Mavic 3 - ?? Maybe Global Shutter (mechanical) ready for Mapping / Modeling possible?

[Dougcjohn]

The thing is, dji firmware api is almost identical to all drones. As documented here:

dji-firmware-tools/dji-dumlv1-flyc.lua at master · o-gs/dji-firmware-tools

Tools for handling firmwares of DJI products, with focus on quadcopters. - dji-firmware-tools/dji-dumlv1-flyc.lua at master · o-gs/dji-firmware-tools

github.com

My question is you know for sure waypoints isnt in the firrmware? It will save me some time if you have tested that.
Many functions are not visible in the app or sdk, however they still exist. For example the fpv racing drone supports virtualsticks (and much more), but have no sdk, or app that supports it. And not to forget setting gimbal in free mode

My very basic 3d-mapping with the mini2, without sdk.

Mapping

All I have available is a Mini 2 and I need to create a map of 40 acres. Because there isn't SDK support I'll have to do this manually. Any suggestions on how much overlap I should have for each picture? Software I should use to stitch them together? Any other tips that will help me succeed...

mavicpilots.com

I'd highly agree with ya... I reread my post and noticed I didn't indicate that too. Yes, SDK in a company is extremely common and often the Base to add unique subsection for device (drone). No company will use different calls or library procedure names different on each product... results in mass programming confusion and huge libraries.

Thanks for your reference too... nice!

My point was any program procedure needs to be configured on device in some fashion to be usable in program. If not, the procedure needs to call each procedure or stage seperate or create a named procedure within program.

Back to M3... personally I have no reference at this stage. I haven't coded hardware for years and I'm going off video hearsay in a hopeful positive position. Logically, I would assume the M3 FW contains a large library of procedural calls to execute future features... or DJI will slip it in with a new FW update. This Jan 24th FW update was pretty large, I'm assuming it contained undocumented features yet to be announced or activated.

 

[ianwood]

Just to clarify, global shutter and mechanical shutter are related but different beasts. Both mitigate rolling shutter artifacts. Global shutter is a fairly exotic feature limited to high-end cameras, e.g. Arri Alexa, RED Komodo where the whole sensor is read at once. It entirely eliminates the need for a mechanical shutter.

Mechanical shutter is a physical shutter like on a film camera that opens and closes at the designated shutter speed to limit the impact from the delay involved with reading a rolling shutter sensor. It is the noise that is annoyingly simulated on so many non-mechanical cameras and phones.

Lastly, the speed at which a rolling shutter sensor can be read dictates how much the delay impacts the overall image.

 

[EpicFlight]

...If anything I hope it lights a fire under their ***** and wakes them up to the huge market share of aerial mappers that they will be missing out on if they don't support it in GSPro and other mission planning apps.

Well, that's the problem... That market is miniscule, in the noise, compared to all the amateurs they sell to. It's not huge.

Of the M1P drones shipped, what percentage of users have flown WP missions? What percentage even know what that means? Few enough that they can be ignored without meaningfully affecting sales.

And lo and behold, that's what we're seeing. DJI didn't become #1 by being stupid.

I'm one of those ratified customers. I'm frustrated too. But I don't blame DJI. I obviously wasn't a target customer for the M3. They sell other solutions to meet those requirements.

 

[Dougcjohn]

I enjoy the DroneU podcasts, but the messages on M3 are definitely confusing. There is no SDK yes for this drone that is publicly available. Maybe they have some knowledge under NDA? I don't know, but it would be nice to just put this issue to bed and have DJI commit to an SDK for the M3.

I'd agree with your statement... an NDA is probably apparent, Paul has much exposure with several sources often. I think he has access or knowledge prior to announcement on several occasions.

In his previous video he indicated they were going to look into the camera, to explore if it could be a global in an upcoming manual test & mapping. That's an odd statement, unless they were aware of something new. For all his mapping classes, and close relationship with Pix4D personnel... It would be something to consider.

