Mavic Pro-2 vs Zoom and P4P - Line Skipping + Sensor Heat?

[Robbyg]

I gave DJI one week to view the video privately. They watched it almost 80 times in that week but refused to answer me or say anything. If they had said I was right or wrong or just called me an @ss hole, I would have added whatever they said in my video to be fair.

This was never intended to be a "hit peice" on the Mavic-2 Pro. I own both Mavic-2's and I'm flying both.

Honestly, I'm VERY supprised this video has gotten the traction that it has. I figured that it would get 100 views and people would say "I dont understand this and I dont care anyway!" I was wrong about that!

Part of the problem is "us". We actually had these very high expectations that we wanted a 1inch-type sensor with PHANTOM 4 PRO performance in a Mavic. Those are VERY UNREALISTIC expectations from a technical AND marketing point of view! We just want DJI to "simply" cannibalize their P4P....and DJI does not agree with us!

Look.....the fact that DJI jam packed a 1inch-type into a tiny cube and kept it from overheating is amazing. Its possibly the smallest 1inch-type sensor camera ever created by anybody to date. No....they are not giving us the pixel count that we want. But the INDIVIDUAL quality, size, signal to noise ratio and dynamic range of each photsites is very good. (We just dont like the low number of them being used)

From a "legal" point of view, line skipping does not use the full sensor. But "legally", pixel binning does. Its funny though....the mode that reads all the photosites (FOV) is the worst performer and the mode that reads a FRACTION of the sensor is the best mode!! (HQ) Thats ironic!!!

We are all demanding that DJI cannibalize the Phantom 4 and future Phantom 5 series right now....today!! DJI just does not want to do this folks!

Actually,...I cant say I blame them.

Cliff

Cliff while I get your point and it may be true if one looks at it from a certain angle. I would still say that DJI is mainly the one at fault. All to often they use a lot of marketing hype including certain YouTube reviewers to paint a picture that is not complete and then let us fill in the blanks with what would seem to be the logical answers. Then when we find out we are completely wrong, DJI is some how is off the hook.

This marketing scheme is used way to often of late by DJI and while it may be legal it's really not fair to those who paid out their money with an expectation that everything would be the way they implied it would be.

I don't know about the rest of you who have not bought a MP Pro but I will wait until they either fix the problem or they come out with an MP that uses the 1" sensor completely just like the P4P.


Rob

 

[yysc]

Presumably the increased aliasing is due to the low-pass filter being optimized for 4k. If the sensor site data are being combined on the sensor to allow faster recording at lower resolution then the filter will not be set low enough to prevent aliasing. That's a tradeoff that is difficult to get around.

The low pass filter (or anti aliasing filter) should be optimized for 20MP otherwise there would be a decrease of quality in photo mode and there isn't. Photos look pretty much identical to the P4P. The Mavic 1 Pro (with fixed HW) has a 12MP sensor and a better 1080p60 which doesn't negatively impact the 4K quality.

DJI in the M2P is simply in all video modes skipping lines, groups of pixels or whatever pattern they use, which is one of the worst ways to subsample a sensor.

 

[sar104]

The low pass filter (or anti aliasing filter) should be optimized for 20MP otherwise there would be a decrease of quality in photo mode and there isn't. Photos look pretty much identical to the P4P. The Mavic 1 Pro (with fixed HW) has a 12MP sensor and a better 1080p60 which doesn't negatively impact the 4K quality.

DJI in the M2P is simply in all video modes skipping lines, groups of pixels or whatever pattern they use, which is one of the worst ways to subsample a sensor.

Agreed - it should be opimized for 20 MP, not 4k, but that just makes it even worse. The low-pass filter will be set higher than for the 12 MP sensors and for anything less than full sensor readout (rather than combining sites on the sensor chip and reading out subsampled raw data) any given reduced resolution, including 1080p60, will look better with a lower-MP sensor with a lower low-pass filter.

I think perhaps we are saying the same thing and you are simply arguing that sensor sub-sampling, which will always result in increased aliasing, should not be used to get 1080p video. Is there similar aliasing in 1080p30 or are they reading out at full density in that mode and downsampling in software?

 

[chunky]

I would absolutely agree with you IF DJI had set these expectations. They did not and it doesn't matter how many times you guys accuse them of it. They claimed one inch sensor, check. They claimed 4k 30 fps, check. They flat out told us that the HQ mode uses less of the sensor and gives better quality, check. They claim better low light photos using hyperlight, check. Etc. I'm not saying it's perfect but to blame DJI for the expectations YouTube set is ridiculous. DJI never claimed that the M2P was a foldable P4P. They claimed it was an improvement over the Mavic Pro. I watched the launch event twice, I read the manual and I looked at the specs before I dropped $2000. I knew what to expect and I got exactly that.

They write 4K resolution - CHECK!!

