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

Cliff_622 · Oct 5, 2018

nomnomnom said:
Does anyone know in FOV mode, is the camera even an improvement from mavic 1 at all? Can someone please let me know thanks

The original Mavic Pro had a dinosaur Ambarella A9 processor that literally dates all the way back to the GoPro Hero 3!! It had a TERRIBLE temporal noise reduction algorithm that turned shadows in to a waxy soup. It's multiframe noise reduction caused "chunks" of pixels to cancel out noise and leave blotchy artifacts everywhere. Its h.264 codec maxed out at 60mbp/s.

I can say without a doubt that either Mavic-2 is light years beyond the original Mavic Pro!

nomnomnom · Oct 5, 2018

Cliff_622 said:
The original Mavic Pro had a dinosaur Ambarella A9 processor that literally dates all the way back to the GoPro Hero 3!! It had a TERRIBLE temporal noise reduction algorithm that turned shadows in to a waxy soup. It's multiframe noise reduction caused "chunks" of pixels to cancel out noise and leave blotchy artifacts everywhere. Its h.264 codec maxed out at 60mbp/s.

I can say without a doubt that either Mavic-2 is light years beyond the original Mavic Pro!

I hear what you are saying, but in FOV mode the image is really soft, like literally not wearing glasses

Lon Denard · Oct 5, 2018

nomnomnom said:
Does anyone know in FOV mode, is the camera even an improvement from mavic 1 at all? Can someone please let me know thanks

That's a tough one. I believe with some post it can be. However according to many of these guys, that's a no. For the same money, you may want the P4P. I personally fly in HQ mode and am pretty happy.

I messed up the editing and I'm no cinematographer but it's decent anyway.

dwallersv · Oct 6, 2018

Category 5 said:
Certainly possible. One of my first thoughts when i saw Mavic 2 was that the Phantom was obsolete. Hopefully they can bring out then P5, and then offer some amazing firmware upgrade to jumpstart M2 sales again when that happens. Of course, they will probably just release a Mavic 2 Pro 2.0 with the fix baked in, and ask everyone to upgrade. Either way, the M2P produces the most professional looking images I've seen from a consumer drone, so resolution aside, I'm happy. Don't forget that the low light performance makes great use of the sensor as well.

Well, it DOES have much better low-light performance than the MP, but not as good as the P4P. And part of this is because it doesn't use the entire sensor to gather light. If it actually did read the entire sensor and downscale to the 16:9 aspect, we'd have even better -- much -- low-light performance.

The 1" sensor, as it is current utilized, is pretty much a gimmick. We're not really using it for it's superior light gathering capabilities, in any meaningful sense.

sar104 · Oct 6, 2018

dwallersv said:
Well, it DOES have much better low-light performance than the MP, but not as good as the P4P. And part of this is because it doesn't use the entire sensor to gather light. If it actually did read the entire sensor and downscale to the 16:9 aspect, we'd have even better -- much -- low-light performance.

The 1" sensor, as it is current utilized, is pretty much a gimmick. We're not really using it for it's superior light gathering capabilities, in any meaningful sense.

It is being fully used for photography. The full width is also being used for FOV video but, since we don't know exactly what is being read and how it is being processed, it's really hard to say whether all the light captured by the sensor is being used. For example - if it were pixel binning in some way - it would be taking advantage of the sensor in terms of low-light signal-to-noise.

dwallersv · Oct 6, 2018

sar104 said:
It is being fully used for photography. The full width is also being used for FOV video but, since we don't know exactly what is being read and how it is being processed, it's really hard to say whether all the light captured by the sensor is being used. For example - if it were pixel binning in some way - it would be taking advantage of the sensor in terms of low-light signal-to-noise.

All true.

I'm speaking of video, not stills, and while we do not know with certainty what's going on, the forensics have convinced me that what Cliff has deduced is what has actually been engineered.

There may be another explanation, but it's not obvious.

sar104 · Oct 6, 2018

dwallersv said:
All true.