He's suggested a DJI strategic comment, that the M3 may have substantial growth possibly that isn't openly shared that can be activated in future... possibly driven by competition.

Taking the above comment, I ponder the reality of global shutter possibility. The P4P & X4S is an electronic rolling shutter, but is a global shutter below 1/2000s, and same for X7 below 1/1000s.

Several articles on X7 have sited:
Quote "Set the shutter speed priority to 1/1000″ or slower. Any speed faster than 1/1000″ will result in a rolling shutter." Unquote.

Pix4D indicates "Gloabal or Fast Readout" for M3 camera images and from what I understand is the same indicator for X7, X4S, & P4P.

So I question, if a rolling shutter captures an image as a global only in slower speeds... does the shutter technology change? A rolling shutter is a linear shutter... not exposing the full sensor at once... different point in time. But does it at slower speeds? How is that different than "fast readout" if rolling shutter correction is not being applied to image. Wouldn't a rolling shutter still expose image linearly requiring shutter corrections with drone moving quickly over surface?

I find that confusing in terminologies... little contradictory. Particularly when the X7 is indicated mechanical under w/1000s with 24mm, 35mm & 50mm; but the 16mm lens is not mechanical... why too wide, only other difference is it lacks a leaf aperture? Fast Readout doesn't negate a partially exposed sensor in a rolling shutter. Fast Readout indicates ability to capture full image at once.

In my searches, I didn't find any reference where a rolling shutter could provide global shutter single point of time utilizing a fast readout... rolling shutter still has different points of time resulting in need to correct image in mapping programs. The M3 while moving required no rolling shutter correction.

Personally, I don't know... sounds like more info is needed. But it's intriguing why M3 mapping images didn't require Rolling shutter corrections.

 

[Dougcjohn]

Just to clarify, global shutter and mechanical shutter are related but different beasts. Both mitigate rolling shutter artifacts. Global shutter is a fairly exotic feature limited to high-end cameras, e.g. Arri Alexa, RED Komodo where the whole sensor is read at once. It entirely eliminates the need for a mechanical shutter.

Mechanical shutter is a physical shutter like on a film camera that opens and closes at the designated shutter speed to limit the impact from the delay involved with reading a rolling shutter sensor. It is the noise that is annoyingly simulated on so many non-mechanical cameras and phones.

Lastly, the speed at which a rolling shutter sensor can be read dictates how much the delay impacts the overall image.

That's how I've interpreted the difference too... but to add to the confusion, DJI states "Mechanical" shutter below 1/2000s.

 

[brett8883]

I'd agree with your statement... an NDA is probably apparent, Paul has much exposure with several sources often. I think he has access or knowledge prior to announcement on several occasions.

In his previous video he indicated they were going to look into the camera, to explore if it could be a global in an upcoming manual test & mapping. That's an odd statement, unless they were aware of something new. For all his mapping classes, and close relationship with Pix4D personnel... It would be something to consider.

He's suggested a DJI strategic comment, that the M3 may have substantial growth possibly that isn't openly shared that can be activated in future... possibly driven by competition.

Taking the above comment, I ponder the reality of global shutter possibility. The P4P & X4S is an electronic rolling shutter, but is a global shutter below 1/2000s, and same for X7 below 1/1000s.

Several articles on X7 have sited:
Quote "Set the shutter speed priority to 1/1000″ or slower. Any speed faster than 1/1000″ will result in a rolling shutter." Unquote.

Pix4D indicates "Gloabal or Fast Readout" for M3 camera images and from what I understand is the same indicator for X7, X4S, & P4P.

So I question, if a rolling shutter captures an image as a global only in slower speeds... does the shutter technology change? A rolling shutter is a linear shutter... not exposing the full sensor at once... different point in time. But does it at slower speeds? How is that different than "fast readout" if rolling shutter correction is not being applied to image. Wouldn't a rolling shutter still expose image linearly requiring shutter corrections with drone moving quickly over surface?

