 are quite impressive although it falls behind non-drone 1" cameras, like the G7x.

 

[With The Birds]

An example how it works. A similar test could be done with the P4P (and P4/mavic), check the edge and center noise at f/2.8 and f/5.6 under the same conditions. Best to use ISO800 or more.

Anyway, DXO results of the P4P over the P4 are quite impressive although it falls behind non-drone 1" cameras, like the G7x.

I’m looking forward to seeing the DX0 mark on the M2 Pro. I’m not disputing that manufacturers implement all sorts of corrections in hardware. A good example being the claim that the M2 uses Hasselblads proprietary colour profile. I expect (hope) your wrong in your claim that vignetting correction is achieved by increasing the sensor gain at the sensor edges. A better result would be obtained in digital manipulation. Luckily your assumption is almost certainly wrong as to drive individual transistors at higher gain in the sensor would be very difficult to implement. Increasing the gain across the whole array (higher ISO) by comparison is easy.