I'm speaking of video, not stills, and while we do not know with certainty what's going on, the forensics have convinced me that what Cliff has deduced is what has actually been engineered.

There may be another explanation, but it's not obvious.

What Cliff posted is real, in that the aliasing exists. His explanation is pure speculation though, and whether the effect is a deliberate degrading of the image by DJI (I find that really unlikely but maybe) or a limitation of the hardware in the smaller M2P package, or an unintended consequence of DJI switching to their own processor chip (if that's true) is impossible to determine.

Lon Denard · Oct 6, 2018

sar104 said:
What Cliff posted is real, in that the aliasing exists. His explanation is pure speculation though, and whether the effect is a deliberate degrading of the image by DJI (I find that really unlikely but maybe) or a limitation of the hardware in the smaller M2P package, or an unintended consequence of DJI switching to their own processor chip (if that's true) is impossible to determine.

I never argued the video image quality angle because I'm not an expert on it but as I understand it, the camera does initially use the 16x9 portion of the sensor in FOV mode and either line skips, pixel bins, sub samples or whatever down to either 4k (says DJI) or 2.7k (says Cliff). In either case isn't the ENTIRE 16x9 portion of the sensor used initially? I mean in HQ mode I recognize that only a portion of the sensor is utilized but in FOV mode I thought that the entire sensor was used at first and that the issue is how the data is processed afterwards.

If that's the case, apart from aperture, lens, focal point, shutter speed, etc, wouldn't the camera capture the same amount of light?

sar104 · Oct 6, 2018

Lon Denard said:
I never argued the video image quality angle because I'm not an expert on it but as I understand it, the camera does initially use the 16x9 portion of the sensor in FOV mode and either line skips, pixel bins, sub samples or whatever down to either 4k (says DJI) or 2.7k (says Cliff). In either case isn't the ENTIRE 16x9 portion of the sensor used initially? I mean in HQ mode I recognize that only a portion of the sensor is utilized but in FOV mode I thought that the entire sensor was used at first and that the issue is how the data is processed afterwards.

Yes - the image in question covers the full width of the sensor. It's obviously not sub-sampling to 2.7k and then upsampling to 4k, but what it is doing is still a mystery. Although Cliff did argue that the 4k and 2.7k resolutions look the same - I'll be testing that this weekend.

Lon Denard · Oct 6, 2018

sar104 said:
Yes - the image in question covers the full width of the sensor. It's obviously not sub-sampling to 2.7k and then upsampling to 4k, but what it is doing is still a mystery. Although Cliff did argue that the 4k and 2.7k resolutions look the same - I'll be testing that this weekend.

Awesome! I look forward to it from a strictly academic standpoint.

Btw I edited my initial reply. The new point he was making concerned light gathering. I'm not sure how what happens in the soc would affect low light capabilities as far as light capture is concerned.

Lon Denard · Oct 6, 2018

dwallersv said:
Well, it DOES have much better low-light performance than the MP, but not as good as the P4P. And part of this is because it doesn't use the entire sensor to gather light. If it actually did read the entire sensor and downscale to the 16:9 aspect, we'd have even better -- much -- low-light performance.

The 1" sensor, as it is current utilized, is pretty much a gimmick. We're not really using it for it's superior light gathering capabilities, in any meaningful sense.

What does the type of sub sampling utilized have to do with how much light is captured? No one has argued that the drone doesn't start with the full 16x9 portion of the sensor initially. The argument is what happens to the data post capture. Line skipping doesn't affect how much light is taken in as far as I know.

c.m.s · Oct 6, 2018

sar104 said:
Yes - the image in question covers the full width of the sensor. It's obviously not sub-sampling to 2.7k and then upsampling to 4k, but what it is doing is still a mystery. Although Cliff did argue that the 4k and 2.7k resolutions look the same - I'll be testing that this weekend.