I find that confusing in terminologies... little contradictory. Particularly when the X7 is indicated mechanical under w/1000s with 24mm, 35mm & 50mm; but the 16mm lens is not mechanical... why too wide, only other difference is it lacks a leaf aperture? Fast Readout doesn't negate a partially exposed sensor in a rolling shutter. Fast Readout indicates ability to capture full image at once.

In my searches, I didn't find any reference where a rolling shutter could provide global shutter single point of time utilizing a fast readout... rolling shutter still has different points of time resulting in need to correct image in mapping programs. The M3 while moving required no rolling shutter correction.

Personally, I don't know... sounds like more info is needed. But it's intriguing why M3 mapping images didn't require Rolling shutter corrections.

It’s a rolling shutter it’s not a global shutter and can never be a global shutter. If it was a global shutter trust me DJI would be shouting that from the roof top. It’s actually quite easy to test for rolling shutter by using a strobe light or camera flash so I have questions about these “experts” on YouTube that are going through all this trouble trying to figure it out.

A mechanical shutter doesn’t read all the pixels at once either so it’s not a global shutter but it reads out so fast it doesn’t make a difference in most cases. All of this is relative. If you really want to get into the minutia of it if the drone is moving when the photo is taken then the drone has moved from when the exposure started to when it ended even with a global shutter so at some point you just gotta say it’s good enough.

The 16mm lens for the X7 has a built in ND filter instead of a leaf shutter so that’s why there’s no mechanical shutter with that one. It’s also an APS-C lens not a full frame lens.

 

[brett8883]

That's how I've interpreted the difference too... but to add to the confusion, DJI states "Mechanical" shutter below 1/2000s.

You are confusing shutter speed with read out speed. Shutter speed is the length of time each pixel is active and exposed to light. Readout time is how fast it takes between when the data from the first pixel is exposed to when the last pixel is exposed.

A Mechanical shutter must close then open then close again to take a photo. Read out time is how quickly it opens or closes. Shutter speed is the time between an closed open then closed. Generally mechanical shutters can read out faster than rolling electronic shutters but can’t have as fast of a shutter speed as electronic shutters because the electronic sensor can expose one pixel at a time. That’s why the X7 and P4P use mechanical shutters for slower shutter speeds but have to revert to electronic rolling shutter for faster shutter speeds.

 

[brett8883]

Here’s a cool video from slow mo guys to show a mechanical shutter in slow motion. Notice how at faster shutter speeds the rear curtain, the closing shutter, has already started blocking the light of the top pixels before the front curtain, the opening shutter, has started exposing the bottom pixels. This is because the shutter speed is above the sync speed. The sync speed is the fastest shutter speed all the pixels are exposed to light at the same time.

Read out time is how fast those blades are moving and shutter speed is the distance from the front curtain to the rear curtain.

 

[Dougcjohn]

It’s a rolling shutter it’s not a global shutter and can never be a global shutter. If it was a global shutter trust me DJI would be shouting that from the roof top. It’s actually quite easy to test for rolling shutter by using a strobe light or camera flash so I have questions about these “experts” on YouTube that are going through all this trouble trying to figure it out.

A mechanical shutter doesn’t read all the pixels at once either so it’s not a global shutter but it reads out so fast it doesn’t make a difference in most cases. All of this is relative. If you really want to get into the minutia of it if the drone is moving when the photo is taken then the drone has moved from when the exposure started to when it ended even with a global shutter so at some point you just gotta say it’s good enough.

The 16mm lens for the X7 has a built in ND filter instead of a leaf shutter so that’s why there’s no mechanical shutter with that one. It’s also an APS-C lens not a full frame lens.

I agree from what we are informed by DJI, they don't state mechanical or global... and reconize that's probably correct. Although until FW and SDK support features supporting mapping & modeling it's a moot point, and if that's an unknown time line the need to clarify camera shutter may be more beneficial not to shout it out for competitor reasons. Do I know with any certainty... absolutely not... wouldn't be the first time DJI concealed a feature or circuit initially.

