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

[yysc]

How are they going to give a full readout and 10-bit and stay under 100Mbps?

That's not some arbitrary limit, you have to have media that can accept all that information and write it in real time.

The Phantom 4 Pro and the SONY RX100 both do a full 1" sensor readout and both supersample (not subsample) that down to 4K 100Mbps.

The video resize, plus correction, filtering etc, is done in the SOC prior for it to export the final video stream at 100Mbps. Unless the SOC used in the M2P is very weak there should be no issue following the same approach, as even the $300 Yi 4K+ action cam uses the Ambarella H2 which is far superior to the H1 used in the P4P.

 

[Category 5]

Hasselblad was a complete disaster before DJI bought them. They were owned and by an investment bankers group. They had laid off employees and were down to a small stadf. They were rebranding 100% Sony made cameras by bolting on wood grips and sticking goofy looking dials to them and trying to charge 5x the Sony selling price. They actually tried to say that their wood grips made these cameras "exotic". Hassy was THE joke of the camera industry. And, when they hit rock bottom, they pratically tried to give all their camera production stock away because nobody wanted them.

Just Google their history. Hassy has never been a true electronics company. They are just a camera "fashion" brand.

Today, it doesnt take ANY camera manufacturing experience to build a camera! My grandmother could do it. She goes to Sony or several other companies to design and manufacture everything. The hardest thing my grandmother needs to do is come up with a logo that she likes. After that, those companies brand it with her logo and BANG....she now has her own "GrandMother-Blad" camera company!

Im serious. "You" can easily have your own camera designed and manufactured completely by third party contracts today. You only need the capital to put down to get it started.

Hasselblad is just a logo that DJI now ownes.

They have never been a consumer oriented manufacturer. Their medium format digital hardware has been, and continues to be the best you can get for image quality. Not sure where you got the data about Sony re-brands at 5 times the price, but any foray into the consumer market would have been an experiment at best. Sony is doing AMAZING things in digital imaging, but Hasselblad still holds the top spot according to DXO.

 

[sar104]

It depends in what you shoot. There are other side effects as more aliasing in the M2P at 4K FOV. If you shoot mostly nature then the effect will be minimized. If you shoot scenes with straight lines (buildings, urban) the subsampling they use in 4K creates quite some aliasing/moire in objects around 300-500m from the drone. You can see it in the video below.

I've watched it my 4K TV in 4K, and the M2P in 4K FOV has far more aliasing/moire than the Autel Evo. In a city landscape it becomes a bit annoying.

What on earth did he do to the M2P output? The colors are far too saturated to make any reasonable comparison.

 

[sar104]

The Phantom 4 Pro and the SONY RX100 both do a full 1" sensor readout and both supersample (not subsample) that down to 4K 100Mbps.

The video resize, plus correction, filtering etc, is done in the SOC prior for it to export the final video stream at 100Mbps. Unless the SOC used in the M2P is very weak there should be no issue following the same approach, as even the $300 Yi 4K+ action cam uses the Ambarella H2 which is far superior to the H1 used in the P4P.

Yes - this seems to be the central issue. The FOV resolution is actually quite good, but has some spatial frequency artifacts that don't seem to follow the usual, expected aliasing behavior, and implies to me that they are doing something other than the regular full readout and downsample to 4k.

 

[yysc]

What on earth did he do to the M2P output? The colors are far too saturated to make any reasonable comparison.

Yes colors are ugly but the point is aliasing/moire. I have also seen similar aliasing in my own M2P videos shot in 4K FOV.

 

[sar104]

Yes colors are ugly but the point is aliasing/moire. I have also seen similar aliasing in my own M2P videos shot in 4K FOV.

The original complaint in this thread was that the resolution is poor. Aliasing gets worse as optical resolution increases near the resolution limit. I've seen aliasing on all platforms, including the RX100. The question is whether it is actually worse on the M2P than comparable cameras and, if so, with which object spatial frequencies since it doesn't seem to be monotonic. Cherry-picking random videos is not a good way to do a controlled comparison. Why did you link to a poorly processed YouTube video rather than your own videos, or full-resolution 4k frame grabs, if you are seeing a significant problem vs. other cameras?

 

[yysc]

The original complaint in this thread was that the resolution is poor. Aliasing gets worse as optical resolution increases near the resolution limit. I've seen aliasing on all platforms, including the RX100. The question is whether it is actually worse on the M2P than comparable cameras and, if so, with which object spatial frequencies since it doesn't seem to be monotonic. Cherry-picking random videos is not a good way to do a controlled comparison. Why did you link to a poorly processed YouTube video rather than your own videos, or full-resolution 4k frame grabs, if you are seeing a significant problem vs. other cameras?

