MAVIC 3 - ISO TEST 4K (Low Light) - Mavic 3 vs Mavic 2 vs Inspire X5S

[brett8883]

Well guys, please accept my apologies. I assumed the codec specs were the same between the two.

I’m not convinced it isn’t the same, the DJI documentation has a lot of errors and omissions but I have an X5S but don’t have an X7 at the moment so I can’t test the X7.

 

[kodiak]

I had a X5S early on but sold it when we migrated to the X7. Prior to PRR I pretty much only shot CDNG along with some 4444XQ. I don't do a lot of post but I do enough that I don't like the CDNG workflow. I prefer ProRes. But was disappointed that it was only 10 bit. On this topic I figured the X5S wouldn't be different- and especially not better than the X7 in terms of encoding. Looking at the X5S specs on the DJI site they don't have an equivalent chart that shows bit depth like the X7 page does. Unfortunate but like you said, DJI documentation has a lot of inconstancies. It does mention 12 bit for RAW specifically, tho.

I joined this forum because I have a M3C on order. I'm still waiting for it to be delivered. I look forward to doing some tests between it and the output of the X7. I know they will be different but as long as the log ProRes footage even somewhat matches the level of detail and dynamic range the X7 provides it will be pretty exciting.

Also, Netvise, thanks for documenting the test.

 

[SkywalkerFeng]

Hi!

Many people are interested in how the sensor works in Mavic 3.
I compared Mavic 3 vs Mavic 2 vs Inspire 2 X5S

I shot everything on video for you.
Enjoy watching and subscribe (I'm preparing a flight time test).

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

Accept third party cookies

Thanks for doing this test. I understand it is focused on video quality only. Is it possible to do another test to compare the image quality difference, especially under high ISO, between the three?

 

[brett8883]

I had a X5S early on but sold it when we migrated to the X7. Prior to PRR I pretty much only shot CDNG along with some 4444XQ. I don't do a lot of post but I do enough that I don't like the CDNG workflow. I prefer ProRes. But was disappointed that it was only 10 bit. On this topic I figured the X5S wouldn't be different- and especially not better than the X7 in terms of encoding. Looking at the X5S specs on the DJI site they don't have an equivalent chart that shows bit depth like the X7 page does. Unfortunate but like you said, DJI documentation has a lot of inconstancies. It does mention 12 bit for RAW specifically, tho.

I joined this forum because I have a M3C on order. I'm still waiting for it to be delivered. I look forward to doing some tests between it and the output of the X7. I know they will be different but as long as the log ProRes footage even somewhat matches the level of detail and dynamic range the X7 provides it will be pretty exciting.

Also, Netvise, thanks for documenting the test.

I probably would have come to the same conclusion as you if I hadn’t tested it myself. Did you actually test the ProRes 4444XQ to see if it really is 10 bit? At this point I wouldn’t assume anything on that graph is correct.

 

[Netvise]

Thanks for doing this test. I understand it is focused on video quality only. Is it possible to do another test to compare the image quality difference, especially under high ISO, between the three?

In the description of the video, I added a link to the source code. Subscribe

 

[SkywalkerFeng]

In the description of the video, I added a link to the source code. Subscribe

Will do

 

[kodiak]

Fired up the RAID and took a look at some older files and one new one. Quicktime shows the 422HQ and 4444XQ files from my X5S as "up to 12-bit". But it shows the same "up to 12-bit" with 4444XQ files from the X7 as well. If I open one of the 4444XQ files in VLC it seems to show 10 bit. I'm not sure if I'm interpreting that correctly.

X7 ProRes RAW HQ:


X5S ProRes 4444XQ:


X5S ProRes 422HQ:


X5S ProRes 4444XQ (VLC media info):

 

[brett8883]

Fired up the RAID and took a look at some older files and one new one. Quicktime shows the 422HQ and 4444XQ files from my X5S as "up to 12-bit". But it shows the same "up to 12-bit" with 4444XQ files from the X7 as well. If I open one of the 4444XQ files in VLC it seems to show 10 bit. I'm not sure if I'm interpreting that correctly.

