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

[Dronefriend]

Thanks for good discussion guys. Could you please explain if recording in 2.7K FOV mode on M2P is equal to 2.7K record on P4P?

Or does it differ because of binning vs downsampling? If so, how large is the difference would you think?

Reason I am asking is I am fine with 2.7K on P4P. And would only use the M2P for private shots, not professional.

I should just take it out and fly but Im so lazy..

 

[chunky]

Thanks for good discussion guys. Could you please explain if recording in 2.7K FOV mode on M2P is equal to 2.7K record on P4P?

Or does it differ because of binning vs downsampling? If so, how large is the difference would you think?

Reason I am asking is I am fine with 2.7K on P4P. And would only use the M2P for private shots, not professional.

I should just take it out and fly but Im so lazy..

I could inspect this with reports the next week since I‘m on vacation without my P4pro.
I also never used 2.7K before.

But we also may have other users with P4pro here.

 

[Sockeye]

Ooops, Italy has woke up....

 

[Glomby]

So they are basically never properly down sampling (down scaling), it's always some kind of line skipping or pixel binning and therefor an artefact introducing compromise. I don't see how this makes sense with such a sensor and the processing power on board. I really hope they can fix a lot of this.

 

[arichmondphoto]

Just curious, but is it at all possible that the artifacts you all are arguing about has more to do with compression than what the sensor is capable of?

I mean, wouldn't the best way to test a sensor be to take a raw readout from the sensor directly? Once it's compressed you aren't seeing the full picture (no pun intended). Given that H.264 and H.265 are highly compressed codecs, it seems reasonable that there would be data loss on that end of the pipeline. This is why high-end production equipment records in codecs like ProRes, CinemaDNG - no/less compression...

 

[chunky]

No, P4pro has a weaker encoder on board but delivers higher resolution.

 

[Sockeye]

These bad things happen before the data compression by the SOC. Artefacts coming from the .264 & h.264 codec look completely different.
The sensor itself can do much more... its the way DJI reads out the information from the sensor, basically ignoring every second pixel for the sake of who knows what...

 

[chunky]

I hope DJI knows what to do now since their specification is obvious!

 

[Sockeye]

I hope DJI knows what to do now since their specification is obvious!

Yes, for sure. They closed the specific threads in the DJI forum....

 

[chunky]

Yes, for sure. They closed the specific threads in the DJI forum....

Specs and FAQs still speak about „resolution“ not video format or image size.
I‘m wondering when this will be changed !?

Gentlemen, I hope you all made screenshots of the specs and faq sites as also of the version 1.0 pdf manual !!


...and I thought the problems with their Ronin-S were their major problems...

 

[chunky]

Hmm, I don‘t see the threads closed:
Mavic 2 Pro Sensor Issues

I also don‘t see changes in the new v1.2 manual, still 4K videoresolution:
https://dl.djicdn.com/downloads/Mavic_2/20180920/Mavic 2 Pro Zoom User Manual_v1.2.pdf

They must be working on it, in a few days we‘ll have a new FW and oversampling back in FFOV - at least I hope so ;-)!

 

[mcsenerd]

To me, this is SOP for DJI. I don't known anyone in their right mind that thinks of them as post-sales friendly company... In any case, I'm not going to throw the baby out with the bath water here, and I'm going to give them some time to try to correct some of this (assuming it can be "corrected").

 

[Sockeye]

Hmm, I don‘t see the threads closed:
Mavic 2 Pro Sensor Issues

Good, that on is alive... I was referring to the other one "When will DJI fix the Pro camera?" (Originally started on the lens distortion, but then drifted to FOV)
It made it to the #1 "hot topics" on the starting page...

 

[Cliff_622]

Why? Seriously.... call it whatever you like. I already posted the table you refer to when I first made the point. Binning is combining photo sites on the sensor to create a larger “bin”.... effectively the summing is done in the analogue domain prior to AD conversion. You get at a true increase in SNR. I’m sorry you feel you have to keep labouring this point... I’m happy to say every technique that arrives at a lower resolution without changing FOV is binning...

