Mavic 2 Pro 4K60 what are the odds?

[mcsenerd]

I’m sure the move to the likely cheaper per unit All-loser SOC has nothing to do with anything...

 

[m0j0]

Go for it. But remember watching stuff on YouTube is not going to much good. You need original comparable footage from each of the drones. Gets complicated and expensive.

Have you compared comparable footage of the P4P between 24, 30 and 60 FPS? I suggest to start there.

I have done the research. MP2 is in a different class than the EVO. Video is fantastic. The only problem I have seen is the barrel distortion in D-LOG. The 10 bit video is amazing. I worked with some RAW data. Holy crap it is good. It's a pro tool. Most of the comparisons on the MP2 don't use corrected D-log video because the other birds don't have it. Well the 10bit color in D-LOG is the difference maker. EVO as I see it is about the same as my current Mavic. Buying it I would not see an appreciable difference in quality they are virtually the same.

So sir - I will concede that the MP2 video surpasses the EVO add to that DJI's 3rd party support and the choice is clear. Looks like I'm holding off awhile for a new bird but when I do get one it will be DJI P5p or Mp2.

 

[NorthGADronePilot]

Seems like a bad decision to exclude it. The other foldable has it. Why would you leave it out solely to protect the P4?

Why would DJI continue to build upon the Phantom line AT ALL now that the Mavic is firmly entrenched and awesome? It makes zero sense to me.

 

[Raymondo1]

I’m getting the M2Pro soon and I think that once the Phantom 5 has been put on sale the M2P will receive image/video quality upgrades. As it stands if the video quality on the M2P was better than the P4 they wouldn’t be able to sell what they have left. Just my opinion.

 

[Cliff_622]

What older foldable would that be? The Mavic Air and the Mavic 2 Zoom do not have 4K60P. The only DJI "consumer level" drone that has it is the Phantom 4 Pro / 4 Pro 2.0 without making the leap into the Inspire line.

If you believe DJI it is primarily a hardware limitation, but if the Mavic 2 Pro had an identical camera to the Phantom 4 Pro with 4K60P and a mechanical shutter, I would struggle to think of one reason to buy the Phantom. At least in the current state of things, then Phantom 4 Pro remains the best image quality you can get in the general category so there is some value there.

The Autel Evo can shoot 4K60P but not with a 1" sensor (it is much, much, easier to get faster readout from a sensor ~4X smaller), so it can't be compared to the M2P, and to my knowledge it is riddled with problems.

The "physical" size of a CMOS sensor doesnt have anything to do with readout speed (directly) A full frame 20mp sensor and a 1inch or smaller 20mp sensor produce the exact same raw sensor data amount. So image assembly and processing ability is identical across all of them.

The sensor size DOES affect "heat". All sensors get hot as you read them. The larger the sensor, the more the heat. To make matters worse, the faster you scan them, the hotter they get. (30p vs 60p)

A large sensor can scan just as fast and as a small one....provided it has the larger cooling mechanisms in place to protect it. The image processor? It doesnt care. 20mp of raw sensor data from a tiny sensor is exactly the same amount as 20mp from a pizza box sized 20mp sensor.

Its all about controlling heat.

 

[CanadaDrone]

The "physical" size of a CMOS sensor doesnt have anything to do with readout speed (directly) A full frame 20mp sensor and a 1inch or smaller 20mp sensor produce the exact same raw sensor data amount. So image assembly and processing ability is identical across all of them.

The sensor size DOES affect "heat". All sensors get hot as you read them. The larger the sensor, the more the heat. To make matters worse, the faster you scan them, the hotter they get. (30p vs 60p)

A large sensor can scan just as fast and as a small one....provided it has the larger cooling mechanisms in place to protect it. The image processor? It doesnt care. 20mp of raw sensor data from a tiny sensor is exactly the same amount as 20mp from a pizza box sized 20mp sensor.

Its all about controlling heat.

Exactly, which is why BSI and Stacked designs generally favor the smaller sensors and have diminishing returns as you get to the larger sensors with more wafer real estate.

 

[Cosmo SIX5]

Protecting the p4 is nonsense. It's been oos since I began searching for a drone in August.

 

[bkushner]

Would have happened by now.

 

[VooDooZG]

I bet M2P Platinum will get 4k60 - I'am waiting for it

 

[Cliff_622]

At least they have the common sense of comparing two cameras that have much in common and I am not impressed by the EVO's video quality looking at it here and the EVO zoom is a joke.

