4k60p is.. not true 4k.

[Roamer105]

I do see a difference on the 30fps set of frame grabs. Frankly I am surprised and disappointed at the same time not to see more of a difference. Maybe it is due to the h265 encoding parameters set. As I had said before, I am surprised to see DJI using H265 encoding on a package that small along with power constraints knowing well how much compute power it requires for H265 encoding.


Sent from my iPad using PhantomPilots

 

[jcknows0]

I don't have the P4 Pro (boys am I tempted!) but I do have an NX500 that does HEVC h265 encode. I have found that there is no way h264 can compete with improved codec. Hopefully the rest of the market will catch up as at least Adobe supports HEVC editing (Shame on you Apple with your gimmick TouchBar editing!)

 

[Doctorcobani]

Have you tested these camera improvement? Re: #2, I ran more tests this morning and I see zero difference in h265 vs h264 encoding quality at 24, 25 or 30p. See attached framegrabs. Granted, 100mb/s is still a noticeable improvement oven 60mb/s, so there is that, not to mention resolution, noise and dynamic range improvements owed to the larger sensor, which I don't think are completely negated, even at low bitrates. (compare p4/p3p/x3 video with x5 video sometime -- the differences are obvious even at 60mbs).

View attachment 69927 View attachment 69928 View attachment 69929 View attachment 69930 View attachment 69931 View attachment 69932

As to #1, until someone demonstrates otherwise, I'm working on the assumption that the mechanical shutter is only used for photos, not for video (as is the case with every other camera out there). DJI never advertised that the mechanical shutter is used to eliminate rolling shutter effects -- they just advertised that they incorporated a mechanical shutter into the physical camera module, where the p4/p3 had none.

I see difference in this JPGs between H264 and H265. More fine detail in H265, you can see that on the leaves.

Respect to the mechanical shutter, it's OBVIOUS that they works only in stills. Even Canon 1DX Mark II DSLR (U$S 6000 body only) or Cinema Canon C300 MKII don't have mechanical shutter for video.

Make numbers: The best mechanical shooter actuator in DSLR professional range can make up to 300.000 shutter actuations (actuations = open and close shutter). If the P4P camera will use the mechanical shutter to record video, for 60fps must have 60 actuations per second, it's absolutely impossible to achieve that a mechanical shutter (the most fastest are in the 17 actuations per second).
And if they can achieve these speeds in the future, the mechanical shutter will be broken in only 3 flights with 28 minutes each of recorded video, because he will use the 300k actuations of the life expectancy in only 84 minutes in 60fps!!

And I don't name the vibrations and movements and noise that cause a mechanical shutter... it's absolutely impossible in every way. These mechanical shutter are made for photos!!!!

Enviado desde mi Nexus 6P mediante Tapatalk

 

[Roamer105]

Have you tested these camera improvement? Re: #2, I ran more tests this morning and I see zero difference in h265 vs h264 encoding quality at 24, 25 or 30p. See attached framegrabs. Granted, 100mb/s is still a noticeable improvement oven 60mb/s, so there is that, not to mention resolution, noise and dynamic range improvements owed to the larger sensor, which I don't think are completely negated, even at low bitrates. (compare p4/p3p/x3 video with x5 video sometime -- the differences are obvious even at 60mbs).

View attachment 69927 View attachment 69928 View attachment 69929 View attachment 69930 View attachment 69931 View attachment 69932

As to #1, until someone demonstrates otherwise, I'm working on the assumption that the mechanical shutter is only used for photos, not for video (as is the case with every other camera out there). DJI never advertised that the mechanical shutter is used to eliminate rolling shutter effects -- they just advertised that they incorporated a mechanical shutter into the physical camera module, where the p4/p3 had none.