Just speculation on my part... not suggesting with any certainly. I am impressed how the images were supposedly captured & Quality and how Pix4D processed & reported.

 

[Dougcjohn]

You are confusing shutter speed with read out speed. Shutter speed is the length of time each pixel is active and exposed to light. Readout time is how fast it takes between when the data from the first pixel is exposed to when the last pixel is exposed.

A Mechanical shutter must close then open then close again to take a photo. Read out time is how quickly it opens or closes. Shutter speed is the time between an closed open then closed. Generally mechanical shutters can read out faster than rolling electronic shutters but can’t have as fast of a shutter speed as electronic shutters because the electronic sensor can expose one pixel at a time. That’s why the X7 and P4P use mechanical shutters for slower shutter speeds but have to revert to electronic rolling shutter for faster shutter speeds.

Agree with your statements, and I wasn’t confused but I may not have communicated correctly... often my error on posts. Yes, I did know the 16 DL-S was similar to APS-C and as indicated, lacked a leaf shutter (my error, stated aperture).

Although the mechanical shutter you’re referring and the video you reference is a focal plane shutter (2 curtain)… which do introduce a rolling shutter at higher speeds & not exposing the full sensor. The DJI X4S, X7 and P4P have leaf shutters that don’t create a rolling shutter effect… one advantage of a leaf shutter over dual curtain. When speed is increased (above design based on image size, pixel sensitivity & readout speed: having shutter 1/2000 & 1/1000s) the leaf shutter stops and sensor collects as a rolling shutter. (Also believe video mode is always rolling and you used to be able to switch mechanical off).

A mechanical shutter can be a physical or electronic, it doesn’t always mean a physical shutter. Correct, Readout speed/time isn’t correlated with the shutter speed, although if readout is longer than shutter exposure time it is effected by shutter time. Likewise, if sensor has design of very fast readout, on a slower shutter speed it can capture similar to a global shutter… and as a leaf shutter, there isn’t a rolling effect by shutter.

Another test of rolling vs global Is a prop. If rolling, prop appears as bent or curved, if global, prop straight & blurred (excluding high shutter speed).

That’s where I was entertaining & questioning the global shutter for the M3 and where you indicated not a global, but fast readout. In the situation of having an extremely fast readout sensor, capable of reading a full sensor capture and an electronic shutter that behaves as a mechanical in slower speeds, that would in essence be capturing like a global shutter. When new cameras with newer electronics, no physical shutter involved, it’s then more design of sensor capabilities & programming.

Does that mean truly global shutter or that it performs like a global shutter? I think the DroneU video and the Pix4D software were indicating probably Global capable results... weather a true global shutter or not. Similar to P4P, X4S & X7… they are often referred to as Mechanical (Global) shutter… does that mean Global or global capable? The Pix4D indicated the same for M3 & P4P images: Global or Fast Readout, and didn’t activate or suggest rolling shutter correction algorithms.

Don’t know… Agree, for now it‘s not spec’ed as Global, but if all electronic and shown as very fast readout… I fail to see it not being similar to existing models but with electronic shutter. As competition increases, a FW change may be able to make a change if sensor is currently capable. Both Sony & Canon have improved older models via FW tapping deeper into sensors capabilities.

 

[brett8883]

Agree with your statements, and I wasn’t confused but I may not have communicated correctly... often my error on posts. Yes, I did know the 16 DL-S was similar to APS-C and as indicated, lacked a leaf shutter (my error, stated aperture).

Although the mechanical shutter you’re referring and the video you reference is a focal plane shutter (2 curtain)… which do introduce a rolling shutter at higher speeds & not exposing the full sensor. The DJI X4S, X7 and P4P have leaf shutters that don’t create a rolling shutter effect… one advantage of a leaf shutter over dual curtain. When speed is increased (above design based on image size, pixel sensitivity & readout speed: having shutter 1/2000 & 1/1000s) the leaf shutter stops and sensor collects as a rolling shutter. (Also believe video mode is always rolling and you used to be able to switch mechanical off).