I used that video as it demonstrated the issue. My YT account is for private videos, but since you asked, below 200% crops from my M2P. All stock settings. Better see them at 100% size.

4K FOV (note continuity issues in road lines, guard rails. This in motion is noticiable when play back at 100% size)

4K HQ (since this is 1:1 readout of the 20MP sensor with the AA filter being optimized for 20MP aliasing is really low).

While the perspective is a bit different both roads were at a similar distance (size of pictured cars is similar).

 

[BlueBay]

What on earth did he do to the M2P output? The colors are far too saturated to make any reasonable comparison.

I agree, the M2P colours are far too saturated in the daylight scenes to make a comparison. The sunset footage is better.

 

[BlueBay]

I used that video as it demonstrated the issue. My YT account is for private videos, but since you asked, below 200% crops from my M2P. All stock settings. Better see them at 100% size.

4K FOV (note continuity issues in road lines, guard rails. This in motion is noticiable when play back at 100% size)

4K HQ (since this is 1:1 readout of the 20MP sensor with the AA filter being optimized for 20MP aliasing is really low).

While the perspective is a bit different both roads were at a similar distance (size of pictured cars is similar).

Wow! Thank you, that's very interesting.

 

[Trees138]

The Phantom 4 Pro and the SONY RX100 both do a full 1" sensor readout and both supersample (not subsample) that down to 4K 100Mbps.

The video resize, plus correction, filtering etc, is done in the SOC prior for it to export the final video stream at 100Mbps. Unless the SOC used in the M2P is very weak there should be no issue following the same approach, as even the $300 Yi 4K+ action cam uses the Ambarella H2 which is far superior to the H1 used in the P4P.

Oh, my mistake, I would have assumed 10-bit would have to be factored in bitrate, boy do I feel silly.

 

[CanadaDrone]

They have never been a consumer oriented manufacturer. Their medium format digital hardware has been, and continues to be the best you can get for image quality. Not sure where you got the data about Sony re-brands at 5 times the price, but any foray into the consumer market would have been an experiment at best. Sony is doing AMAZING things in digital imaging, but Hasselblad still holds the top spot according to DXO.

Hasselblad has been caught many times over the years re-branding Sony cameras, including the Sony NEX-7, RX100, MK 1 & MK 2, A99 and A7R with hilarious price premiums. Leica has played the same games with re-branded Panasonic Lumix cameras. They weren't high volume but neither is anything Hasselblad makes.

Regarding image quality, there aren't really any unique sensors of note in the MF world. Sony pretty much makes them all. In the 50MP cameras, the Pentax 645Z, Hasselblad 50C, and Phase One IQ 350 all use the same sensor. The 100mp Sony IMX211 sensor is currently used in the Hasselblad H6D 100C and and Phase One IQ3 100. It's all pretty much the same - any rational purchase decision would come down to lenses and the system as a whole.

 

[CanadaDrone]

I am still planning on buying a MP2 but I'd be lying if I said it didn't bother me how bad the video IQ is (relative to other 1" sensor units). It's better than the 1/2.3" stuff but in good light the difference isn't nearly as big as it should be. Still images are another story, and certainly help to bolster the value proposition. All the excitement I had about DJI Putting a 1"sensor in a Mavic body is pretty much gone though when you look at the night and day difference between it and the P4P. Understandably, I'd guess there were size and most likely heat considerations that ultimately forced compromises in the worst possible areas (image quality).

Maybe the next Phantom will get folding legs and we can (sort of) get the best of both worlds

 

[igor88]

My YT account is for private videos...

Mine is not.
Disregard p4p picture and whole comparison.
Just watch aliasing. Check the question mark at 0:42 min.

 

[WithTheBirds]

I agree, better than the MPP, but M2P has a 1 inch Sensor. We are mostly comparing the 4k resolution with the P4P which also has a 1 inch sensor, but the M2P has less 4k resolution (than the P4P) as it's not doing a full readout (of the sensor).

Therein lies the issue! Whether it's hopping, skipping, cooling, binning or evaporating(!) - please give us the full readout so that we can properly feed our OLEDs!

You want the full readout? Pony up $50K+, that’s the admission price for what you would hope for. This is more a case of unrealistic expectations of some users rather than an DJI failing to deliver.

 

[yysc]

Mine is not.
Disregard p4p picture and whole comparison.
Just watch aliasing. Check the question mark at 0:42 min.

Thanks this shows the same issue in 4K FOV as my crops. I guess DJI will now say the camera was optimized for trees and not roads.