Sorry to harp on this a little more.... Once again, if you talk stills - yes, 20MP is a wonderful improvement! However, when it comes to 4K video, the increased size of the sensor has little or no impact (short of possible optical advantages). It definitively has no impact in resolution and dynamic range on video unless the video is encoded at 10 bit color depth instead of 8 bit (Like Inspire 2, but I have not seen that in the P4P specs). 4K resolution is 4K resolution - period. Technically you only need a 16:9 shaped 9MP sensor for that. However, if the 4K video pixel grab is actually performed at the still frame (16:9 Aspect Ratio: 5472 × 3078) and then down-converted to 4K (4096 × 2160) it may yield a somewhat clearer/sharper picture at the target resolution. But that is strictly a guess.

That is why I said earlier: For video, the consequential improvements on the camera/encoder are the mechanical shutter and H264/H265 at 100mb/s. Everything else (including mechanical shutter) is really good and a substantial improvement for still photography. So, to really compare camera quality of optics and sensor, one would have to take RAW stills on the P4 and the P4P same subject/lighting etc.

 

[Roamer105]

I see difference in this JPGs between H264 and H265. More fine detail in H265, you can see that on the leaves.

Respect to the mechanical shutter, it's OBVIOUS that they works only in stills. Even Canon 1DX Mark II DSLR (U$S 6000 body only) or Cinema Canon C300 MKII don't have mechanical shutter for video.

Make numbers: The best mechanical shooter actuator in DSLR professional range can make up to 300.000 shutter actuations (actuations = open and close shutter). If the P4P camera will use the mechanical shutter to record video, for 60fps must have 60 actuations per second, it's absolutely impossible to achieve that a mechanical shutter (the most fastest are in the 17 actuations per second).
And if they can achieve these speeds in the future, the mechanical shutter will be broken in only 3 flights with 28 minutes each of recorded video, because he will use the 300k actuations of the life expectancy in only 84 minutes in 60fps!!

And I don't name the vibrations and movements and noise that cause a mechanical shutter... it's absolutely impossible in every way. These mechanical shutter are made for photos!!!!

Enviado desde mi Nexus 6P mediante Tapatalk

I think the 17 actuations per second limitation is related to the time it takes to process and transfer to image data to storage. The latter is most likely the bottle neck. I have a Canon 7D Mark II. Canon uses an interim fast buffer to allow short bursts of high rate of still repetitions.

When it comes to the notion of "mechanical shutter" as it relates to video, frankly I am equally puzzled as are you.The best thing I can think of is a rolling shutter that really does not roll from scan-line-group to scan-line-group but simply takes the whole picture as one single scan-line-group and the actual exposure is still controlled by an electronic shutter. If the whole sensor can be treated as one scan-line-group, it effectively is the same as a physical/mechanical shutter. However, there may well be such a mechanical shutter - but it may only be used for still photography. There is a possible hint about the P4P shutter: The spec indicates that the mechanical shutter seamlessly transforms into an electronic shutter once it exceeds 1/2000 of a second. So, what does this really mean? Go from a single sensor all-scan-lines grab to a rolling group-wise scan due to line-scan speed limitations? DJI does not really offer much insight into their camera technology - so we probably will never know.

Another clue might be the specs indicating an 8-1/2000 mechanical shutter and an 8-1/8000 electronic shutter - so question is: Which is it? Which method is used for what? If there is a choice, how is it selected?

Here's another puzzler: How is the focus adjustment done? Is there a servo motor mechanism (USM or whatever) moving lenses? I doubt that, but if not - how? I noticed in the demo videos, the focus adjustment is very slow in comparison to regular cameras.

Similarly, how is the iris aperture controlled? Regular cameras all involve servo motors of some kind resulting in mechanical movement for all of these adjustments......

 

[Dronefriend]

Increased size of sensor has no improvement in video recording for dynamic range? You have no clue what you're talking about.

 

[Roamer105]

Increased size of sensor has no improvement in video recording for dynamic range? You have no clue what you're talking about.