X7 ProRes RAW HQ:
View attachment 139319

X5S ProRes 4444XQ:


X5S ProRes 422HQ:
View attachment 139321

X5S ProRes 4444XQ (VLC media info):
View attachment 139322

It looks like VLC player is giving you the Decoded format so it can play on your monitor and I think QuickTime is just showing you the max bit depth of the codec not the actual color depth.

I like to use DaVinci Resolve to figure these things out because what really matters is what color depth the editor can recognize.

 

[Netvise]

It looks like VLC player is giving you the Decoded format so it can play on your monitor and I think QuickTime is just showing you the max bit depth of the codec not the actual color depth.

I like to use DaVinci Resolve to figure these things out because what really matters is what color depth the editor can recognize.

In the description for the video there are sources in ProRes and RAW

 

[brett8883]

In the description for the video there are sources in ProRes and RAW

In the VLC Player codec information under “decoded” format which is what I assume you are referring to, it shows the format the video was decoded to not the encoded format which is at the top under “codec.”

VLC player must decode any codec to play it back and it uses YUV format to do play back the file. Computer monitors aren’t able to display 12 bits so it has to be scaled to 10 bit color.

ProRes 4444 XQ is RGBA not YUV hence the extra 4.

 

[canadian drone videos]

You're right. Before the test, I thought for a long time about putting the 2.8 aperture on Inspire 2 or not. I put 2.0 because this is an undeniable advantage of Inspire 2. If you put less than 2.0, the picture is not clear.

What x5s lens were you using? I take it the Olympus 12mm as I don’t think the DJI 15 has F2.0 if memory serves me correctly. Why wouldn’t you just use the DJI 15 at F2.8 keep it the same for the demonstration. I’d like to see the x5s DJI 15mm at F2.8 verses the M3 at F2.8. Can you make that happen?

 

[brett8883]

What x5s lens were you using? I take it the Olympus 12mm as I don’t think the DJI 15 has F2.0 if memory serves me correctly. Why wouldn’t you just use the DJI 15 at F2.8 keep it the same for the demonstration. I’d like to see the x5s DJI 15mm at F2.8 verses the M3 at F2.8. Can you make that happen?

The stock lens on the X5S is a 15mm f/1.7

The test could have done at f/2.8 but if you are testing low light capability does it make sense to nerf the light gathering capability of the X5s just to match the M3? Realistically you will never use a higher ISO when you can just open the aperture. I personally think it’s important to show this undeniable advantage of the X5S. Otherwise people are going to come to the conclusion that the M3 is as good or better in low light which isn’t the case.

 

[brett8883]

In the description for the video there are sources in ProRes and RAW

I'm finally home now and could do the tests to show you

X5S ProRes 4444 XQ-12 bit


X5s ProRes 422 HQ-10-bit


X5S CinemaDNG-12 bit


For the record I tried it in normal color, D-log, EI Mode on, and EI Mode off and it made no difference to the bit depth. I would be a little surprised if ProRes 4444 XQ was different on the X7 but it is certainly possible.

 

[canadian drone videos]

The stock lens on the X5S is a 15mm f/1.7

The test could have done at f/2.8 but if you are testing low light capability does it make sense to nerf the light gathering capability of the X5s just to match the M3? Realistically you will never use a higher ISO when you can just open the aperture. I personally think it’s important to show this undeniable advantage of the X5S. Otherwise people are going to come to the conclusion that the M3 is as good or better in low light which isn’t the case.

Yes, but we’re already aware the x5s has an advantage so the test doesn’t tell us anything we don’t already know. I just sold my I2, I only ever used 1.7 I think 3 times. I won’t fly in towns or cities so the low light performance was never that important to me. The one time I really used it was to take this picture of the northern lights with that comet that came by a few years ago, it did a nice job.