A mechanical shutter can be a physical or electronic, it doesn’t always mean a physical shutter. Correct, Readout speed/time isn’t correlated with the shutter speed, although if readout is longer than shutter exposure time it is effected by shutter time. Likewise, if sensor has design of very fast readout, on a slower shutter speed it can capture similar to a global shutter… and as a leaf shutter, there isn’t a rolling effect by shutter.

Another test of rolling vs global Is a prop. If rolling, prop appears as bent or curved, if global, prop straight & blurred (excluding high shutter speed).

That’s where I was entertaining & questioning the global shutter for the M3 and where you indicated not a global, but fast readout. In the situation of having an extremely fast readout sensor, capable of reading a full sensor capture and an electronic shutter that behaves as a mechanical in slower speeds, that would in essence be capturing like a global shutter. When new cameras with newer electronics, no physical shutter involved, it’s then more design of sensor capabilities & programming.

Does that mean truly global shutter or that it performs like a global shutter? I think the DroneU video and the Pix4D software were indicating probably Global capable results... weather a true global shutter or not. Similar to P4P, X4S & X7… they are often referred to as Mechanical (Global) shutter… does that mean Global or global capable? The Pix4D indicated the same for M3 & P4P images: Global or Fast Readout, and didn’t activate or suggest rolling shutter correction algorithms.

Don’t know… Agree, for now it‘s not spec’ed as Global, but if all electronic and shown as very fast readout… I fail to see it not being similar to existing models but with electronic shutter. As competition increases, a FW change may be able to make a change if sensor is currently capable. Both Sony & Canon have improved older models via FW tapping deeper into sensors capabilities.

Perhaps the most damning evidence for it not being and not being capable of being a global shutter is the fact that Sony semiconductor doesn’t make a MFT sized global shutter sensor that would meet the M3 dimensions. There’s also no other commercially available global shutter MFT sensors that match the M3’s specs. In all likelihood the M3 uses an IMX272 from Sony which matches all specs from the M3 which means it’s not even a backside illuminated much less a stacked sensor like the Nikon Z9 and Sony A1.

I even think that’s a great possibility that Pix4D just hasn’t created a rolling shutter correction profile for the M3 yet and therefore that’s why the global shutter/fast read out is the only option in that software. If that’s the evidence people are going off of that’s it’s a “global” shutter that’s extremely weak evidence.

A leaf shutter would probably count as a global shutter I’ll give you that but the M3 doesn’t have a leaf shutter.

 

[Dougcjohn]

Perhaps the most damning evidence for it not being and not being capable of being a global shutter is the fact that Sony semiconductor doesn’t make a MFT sized global shutter sensor that would meet the M3 dimensions. There’s also no other commercially available global shutter MFT sensors that match the M3’s specs. In all likelihood the M3 uses an IMX272 from Sony which matches all specs from the M3 which means it’s not even a backside illuminated much less a stacked sensor like the Nikon Z9 and Sony A1.

I even think that’s a great possibility that Pix4D just hasn’t created a rolling shutter correction profile for the M3 yet and therefore that’s why the global shutter/fast read out is the only option in that software. If that’s the evidence people are going off of that’s it’s a “global” shutter that’s extremely weak evidence.

A leaf shutter would probably count as a global shutter I’ll give you that but the M3 doesn’t have a leaf shutter.

Lol.. not arguing ya... just seeing possibilities.
As far as the sensor, I haven't seen or heard that degree of technical specs. You may have a point on Pix4D profile, although if I recall they did mention the camera was identified in Pix4D too and that indicates a profile being applied.

In the end, for Drone image collections; if I can collect global "similar / comparable" MFT quality images that lack rolling shutter artifacts and not need SW corrections... applying rolling shutter SW correction isn't sufficient at times in small detail. Once the SDK and WP incorporated in 3rd party apps... for my needs and quality, that's Great!