But to quickly come back to my original response to the 4K/60fps issue:

A 4K video frame as an 8-bit (YUV) RAW image produces 199,065,600 image information bits (roughly 200Mb/FRAME or just shy of 25 Megabytes of data. That is just ONE FRAME. If you record at 30 frames per second at 100Mb/s bitrate, you allot about 3.3 Megabits per frame on average. That is a whopping 1:60 million reduction. Sure mpg (especially HEVC - H265) compression is smart about motion vectors and understands how to recognize parts of an image that changes versus others and how the changes occur hence is able to perform motion prediction. This allows the compression algorithms to throw away parts of an image recording just deltas. Every some number of frames (GOP), it records the whole frame (I-Frame) and then starts over with the prediction producing only partial frames with progressively degrading quality and detail. (often seen on low quality compression as "GOP PULSING" - a very ugly artifact nearly impossible to correct). Even the I-frame is heavily compressed where the algorithm has to arbitrate fine details and small contrast changes resulting in what sometimes is referred to as "mush" or "water color" (very visible in distant trees and mountains). On the M1 I tried to compensate by pushing both sharpness and contrast during the recording phase (Style 2,1,0) which only works to a degree.

Not to go any further into obscure details about compression suffice it to know that there are numerous different encoder implementations and all of them have numerous parameters to guide internal algorithms favoring one thing over another, one compromise after another etc. What makes matters worse is the COMPUTATIONAL limitations of the silicon in the camera system (not the sensor but the image processor). Higher quality compression requires an exponential increase in processing power and there is simply a limit to that in this tiny system no matter what.

Assuming for a moment that the image processor CAN do an equally good job at twice the speed doubling the amount of data that ultimately HAS to be discarded will inevitably degrade the resulting image quality. So, if it is already very hard to encode a high quality 30fps 4K stream making many compromises already (clearly visible when looking at the exact same size RAW (stil) image) - why would you want to force a 60fps equivalent through this funnel - would you not likely get a much better result recording 2.7K at 60fps or even 2K/60fps under these constraints?

So regardless on how good the camera hardware (lenses, support/gimbal etc) and the underlying sensor (which only needs to have 8.3 Megapixels for 4K (UHD) video), the most deterministic component (for video) is the image processor and the encoder it is running. If you want to compare cameras, it is best to compare RAW stills from each of the same subject since you pretty much exclude the image processor and look at image capture parts only. Once you have that base line, then look at footage to compare the image processing part in context of the established baseline.

Part of this water color "pulsing" effect is NOT just interframe (long GOP) compression. A BIG part of the problem is heavy handed temporal noise reduction.

Noise reduction scrubs and strangles the very life out of an image. It literally SMEARS hi detail elements like grass and bushes and turns them into mud. The more you add the worse it gets.....and DJI ADDS A TON OF IT!!

Temporal noise reduction uses multiple frams to analyze and identify noise. It looks for the same noise patterns across several frames and the tries to phase cancel it out.

In the older Ambarella A9 chip we saw the horrible "pulsing" and water color smearing becase we were seeing this multi frame noise reduction resetting is scrubbing process every certain number of frames.

Originally, many of us blamed the low h.264 bit rate....this was NOT the case. We then realized that when we added +1 sharpening, the noise reduction was PULLED BACK and this reduced the negative artifacts.....while still using the same low h.264 bit rate.

4k 60p?.....Hahaha! We cant even get full pixel readout in 30p!! We are either getting pixel binned garbage in FOV mode...or getting OK 1:1 readout in a CROPPED (40mm) HQ mode.

That says A LOT about the M2P's sensor and image processor combo!

Cliff Totten

 

[CanadaDrone]

4k 60p?.....Hahaha! We cant even get full pixel readout in 30p!! We are either getting pixel binned garbage in FOV mode...or getting OK 1:1 readout in a CROPPED (40mm) HQ mode.

Do you have something that shows DJI is pixel binning rather than subsampling to get 4K in FOV mode? If so I would be genuinely interested in seeing it as that would go against their official explanation.

 

[CanadaDrone]

Protecting the p4 is nonsense. It's been oos since I began searching for a drone in August.

I don't know where you're located but I see the P4 Pro 2.0 in stock at major stores like B&H.