 

[sar104]

I am still planning on buying a MP2 but I'd be lying if I said it didn't bother me how bad the video IQ is (relative to other 1" sensor units). It's better than the 1/2.3" stuff but in good light the difference isn't nearly as big as it should be. Still images are another story, and certainly help to bolster the value proposition. All the excitement I had about DJI Putting a 1"sensor in a Mavic body is pretty much gone though when you look at the night and day difference between it and the P4P. Understandably, I'd guess there were size and most likely heat considerations that ultimately forced compromises in the worst possible areas (image quality).

Maybe the next Phantom will get folding legs and we can (sort of) get the best of both worlds

I'm still really looking forward to someone posting actual examples of these "night and day" differences.

 

[yysc]

You want the full readout? Pony up $50K+, that’s the admission price for what you would hope for. This is more a case of unrealistic expectations of some users rather than an DJI failing to deliver.

The P4P seems to do supersample from full sensor readout and also doesn't have the aliasing issues shown in this very same page of this thread. And the admission price is the same ($1499).

 

[kilomikebravo]

I'm still really looking forward to someone posting actual examples of these "night and day" differences.

Can I get a big AMEN?!

 

[CanadaDrone]

I'm still really looking forward to someone posting actual examples of these "night and day" differences.

Lots of comparisons on YouTube if you want to check those out. Just knowing the way the MP2 reads/processes video compared to how the P4P, you don't even need to see comparisons to know one will be much better than the other, but if you want side-by-side comparisons they are on YouTube or you can download sample footage from some willing to share it and the difference is quite substantial. FOV mode is actually quite awful, HQ mode is better but still well behind the P4P. It is an improvement on the original MP, but if you've owned or processed P4P footage, the difference really stands out. That exact same sensor found in both a Sony RX100 series camera and the P4P produces a much better video file than it does in the MP2. Like I said, I believe they had compromises to make for both size and heat, and unfortunately that means a big hit to image quality (reading a sensor quickly makes a lot of heat). Stills are unaffected.

Can I get a big AMEN?!

Again, they are very easy to find - literally the first google hits. If you aren't finding them you probably haven't looked. If video IQ isn't a big deal to you or simply being better than the original MP is good enough, then it's a non-issue. Those used to P4P or even RX100 footage are likely going to be immediately disappointed as it will stand out like a sore thumb (I am picky about image quality and to me it was extremely obvious). Those coming from an original MP are likely to be quite pleased, especially since the MP footage was overall quite average.

I also think it's an issue with expectations. Many people, including myself, were hoping for P4P image quality from a Mavic. As soon as the leaks surfaced with the 1" sensor, a lot of people got excited about that being a reality. Unfortunately, that is not the case, but it's understandable given the size/heat constraints between the different designs.

 

[sar104]

Lots of comparisons on YouTube if you want to check those out. Just knowing the way the MP2 reads/processes video compared to how the P4P, you don't even need to see comparisons to know one will be much better than the other, but if you want side-by-side comparisons they are on YouTube or you can download sample footage from some willing to share it and the difference is quite substantial. FOV mode is actually quite awful, HQ mode is better but still well behind the P4P. It is an improvement on the original MP, but if you've owned or processed P4P footage, the difference really stands out. That exact same sensor found in both a Sony RX100 series camera and the P4P produces a much better video file than it does in the MP2. Like I said, I believe they had compromises to make for both size and heat, and unfortunately that means a big hit to image quality (reading a sensor quickly makes a lot of heat). Stills are unaffected.

"Just knowing the way the MP2 reads/processes video compared to how the P4P, you don't even need to see comparisons to know one will be much better than the other...."​

​

And there is the single biggest problem with this discussion, in a nutshell. Firstly, you don't know how the M2P reads and processes - you are just guessing based on speculation, some of which is clearly wrong. Secondly, all the stuff I've seen on YouTube is as worthless as the last YT example posted on this thread - mostly uncontrolled, poorly processed video, or examined at ludicrous magnifications. Thirdly, not being interested in actually looking at actual video makes the entire argument pointless.

I've spent way too long now looking at video and frame grab samples, and now doing actual resolution tests. For what it's worth, which is likely not much, the method apparently used on the P4P (20 MP readout to the image processor followed by some form of processing down to 4k) seems to produce cleaner results in terms of aliasing than the M2P in FOV Dlog-M (some kind of downsampling apparently on the sensor itself) at certain spatial frequencies. As spatial frequency increases towards the resolution limit, the full-readout approach seems to give the expected gradual transition from clearly resolved through classical moiré to unresolved, while the M2P goes from well-resolved to moiré to resolved again before reaching unresolved. That may be responsible for the less pleasing appearance of aliasing. It will be interesting to find out whether that is constrained by hardware or whether firmware improvements are possible.