Have a good look at the table and you will see there are only two modes where horizontal and/or vertical binning isn’t performed. It might be a clue in solving this mystery. That was my intent in originally posting it although I almost wish I hadn’t.

While your at it another observation might be the white paper says nothing about line skipping.

I suspect we will ultimately find sensor heating issues are not the issue here.

Sensors only do one thing. They collect photons and produce voltages. A 20 mp sensor will produce 20 million voltage readings per read cycle. They dont bin themselves at all. It's the "image processor" that collects those raw voltage readings and binns those values during the raw data deBayer and image assembly process. The image processor can address columns and rows on the grid as it wants too. The image processor can read photosites in many different patterns. It can scan some and leave others untouched. The binning process is entirely on the image processor's programming. (Possibly Ambarella H2, H22 or H3 SOC)

Look,...in my video, (yes, I'm Cliff Totten) I state that I think the FOV resolution readout problem is because of heat. Now,...only 5 or 10 people on planet Earth (DJI and Ambarella) actually understand the readout mathematics of the Mavic-2 Pro in FOV mode and they will NEVER tell us why they did it. If it was NOT done for marketing protection of the Phantom 4 Pro v2.0 or future Phantom 5 then it was ONLY done to solve an actual "problem". There is no other explanation why DJI chose,....chose...CHOSE to do it like this. Because it was NECESSARY to solve a "problem". (if it was not a deliberate a marketing trick)

Why else would DJI do this? Anybody care to guess?

FOV wide view "pixel binning" or "line skipping" or other "sub-sampling" stunt is there for a reason. DJI will never tell us what their reason was because it's too embarrassing either way!

CT

 

[WithTheBirds]

Sensors only do one thing. They collect photons and produce voltages. A 20 mp sensor will produce 20 million voltage readings per read cycle. They dont bin themselves at all. It's the "image processor" that collects those raw voltage readings and binns those values during the raw data deBayer and image assembly process. The image processor can address columns and rows on the grid as it wants too. The image processor can read photosites in many different patterns. It can scan some and leave others untouched. The binning process is entirely on the image processor's programming. (Possibly Ambarella H2, H22 or H3 SOC)

Look,...in my video, (yes, I'm Cliff Totten) I state that I think the FOV resolution readout problem is because of heat. Now,...only 5 or 10 people on planet Earth (DJI and Ambarella) actually understand the readout mathematics of the Mavic-2 Pro in FOV mode and they will NEVER tell us why they did it. If it was NOT done for marketing protection of the Phantom 4 Pro v2.0 or future Phantom 5 then it was ONLY done to solve an actual "problem". There is no other explanation why DJI chose,....chose...CHOSE to do it like this. Because it was NECESSARY to solve a "problem". (if it was not a deliberate a marketing trick)

Why else would DJI do this? Anybody care to guess?

FOV wide view "pixel binning" or "line skipping" or other "sub-sampling" stunt is there for a reason. DJI will never tell us what their reason was because it's too embarrassing either way!

CT

Your right- CMOS sensors can’t do binning on the sensor- CCD’s absolutely can and do. By altering the drive clock and readout timing photosites can be combined effectively creating larger “bins” to collect photons with the individual charges being added in the analog voltage domain prior to AD conversion. It wasn’t long ago CCD pulled the pants down on CMOS in performance. Back illuminated CMOS has significantly closed the gap (and they are cheaper to make).

It seems the EXMOR performs line addition on the chip. To the extent your assumptions are correct it is some nasty processing in the SOC absolutely. We should feel cheated and I agree DJI will never admit to anything- they might blame a third party though.