I bought my Mavic 2 Pro after the "issue" with the 4k FOV was already a hot topic. Since I mainly take still photos I wasn't that concerned about it and after using it for some time now i'm very happy with the camera performance of the Mavic 2 Pro compared to my Mavic Air.

With that said I'm technical nerd so I do like to read about this stuff anyway

So all I wanted to add to this thread really is that I'm looking forward to your test result and appreciate the tests @Cliff_622 has done!

Sockeye · Oct 6, 2018

my 2 cents to ths story:

Lon Denard · Oct 6, 2018

Sockeye said:
my 2 cents to ths story:

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Hey that's interesting. So were all of the distances the same for each camera? Where are the aperture settings on the camera with variable aperture? Why adjust the sharpness? Why not shoot in normal rather than dlog?

I'm honestly curious. I'm happy with HQ and don't plan to shoot in any other way. This is a learning experience for me at this point. I won the only argument I had on this topic.

Lon Denard · Oct 6, 2018

This is a test shot I made to work on color grading. I know it's boring and since I hand fly instead of using autonomous, it's a bit clunky but I like the way the video comes out even with ham handed flying and amateur editing. That's all that counts to me. This pixel counting and processing type debate is interesting but since I'm mostly happy with the results, it doesn't really matter. Of course, I loved hot dogs before I knew what was in them and I still eat them now. All the people that loved the drone before but are outraged now don't make a lot of sense to me.

Category 5 · Oct 6, 2018

Sockeye said:
my 2 cents to ths story:

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

I actually see a little more detail at 4K FOV in your real world example. It certainly looks like detail could be better, but all in all I think the situation is a lot less grim than many in this thread suggest.

yysc · Oct 6, 2018

Sockeye said:
my 2 cents to ths story:

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Thanks for the effort of putting this together. As expected in 4K FOV is terribe in terms of resolved detail. Worse than M1P or a GoPro Hero 5... So those of us who feel cheated in the video department with the M2P have all the right to feel this way.

And DJI says it is only noticiable in charts. BS. We've seen the weird artifacts in 4K FOV in linear patterns that are at medium distance thanks to their crappy sophisticated RAW subsampling.

Lon Denard · Oct 6, 2018

Category 5 said:
I actually see a little more detail at 4K FOV in your real world example. It certainly looks like detail could be better, but all in all I think the situation is a lot less grim than many in this thread suggest.

Yea I've seen the words "horrible" and "disgusting" used to describe the situation. Hahaha! I really don't think it's quite that critical.

Oh add "terrible" and "crappy" to that list!

Lon Denard · Oct 6, 2018

yysc said:
Thanks for the effort of putting this together. As expected in 4K FOV is terribe in terms of resolved detail. Worse than M1P or a GoPro Hero 5... So those of us who feel cheated in the video department with the M2P have all the right to feel this way.

And DJI says it is only noticiable in charts. BS. We've seen the weird artifacts in 4K FOV in linear patterns that are at medium distance thanks to their crappy sophisticated RAW subsampling.

Real talk. Why didn't you just send it back if you're this unhappy? You had to have noticed this "terrible" and "crappy" video quality before Cliff told you it was bad. Why did you keep it? Honestly, I don't understand. It's like working with a woman who's husband gets drunk and beats her every weekend. Every Monday there's a new bruise and story... I wonder if we should start a battered drone pilots home or something?

yysc · Oct 6, 2018

Category 5 said:
I actually see a little more detail at 4K FOV in your real world example. It certainly looks like detail could be better, but all in all I think the situation is a lot less grim than many in this thread suggest.

Well considering 2.7K is roughly 4MP and 4K is 8MP. Having double the information should provide more than a very slight detail increase over 2.7K.

I took a picture of my M2P and supersampled to 4K (cropping to 16:9) and also supersampled to 2.7K. Both bicubic. Then I increased the zoom of 2.7K by 150% to match 4K size and similar to Sockeye comparision.

The difference is obvious and this is what we should see, and not the very light detail increase from 2.7K 4MP to 4K 8MP.