Sorry my friend we may talk about different things here, while a larger sensor may be able to capture more light overall, the effective physical size of a single pixel on the 1" 20MP sensor is actually a little smaller than the P4 sensor pixel size (2/3" at 12MP). Having said that, the real dynamic range for video is determined by the bit-depth (number of bits) recorded for each Y-U-V signal information component per pixel. That is the limiting factor even if the sensor was physically smaller like on a P4. To my knowledge, the P4P encoder still only records at 8 bits - which is the same for P4, Mavic and P3P/A.

 

[Roamer105]

As far as the shutter for video is concerned - here is an interesting piece of information on the $3000 Zenmuse X5R rolling shutter.
How Does the Osmo RAW Compare to Professional Cinema Cameras? | cinema5D
Even very expensive professional cameras with CMOS sensors such as the Arri have a rolling shutter, but it all boils down to "read-out-time". The latter is at 2ms while the former is at 25ms. Needless to say, the faster-the-better. How that relates to the P5P? No idea about that sensor speed. I am quite sure that it is possible to get some rolling shutter artifacts on the P4P video we will find out over time. I suspect the mechanical shutter is only used for stills, just like any other professional camera.

 

[Dronefriend]

Sorry my friend we may talk about different things here, while a larger sensor may be able to capture more light overall, the effective physical size of a single pixel on the 1" 20MP sensor is actually a little smaller than the P4 sensor pixel size (2/3" at 12MP). Having said that, the real dynamic range for video is determined by the bit-depth (number of bits) recorded for each Y-U-V signal information component per pixel. That is the limiting factor even if the sensor was physically smaller like on a P4. To my knowledge, the P4P encoder still only records at 8 bits - which is the same for P4, Mavic and P3P/A.

I like your humble approach.

I've read h265 can record 10bit, if it does on the P4P, I am not sure.

I still fail to see how 20 MP on 1 inch are made up of smaller pixels than 12 MP on 2/3. Does it even matter? A larger sensor area means larger area that captures light, which in turn allows for higher sensitivity.

A 1 inch sensor is the same size as four 2/3 sensors, but 20 MP is not four times the pixels from 12 MP. That would be 50.6 MP.

 

[Roamer105]

I like your humble approach.

I've read h265 can record 10bit, if it does on the P4P, I am not sure.

I still fail to see how 20 MP on 1 inch are made up of smaller pixels than 12 MP on 2/3. Does it even matter? A larger sensor area means larger area that captures light, which in turn allows for higher sensitivity.

A 1 inch sensor is the same size as four 2/3 sensors, but 20 MP is not four times the pixels from 12 MP. That would be 50.6 MP.

You are right - i made a calculation mistake: A single pixel on the P4P 1" (at 20MP) is actually 37% larger than a single pixel on the 2/3" 12MP sensor of the P4. But as you said - does it even matter? Probably not for video for obvious compression constraints reasons. However, it does matter a lot for stills especially if they are recorded RAW. The RAW recorded dynamic range is orders of magnitude larger than for compressed video - not sure how much, but I guess 12 bits. Yes, H265 does allow 10 bit color depth, but the P4P encoder does not do that. I am sure if it did, it would be spelled out in bold letters by DJI - and it would create a product positioning conflict with Inspire 2.

 

[Doctorcobani]

I like your humble approach.

I've read h265 can record 10bit, if it does on the P4P, I am not sure.

I still fail to see how 20 MP on 1 inch are made up of smaller pixels than 12 MP on 2/3. Does it even matter? A larger sensor area means larger area that captures light, which in turn allows for higher sensitivity.

A 1 inch sensor is the same size as four 2/3 sensors, but 20 MP is not four times the pixels from 12 MP. That would be 50.6 MP.

Agree.

And another thing, the quality and technology of the sensor matters, the software behind matters, and so on.