 

[star4870]

Sensors only do one thing. They collect photons and produce voltages. A 20 mp sensor will produce 20 million voltage readings per read cycle. They dont bin themselves at all. It's the "image processor" that collects those raw voltage readings and binns those values during the raw data deBayer and image assembly process. The image processor can address columns and rows on the grid as it wants too. The image processor can read photosites in many different patterns. It can scan some and leave others untouched. The binning process is entirely on the image processor's programming. (Possibly Ambarella H2, H22 or H3 SOC)

Look,...in my video, (yes, I'm Cliff Totten) I state that I think the FOV resolution readout problem is because of heat. Now,...only 5 or 10 people on planet Earth (DJI and Ambarella) actually understand the readout mathematics of the Mavic-2 Pro in FOV mode and they will NEVER tell us why they did it. If it was NOT done for marketing protection of the Phantom 4 Pro v2.0 or future Phantom 5 then it was ONLY done to solve an actual "problem". There is no other explanation why DJI chose,....chose...CHOSE to do it like this. Because it was NECESSARY to solve a "problem". (if it was not a deliberate a marketing trick)

Why else would DJI do this? Anybody care to guess?

FOV wide view "pixel binning" or "line skipping" or other "sub-sampling" stunt is there for a reason. DJI will never tell us what their reason was because it's too embarrassing either way!

CT

I don't believe in overheating, I had Phantom 4 pro and was opening gimbal, there was a lots of unused space inside and 2 MCB board one for gimbal motor control, second for the camera. No heat sink, no adhesive to sides of the case, nothing.
They could easily use original design, reduced it and fit into new gimbal, if they wanted.
There is another reason why they did it. Like completely new image procesor with new hasselblad color space, requiring new algorithm to process the image and they still need to polish it.
Soon we will know the true. As this become more and more issue they will have to address it. How they approach it will give us hint what was the real reason.

 

[WithTheBirds]

I don't believe in overheating, I had Phantom 4 pro and was opening gimbal, there was a lots of unused space inside and 2 MCB board one for gimbal motor control, second for the camera. No heat sink, no adhesive to sides of the case, nothing.
They could easily use original design, reduced it and fit into new gimbal, if they wanted.
There is another reason why they did it. Like completely new image procesor with new hasselblad color space, requiring new algorithm to process the image and they still need to polish it.
Soon we will know the true. As this become more and more issue they will have to address it. How they approach it will give us hint what was the real reason.

I’m with you- the overheating speculation probably is a no issue. No way DJI/Hasselblad would have intended it to be like this and we will see a resolution or good explanation. The stills out of this thing are ridiculously good for the price and other features.

 

[Cliff_622]

I was reading a stock market article on Ambarella. It stated that GoPro has left Ambarella and that DJI is parting ways with them in favor of developing their own custom SOC image processors.

We all are expecting the Mavic-2's to have the Ambarella H2, H22 or H3 image processors. Is it possible that DJI is NOT using Ambarella here? Could DJI have attempted to design their own chip and this is why we are seeing this?

I dunno,...would LOVE to see Mavic-2 tear-down and motherboard photo with the metal covers removed from that image processor. We know what off the shelf Ambarella image processors are capable of. If DJI has developed their own, than we will never know.

CT

Addition:
Q2 2019 Ambarella Earnings Conference Call:
Aug. 30, 2018 4:30 p.m. ET - This is a portion from their stock investor earnings call......

"And as you can see from our weak Q3 guidance, weeks by revenue from these consumer applications to continue to be under pressure, reflecting lower demand for non-GoPro sports camera, the VR market remaining nascent, and the drone market leader, DJI, mostly using its own silicon solutions. "

Woah!!!!!,...could the Mavic-2 be using a NEW DJI-owned and designed image processor? If so,...DJI is REALLY screwing it up! LOL

Hmmm,....we know Qualcom is out there for sale too.

CT

 

[chunky]

Guys, as previously stated, Sony Exmor CAN bin on the sensor. It is available on many of their sensors and saves readout time or enables higher frame count by the cost of resolution.

...but that‘s really not the point here.

 

[WithTheBirds]

Guys, as previously stated, Sony Exmor CAN bin on the sensor. It is available on many of their sensors and saves readout time or enables higher frame count by the cost of resolution.

...but that‘s really not the point here.

Previously stated by who? Vertical and horizontal line addition isn't binning- it has the same outcome with respect to grouping pixels but it is done differently.