 

[Cliff_622]

Do you have something that shows DJI is pixel binning rather than subsampling to get 4K in FOV mode? If so I would be genuinely interested in seeing it as that would go against their official explanation.

Pixel binning is a form of sub sampling. Its taking RGGB clusters and subsampling them...by....averaging those clusters DOWN to one single value. So yes, DJI is "subsampling" by binning values. Is it "reading" every pixel? Possibly yes!...thats the INPUT. The OUTPUT that goes to video is cluster averaged. (Binned) This is why its so badly soft in FOV mode. FOV is BARELY higher than 1080 in actual resolving power. Its certainly nowwhere close to "4k"......

This is why DJI made an "HQ" mode. But "HQ" only uses a center portion of the 1inch-type sensor. So, HQ does not use the "real" sensor surface area.

Im sorry but FOV is just "soup" and "mud" to me and I wont ever use it.

 

[CanadaDrone]

Pixel binning is a form of sub sampling. Its taking RGGB clusters and subsampling them...by....averaging those clusters DOWN to one single value. So yes, DJI is "subsampling" by binning values. Is it "reading" every pixel? Possibly yes!...thats the INPUT. The OUTPUT that goes to video is cluster averaged. (Binned) This is why its so badly soft in FOV mode. FOV is BARELY higher than 1080 in actual resolving power. Its certainly nowwhere close to "4k"......

This is why DJI made an "HQ" mode. But "HQ" only uses a center portion of the 1inch-type sensor. So, HQ does not use the "real" sensor surface area.

Im sorry but FOV is just "soup" and "mud" to me and I wont ever use it.

OK sounds like we're talking about the same thing then

 

[wco81]

I bet M2P Platinum will get 4k60 - I'am waiting for it

Is that likely this year?

If they're planning to release this year, you'd think in a couple of months, you'd hear about Mavic 2 inventories going down or even price promotions or outright cuts.

I don't know how frequently this Ambarella processor iterates. Do they come out with more powerful chips every year? Do they use smaller-die processes like they do for smart phone chips, which continually drive double-digit performance increases or better battery life?

For one thing, it sounds like they still use H.264, not H.265? But H.265 codecs may require more processing power as well as higher royalties. There is now a royalty-free codec, AV1, which is going to be used more and more but that's really new so there probably aren't optimized encoders for it and the processors used in drones may not have the horsepower for it yet.

 

[Kilrah]

DJI isn't using Ambarella chips anymore, the M2 uses an Allwinner processr for the camera.
I believe Ambarella have pretty restrictive / aggressive commercial terms that DJI probably didn't want to deal with anymore, so they went elsewhere and thus had to take a small step back on features.
M2 does both H.264 and H.265 already, so not related.

That lack of 60fps could also very well have nothing to do with actual hardware capabilities, just marketing and "leaving something to improve for the next generation to encourage people to upgrade" since the list of possible improvements in this form factor is becoming pretty short.

 

[brett8883]

Why would DJI continue to build upon the Phantom line AT ALL now that the Mavic is firmly entrenched and awesome? It makes zero sense to me.

The Phantom 4 V2 has a mechanical shutter and it’s heavier weight allows it to fly better in wind.

I get the 10-bit Log thing but in real world side by side my experience the Phantom actually produces better quality video then Mavic 2 Pro. We can talk specs all day but what I’m talking about is actual quality. In low light the phantom blows the Mavic 2 Pro out of the water.

The wind thing I big for me. Even with a 10 MPH wind the Mavic has a real hard time panning smoothly.

 

[brett8883]

Do you have something that shows DJI is pixel binning rather than subsampling to get 4K in FOV mode? If so I would be genuinely interested in seeing it as that would go against their official explanation.

I think this is right up your alley. Personally I don’t care that much but he makes a compelling argument.

 

[CanadaDrone]

I think this is right up your alley. Personally I don’t care that much but he makes a compelling argument.

Thanks - I have seen that but I don't get how it shows a specific difference between pixel binning and subsampling, which are similar but 2 different things. The P4P without a doubt puts out better video than the M2P, I just wish it folded up just as small

 

[brett8883]

Thanks - I have seen that but I don't get how it shows a specific difference between pixel binning and subsampling, which are similar but 2 different things. The P4P without a doubt puts out better video than the M2P, I just wish it folded up just as small

Me too