The M2P also doesn‘t cannibalize the P4pro since it has no „professional“ 4K60p, adding the ascent and descent numbers are higher with the P4pro, the controller is more ergonomic with better sticks and a better gimbal wheel.

 

[yysc]

Agreed - it should be opimized for 20 MP, not 4k, but that just makes it even worse. The low-pass filter will be set higher than for the 12 MP sensors and for anything less than full sensor readout (rather than combining sites on the sensor chip and reading out subsampled raw data) any given reduced resolution, including 1080p60, will look better with a lower-MP sensor with a lower low-pass filter.

I think perhaps we are saying the same thing and you are simply arguing that sensor sub-sampling, which will always result in increased aliasing, should not be used to get 1080p video. Is there similar aliasing in 1080p30 or are they reading out at full density in that mode and downsampling in software?

Yes we are saying the same thing, the right way to do this is supersample (like the P4P, Parrot Anafi both 20MP drones do), reading all the pixels (around 16.6MP in 16:9 crop) and then combining all these values into the 8MP frame for 4K FOV or 2MP for 1080p. Granted the sensor cannot read 16.6MP@60fps so full supersampling cannot be done at 1080@60, but at least it should be as good as in other drones with 20MP sensor and it isn't currently.

 

[Lon Denard]

Presumably the increased aliasing is due to the low-pass filter being optimized for 4k. If the sensor site data are being combined on the sensor to allow faster recording at lower resolution then the filter will not be set low enough to prevent aliasing. That's a tradeoff that is difficult to get around.

Nope. The M2P is garbage and DJI are a bunch of criminals! We won't be having any reason injected into this discussion mister!

So to sum it up, the M2P doesn't shoot decent video in any setting and anyone that says otherwise is a DJI fanboy.

 

[Lon Denard]

They write 4K resolution - CHECK!!

The M2P also doesn‘t cannibalize the P4pro since it has no „professional“ 4K60p, adding the ascent and descent numbers are higher with the P4pro, the controller is more ergonomic with better sticks and a better gimbal wheel.

Agreed. So go buy a P4P is what I'm saying. If they produce a Spark Pro, I'm not going to expect it to be as good as the Mavic Air.

 

[yysc]

They write 4K resolution - CHECK!!

The M2P also doesn‘t cannibalize the P4pro since it has no „professional“ 4K60p, adding the ascent and descent numbers are higher with the P4pro, the controller is more ergonomic with better sticks and a better gimbal wheel.

And the user manual says (page 38)

Mavic 2 Pro/Zoom
User Manual

2018 DJI All Rights Reserved.

Video Edit
The Mavic 2 Pro supports video formats MP4 and MOV, providing Normal, D-Log, and HLG color
modes. In Normal mode, H.264 and H.265 code formats are supported. In D-Log or HLG mode, only
H.265 is supported.

For the Mavic 2 Pro, Full FOV down samples from the 5.7K sensor to 4K resolution while HW crops in
the center for finer image quality but less FOV. Full FOV view is 75° and HQ view is 55°. You can select
between them according to your actual shooting demands.

https://dl.djicdn.com/downloads/Mavic_2/20180823/Mavic 2 Pro Zoom User Manual V1.0.pdf

Down sample, NOT CHECK! , this is subsample.

 

[yysc]

Agreed. So go buy a P4P is what I'm saying. If they produce a Spark Pro, I'm not going to expect it to be as good as the Mavic Air.

At least for the US market the P4P V2.0 and the M2P have the same price ($1499). One would expect similar quality.

The Spark is $399 whereas Mavic Air is $799 so yes, I would expect better quality paying double. They will sell no Spark Pros if they price them at $799 considering the platform limitations due to the size.

 

[Lon Denard]

Here's the video I watched before I got mine.

 

[Sockeye]

So to sum it up, the M2P doesn't shoot decent video in any setting and anyone that says otherwise is a DJI fanboy.

My Gopro Hero 5 also shoots decent videos... But would I pay 1500$ for it?
(btw. with more details in 4K than the FOV)

 

[Zukoota]

So basically, either buy a mavic 2 pro where you would be forced to use a narrow field of view and accept more noise (part of buying a bigger sensor is for improved low light performance) and don't use the full field of view which provides softer images than the mavic air. Or, buy the substantially cheaper zoom where you can adjust your field of view with an optical zoom and retain the same image quality.

Cool, decision made - im returning the pro and getting the zoom.

 

[yysc]

So basically, either buy a mavic 2 pro where you would be forced to use a narrow field of view and accept more noise (part of buying a bigger sensor is for improved low light performance) and don't use the full field of view which provides softer images than the mavic air. Or, buy the substantially cheaper zoom where you can adjust your field of view with an optical zoom and retain the same image quality.

Cool, decision made - im returning the pro and getting the zoom.

Yep, in HQ mode the recorded area is barely bigger than the 1/2.3" sensor of the Zoom. Still slightly better SNR but far less than a proper FOV supersample. A waste of a 1" sensor.