If the sensor size doesn't matter in video, a film can be achieved with a cellphone. So why buy a Canon C300, a Phantom or Red camera? That's non sense.
Why buy a RX100 if a GoPro can bring the same quality, because the sensor size doesn't matter... (It's ironic)

Compare a P4 footage with P4P footage, both in H264 4k 30fps, and you will see the difference. Compare a night footage between they, compare a video in high ISO, and you will see.

The sentence "sensor size doesn't matter in DR and IQ if both will record in 8 bits" is ABSOLUTELY WRONG!!!!

Enviado desde mi Nexus 6P mediante Tapatalk

 

[Doctorcobani]

That is why I said earlier: For video, the consequential improvements on the camera/encoder are the mechanical shutter and H264/H265 at 100mb/s.

Mechanical shutter it has nothing to do with vídeo!!!! It's only used for stills!

Every professional camera use global shutter or electronic shutter, the readout time is the game changer on rolling shutter, with very high read pixel speeds, but this high readout happens in ultra professional cameras, and cost like 20 units or more of P4P.

It's a very intelligent drone with very high quality video and stills, and they IQ has never seen on a consumer drone. For U$S 1500, there's nothing more to say or discuss...


Enviado desde mi Nexus 6P mediante Tapatalk

 

[wsalopek]

For stills.... I've read that, generally speaking, dynamic range is better with a mechanical shutter as opposed to an electronic shutter. There may be other pros and cons... I don't have time to look it up right now.

-

Bill

 

[Andrei V]

Sorry to harp on this a little more.... Once again, if you talk stills - yes, 20MP is a wonderful improvement! However, when it comes to 4K video, the increased size of the sensor has little or no impact (short of possible optical advantages). It definitively has no impact in resolution and dynamic range on video unless the video is encoded at 10 bit color depth instead of 8 bit (Like Inspire 2, but I have not seen that in the P4P specs). 4K resolution is 4K resolution - period. Technically you only need a 16:9 shaped 9MP sensor for that. However, if the 4K video pixel grab is actually performed at the still frame (16:9 Aspect Ratio: 5472 × 3078) and then down-converted to 4K (4096 × 2160) it may yield a somewhat clearer/sharper picture at the target resolution. But that is strictly a guess.

That is why I said earlier: For video, the consequential improvements on the camera/encoder are the mechanical shutter and H264/H265 at 100mb/s. Everything else (including mechanical shutter) is really good and a substantial improvement for still photography. So, to really compare camera quality of optics and sensor, one would have to take RAW stills on the P4 and the P4P same subject/lighting etc.

I think you're conflating a number of things here. The dynamic range that can be captured by a camera (in video) isn't limited by the color depth of the recording codec. You can capture more than 8 stops of dynamic range in an 8-bit codec by compressing the luma data, either linearly or logarithmically (e.g. Sony SLOG2 on the A7RII captures about 12-13 stops of dynamic range in a 8 bit h.264 codec). Dynamic range is actually limited primarily by the sensor size (e.i. the amount of light the camera can capture). Larger sensor = ability to capture more light = less noise = higher dynamic range). The P4P has been measured around 11 stops of dynamic range. The P4? Around 8.

Similarly, the resolution of the recording codec only acts as a cap on the actual resolution the camera can record, not a floor. Or put another way, just because two cameras record 4000x2000 pixels of data per frame doesn't mean they record the same level of information (fine detail) in those two shots. The quality of the lens and the size of the sensor both impact on how much fine detail you're actually acquiring, independent of the video codec. Larger sensors and optically better lenses will record more fine detail. In fact, this has been specifically tested for the p4, p4p and x5s cameras. The p4 camera can only capture around 1100 lines of vertical resolution, whereas the p4p/x5s are around 1400 lines.

Neither actually matches the true 2000 lines of verticual resolution you would be able to encode in a 4k signal. So no -- 4K resolution is not 4K resolution, when it comes to the amount of fine detail you're recording.

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

 

[Andrei V]

I see difference in this JPGs between H264 and H265. More fine detail in H265, you can see that on the leaves.