I'll be able to use something smaller than my I2's w X4S or X7... if the M3's MFT camera & glass are comparable. Gaining Newer technologies, longer flight, better radio connection & video transmission... much faster setup.. in a package 1/4 the size. Works for me!

Didn't say M3 had a leaf.. I said all electronic.

 

[Wreckhunter]

Hi - Trying to get the word out today. I just got confirmation from a DJI rep the M3 Pro will NOT be able to be used for aerial mapping. "The GSPro would not be supporting the M3, the M3 is not Mobile SDK compatible and it does not support waypoint mission." This is HUGELY DISSAPOINTING for anyone expecting this to replace the Phantom 4 Pro line for aerial photogrammetry. Please pass this on. Since it is past 15 days since purchase we are having issues easily returning. DJI is really trending downhill lately.

Trending downhill na,a few missing features on one dji model,that are now in place and the whole company is on a downhill trend.

 

[JoelP]

I think that I read that DJI will eventually issue an SDK for the Mavic3. If the Mavic 3 FW doesn't support the control signals that a 3rd party app would send to control sticks, then what would be the use for any SDK?

 

[Dougcjohn]

I think that I read that DJI will eventually issue an SDK for the Mavic3. If the Mavic 3 FW doesn't support the control signals that a 3rd party app would send to control sticks, then what would be the use for any SDK?

There will be a SDK... no question, no worries. I think several sources have confirmed some degree of development in the mix.

And correct, once released... soon to follow will be supported 3rd party apps. How that SDK will be processed on craft is unknown... will likely follow the other crafts that are based on FLY App. We can hope the M3 being a higher end craft may include autonomous on-board processing ( no radio connection required) but most likely not.

Over the next 4-8 months we should see the remaining M3 features & tid bits released... I'd assume there was more to come.

 

[Cookedinlh]

I've read a few articles that suggested the M3 MFT camera was a Global Shutter (referred as mechanical) similar to the P4Pv2, Inspire 2 X4S & X7 cameras.
The recent 1/24/22 DroneU video on the M3, they performed a Manual Mapping exercise lacking any currently supported Mapping / Waypoint SW that the SDK will allow.
During the manual mapping exercise & software processing, it was confirmed that the M3's MFT Camera via Pix4D does have a Global shutter... like the P4P. I'm assuming it will be similar to P4P & X4S that it's mechanical up to 1/2000s shutter, and Electronic Shutter above. According to DroneU, the images were outstanding, the MFT sensor exceeding expectations.

It's also confirmed that the M3 SDK has been distributed, there are a few Development shops currently developing 3rd Party software that will include Mapping & Modeling.

Looks like the M3 is gaining more ground with time, gaining use for Mapping & Modeling is another welcome feature as SW is developed.

I did a 6km long spillway manual map with my new M3 and covered it well with maybe 2 1/2 batteries on three flights. I'll be watching for Litchi for the M3 . . anyone know?. M3 does amazing detail even at 300-400ft

 

[Dougcjohn]

I did a 6km long spillway manual map with my new M3 and covered it well with maybe 2 1/2 batteries on three flights. I'll be watching for Litchi for the M3 . . anyone know?. M3 does amazing detail even at 300-400ft

Great to hear… what did you use to process the images?
The SDK for 3rd Party Developers will come…but I doubt before Fall time period.

 

[Gabriel Arcieri]

Estou executando levantamento topográfico e nuvem de pontos com M3. O resultado tem sido excelente mesmo realizando vôo manual. Este ortomosaico representa uma área de 67.000m². Usei apenas uma bateria.

 

[GadgetGuy]

English, please...

 

[Cookedinlh]

English, please...

easy as "google Translate" I am executing topographic survey and cloud of points with M3. The result has been excellent even when performing manual voo. This orthomosaic represents an area of 67,000m². I barely used a battery.