I'm outside of the return window and still hope DJI will improve current FW. If it turns out they fix it with a Mavic 2 Platinum I'll patiently wait with my M2P until somebody else releases a foldable drone with a 1" sensor (2 years?) and buy that and sell the M2P.

 

[chunky]

HQ mode is „ok“, men, it really is.

FFOV is described differently and quality is really low.
This has to be improved by DJI to fulfill the specs and quality demands.
We didn‘t buy a sub 2.7K drone but got one!
FFOV resolution is clear behind an at least „native sampling“ 4K camera (like HQ mode), nothing below is acceptable.
„RAW subsampling“ describes a kind of binning (DJI reps), so in theory max. 2.7K resolution is achievable - that‘s stated nowhere before the buy!
They state 4K resolution that is still nowhere proved by DJI.

 

[sar104]

Nope. The M2P is garbage and DJI are a bunch of criminals! We won't be having any reason injected into this discussion mister!

So to sum it up, the M2P doesn't shoot decent video in any setting and anyone that says otherwise is a DJI fanboy.

That's what people are saying, anyway®.

 

[sar104]

And the user manual says (page 38)

Mavic 2 Pro/Zoom
User Manual

2018 DJI All Rights Reserved.

Video Edit
The Mavic 2 Pro supports video formats MP4 and MOV, providing Normal, D-Log, and HLG color
modes. In Normal mode, H.264 and H.265 code formats are supported. In D-Log or HLG mode, only
H.265 is supported.

For the Mavic 2 Pro, Full FOV down samples from the 5.7K sensor to 4K resolution while HW crops in
the center for finer image quality but less FOV. Full FOV view is 75° and HQ view is 55°. You can select
between them according to your actual shooting demands.

https://dl.djicdn.com/downloads/Mavic_2/20180823/Mavic 2 Pro Zoom User Manual V1.0.pdf

Down sample, NOT CHECK! , this is subsample.

And back to the original question - how do we know that it is sub-sampling rather than down-sampling?

 

[Roamer105]

I would absolutely agree with you IF DJI had set these expectations. They did not and it doesn't matter how many times you guys accuse them of it. They claimed one inch sensor, check. They claimed 4k 30 fps, check. They flat out told us that the HQ mode uses less of the sensor and gives better quality, check. They claim better low light photos using hyperlight, check. Etc. I'm not saying it's perfect but to blame DJI for the expectations YouTube set is ridiculous. DJI never claimed that the M2P was a foldable P4P. They claimed it was an improvement over the Mavic Pro. I watched the launch event twice, I read the manual and I looked at the specs before I dropped $2000. I knew what to expect and I got exactly that.

I agree with you and would like to add this: Judging the camera from its still performance which means optics and sensor, it beats all previous DJI products and it does this without cropping off vignettes and soft corners (Remember the early complaints of the P4P). When it comes to video, given the superior quality of the optics, it is all just a firmware issue dealing with video pre-processing. Since it has a significantly more advanced image processor, there is no reason to believe DJI is not going to improve on that.

Aside from some of the obvious video issues left on the table, It is unfortunate that us drone videographers have been kind-of “brainwashed” looking at and getting used to over sharpened 8-bit footage and now that we get 10-bit video, well it seems to look soft. Yet when comparing individual frames to equivalent stills taken by high quality SLRs they look much more like the stills than a frame taken from the MP 1 for example.

 

[cochitiman]

I have the M2pro. Cliffs video confirmed what I was seeing in 4k. Both settings lacked clarity to the point that even I could see no clearer resolution than my Mp1 got in 4k. More vibrant color and better in dim conditions as well. I compared my Mp2 in 4k to my Mp1 in 2.7. The Mp1 in 2.7 was actually a tiny bit better. The 2 in 4k should have looked way better. That's why I bought it. I just sent it back this morning and ordered a Mp zoom as I think I'll be happier with that

 

[Lon Denard]

My Gopro Hero 5 also shoots decent videos... But would I pay 1500$ for it?
(btw. with more details in 4K than the FOV)

How fast does it fly? The last time I bought a flying GoPro, it cost over $1100, had horrible range, routinely lost gps, zero obstacle avoidance, limited flight modes, meh camera, etc.

 

[cochitiman]

I have the M2pro. Cliffs video confirmed what I was seeing in 4k. Both settings lacked clarity to the point that even I could see no clearer resolution than my Mp1 got in 4k. More vibrant color and better in dim conditions as well. I compared my Mp2 in 4k to my Mp1 in 2.7. The Mp1 in 2.7 was actually a tiny bit better. The 2 in 4k should have looked way better. That's why I bought it. I just sent it back this morning and ordered a Mp zoom as I think I'll be happier with that

I meant the Mp2 has better color and is better in dim conditions in 4k. Is no better in sunny conditions than Mp1 in 2.7