Respect to the mechanical shutter, it's OBVIOUS that they works only in stills. Even Canon 1DX Mark II DSLR (U$S 6000 body only) or Cinema Canon C300 MKII don't have mechanical shutter for video.

Make numbers: The best mechanical shooter actuator in DSLR professional range can make up to 300.000 shutter actuations (actuations = open and close shutter). If the P4P camera will use the mechanical shutter to record video, for 60fps must have 60 actuations per second, it's absolutely impossible to achieve that a mechanical shutter (the most fastest are in the 17 actuations per second).
And if they can achieve these speeds in the future, the mechanical shutter will be broken in only 3 flights with 28 minutes each of recorded video, because he will use the 300k actuations of the life expectancy in only 84 minutes in 60fps!!

And I don't name the vibrations and movements and noise that cause a mechanical shutter... it's absolutely impossible in every way. These mechanical shutter are made for photos!!!!

Enviado desde mi Nexus 6P mediante Tapatalk

Enviado, I believe the hype was around the fact that the P4P uses a leaf shutter rather than a more traditional plane mechanical shutter used in most DSLRs. In theory, a leaf shutter could allow both faster actuation and (possibly) more durability.

A number of websites following DJI's announcement were misleadingly reporting that the P4P uses a mechanical shutter to eliminate rolling shutter effects in video. It was one of the factors that influenced my decision to upgrade from a P4 to a P4P (but fortunately, not the only factor..)

 

[Andrei V]

As far as the shutter for video is concerned - here is an interesting piece of information on the $3000 Zenmuse X5R rolling shutter.
How Does the Osmo RAW Compare to Professional Cinema Cameras? | cinema5D
Even very expensive professional cameras with CMOS sensors such as the Arri have a rolling shutter, but it all boils down to "read-out-time". The latter is at 2ms while the former is at 25ms. Needless to say, the faster-the-better. How that relates to the P5P? No idea about that sensor speed. I am quite sure that it is possible to get some rolling shutter artifacts on the P4P video we will find out over time. I suspect the mechanical shutter is only used for stills, just like any other professional camera.

Correct. Mechanical shutter is only used for stills. And the readout time for the sensor on the P4P isn't that fast when the whole sensor is read out (such as when shooting at 24p), so you definitely get rolling shutter effects such as bent vertical lines. See my testing here:

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

 

[AIRSPACE]

Correct. Mechanical shutter is only used for stills. And the readout time for the sensor on the P4P isn't that fast when the whole sensor is read out (such as when shooting at 24p), so you definitely get rolling shutter effects such as bent vertical lines. See my testing here:

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

Nice video, thanks for sharing your expertise on the subject.

 

[ianwood]

Back to the original topic, I am disappointed by these results. Unless the shot contains a lot of moving detail (like waves or moving grass), a 4K/60 shot should still fit into a 100mb/s with less detail loss than that. The samples provided look like a interpolation. And if 4K/30 can fit into 60mb/s then 4K/48 should definitely fit into 100mb/s with similar quality.

 

[Andrei V]

Interpolation because it's not really reading out every pixel of the sensor, Ian. Actually, I'm really curious if the X5S does the same in 60p -- do you have yours yet?

 

[Tron_X]

Those frame grabs from 4K 60p appear to be from a traditional line-skipping mode rather than codec issue. It's easier to hide in 4K but still results in aliasing problems.

This machine appears to perform best in 4K 24/30p H.265 and raw stills, which is still worth the price of admission, IMO. "Real" 4K 60p will probably arrive in the P5 or P6.

Is there a 2.7K 60p mode anyone can try? Perhaps they've implemented it with binning rather than skipping and it will prove to be superior with less artifacting.

ETA - Apparently there is a 2720 × 1530 60p mode. I wouldn't be surprised if it out-resolves 4K 60p (if they implement binning rather than line-skipping).