 

[Netvise]

Yes, but we’re already aware the x5s has an advantage so the test doesn’t tell us anything we don’t already know. I just sold my I2, I only ever used 1.7 I think 3 times. I won’t fly in towns or cities so the low light performance was never that important to me. The one time I really used it was to take this picture of the northern lights with that comet that came by a few years ago, it did a nice job.

Think of the test as comparing Mavic 2 to Mavic 3. Inspire 2 is a drone from another category and it is an add-on there.

 

[canadian drone videos]

Think of the test as comparing Mavic 2 to Mavic 3. Inspire 2 is a drone from another category and it is an add-on there.

I respectfully disagree, the main camera sensor in the x5s is identical to the main camera sensor in the M3. The I2 x5s and M3 are more comparable then the M2 to the M3. The M3 is basically a flying x5s without interchangeable lenses. Since the X5s and M3 have identical sensors I would be interested to see the difference at F2.8 in low light from both cameras. I already know the 1” sensor on the M2 can’t compete. So id like to see something I don’t know yet?

 

[Netvise]

I respectfully disagree, the main camera sensor in the x5s is identical to the main camera sensor in the M3. The I2 x5s and M3 are more comparable then the M2 to the M3. The M3 is basically a flying x5s without interchangeable lenses. Since the X5s and M3 have identical sensors I would be interested to see the difference at F2.8 in low light from both cameras. I already know the 1” sensor on the M2 can’t compete. So id like to see something I don’t know yet?

Above was the answer to your question:

brett8883 said: The test could have done at f/2.8 but if you are testing low light capability does it make sense to nerf the light gathering capability of the X5s just to match the M3? Realistically you will never use a higher ISO when you can just open the aperture. I personally think it’s important to show this undeniable advantage of the X5S. Otherwise people are going to come to the conclusion that the M3 is as good or better in low light which isn’t the case.

 

[brett8883]

I respectfully disagree, the main camera sensor in the x5s is identical to the main camera sensor in the M3. The I2 x5s and M3 are more comparable then the M2 to the M3. The M3 is basically a flying x5s without interchangeable lenses. Since the X5s and M3 have identical sensors I would be interested to see the difference at F2.8 in low light from both cameras. I already know the 1” sensor on the M2 can’t compete. So id like to see something I don’t know yet?

I mean the test you are suggesting is basically the same as wanting to race a Honda Civic against a Ferrari but you want to limit the throttle on the Ferrari so it can’t go faster than the Civic. What’s the point of racing at that point? If the Civic wins are you going to say the Civic is faster than a Ferrari?

And if you think that’s a bad anology to compare the M3 to a Honda Civic just remember we know the x5S captures 3-4 times more light than the M3. A Ferrari is not 3-4 times faster than a Honda Civic.

 

[canadian drone videos]

I mean the test you are suggesting is basically the same as wanting to race a Honda Civic against a Ferrari but you want to limit the throttle on the Ferrari so it can’t go faster than the Civic. What’s the point of racing at that point? If the Civic wins are you going to say the Civic is faster than a Ferrari?

And if you think that’s a bad anology to compare the M3 to a Honda Civic just remember we know the x5S captures 3-4 times more light than the M3. A Ferrari is not 3-4 times faster than a Honda Civic.

I want a comparison of two entirely different drones using entirely different software but have the identical main camera sensor. Theoretically they should look nearly identical at F2.8, but I have no idea because no one has posted test results. This test didn’t tell me anything I don’t already know, didn’t even need to see the test. I want to know the unknown not the known, what’s so wrong with that?

 

[Netvise]

I want a comparison of two entirely different drones using entirely different software but have the identical main camera sensor. Theoretically they should look nearly identical at F2.8, but I have no idea because no one has posted test results. This test didn’t tell me anything I don’t already know, didn’t even need to see the test. I want to know the unknown not the known, what’s so wrong with that?

Have you seen a test somewhere with all ISO variants and all-color profiles M2 vs M3 vs X5S with attached source files?