Just got pinged by DJI support

Joined
Nov 17, 2016
Messages
1,084
Reaction score
485
Age
67
Location
Utah, USA
I just received a private message from DJI through a YouTube message. They'd seen my YouTube video on the video problems and want me to send the original files. I'll send then the first file of the flight as that contains both the blown highlights with low IRE values and the pulsing artifact -- not possible to send the whole 13GB.


Brian
 
Interesting they are reaching out. I wonder if they will actually fix the issue? I'm doubtful since it's been around so long.
 
We shall see, but as with all thing DJI ... don't hold your breath...


Brian
That's awesome. They are actually requesting for the customer's feedback. That can definitely give the company some insight, so hopefully they can make the adjustments. I think you can upload your file(s) to onedrive, or something similar, then send them a link. They can then download it directly, but have fun waiting for that 13GB to process. :)

Please keep us updated^
 
  • Like
Reactions: westray
That's awesome. They are actually requesting for the customer's feedback. That can definitely give the company some insight, so hopefully they can make the adjustments. I think you can upload your file(s) to onedrive, or something similar, then send them a link. They can then download it directly, but have fun waiting for that 13GB to process. :)

Please keep us updated^


I sent them just the first flight file of 4GB using a share with Google Drive - it's limited to 10GB so I would not be able to send all four files anyway. Here's the text of my email that accompanied the video file.

DJI_0001.MP4
Gentlemen,

In the YouTube video that you refer to (
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
)
I refer to several issues I have with the P4P and I have attached one of the four files of the flight in question. In it you will see the blown highlights in the sky and also the flickering/pulsing artifact that occurs every 30 frames -- flickering most obvious from about 3:30 onward, while flying over the pond. So, let me start at the beginning.


1. The camera settings were D-Log, with a custom setting of -2/-2/-2 for sharpness, contrast and saturation. The exposure was controlled manually with 30fps, f/6.3, ISO100 and a shutter speed determined partly by the Go app histogram and partly by the observation of the video feed.

2. It is always difficult to see the Go app screen when working in bright daylight even with the brightness turned all the way up on my Samsung tablet. When using D-Log in particular the washed out screen is even more washed out due to the DR compression of D-Log making it even more difficult to see the video feed on the Go app. This makes confirming critical focus and exposure very difficult.

3. I set the exposure first by looking at the video feed and then checking the histogram. The histogram looked very low so I adjusted the exposure up a bit and then the feed looked over exposed so I reduced it half way between those points. At this point the peak value in the histogram was less than 75 indicating no clipping and that the exposure was pretty far away from clipping the highlights. The min value was about 20 indication there was no crushing of the blacks.

4. I reviewed the video in Adobe PP CC and found the problems mentioned in the video with terribly blown highlights even the the Go app histogram did not show that the exposure was any where near clipping. I also noted the flickering/pulsing problem and verified it occurs every 30 frames. You can single step the video and when going from frame 29 to frame 0, every second, you see the Lumetri scope indicate a jump in the values.

5. This flickering problem is not limited to the new P4P and also exists with the Inspire 1 Pro with X5 camera. I own both the P4P and the I1 Pro and this flickering problem occurs with both cameras. I have found that the flickering issue becomes more noticeable after doing contrast stretching in post and the more you stretch the contrast the more visible the flickering. D-Log requires a large amount of contrast stretching and seems to produce the greatest amount of visible flickering in the final output.


So, my feeling is that D-Log is not properly handling contrast and that data is being lost. However, even if this turns out not to be the case the operational problem is that it is so hard to determine proper exposure and critical focus when using D-Log due to the washed out screen. Also, the flickering effect is made worse when using D-Log due to the contrast stretching amplifying the artifact.

You will note that I have posted two more videos with the P4P that are not as critical and demonstrate that even with the image processing problems it's still possible to get good video. In this video (
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
) I shoot using D-Cinelike and was able to get much better exposure because the histogram was more accurate and it was easier to see the video feed. My last video (
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
) was also shot using D-Cinelike with a custom setting of -1/0,-1 and also using the new firmware from 09Dec2016. Exposure was once again manual with f/5.6, ISO100 and a shutter speed that was 1/1000 throughout.

For me the biggest problem is the flicker issue but I would also like to see D-Log improved both from a functioning standpoint and an operational one. The Go app histogram needs to be useful to help set the exposure because the video feed, being so washed out, makes visual adjustment of exposure very difficult.


Thank you,

Brian Stirling
 
i thinks it's promising that DJI have reached out to it's customers, for all we know they may not have known about the fault (or am i being naieve).
 
I've done a lot of work with China suppliers. That email is too long and confusing. They'll understand about 20% of that.

Emails in the future need to be short and precise and to the point. English is a very hard language to understand, especially if it's not your first language.
 
  • Like
Reactions: Rasit
I sent them just the first flight file of 4GB using a share with Google Drive - it's limited to 10GB so I would not be able to send all four files anyway. Here's the text of my email that accompanied the video file.

DJI_0001.MP4
Gentlemen,

In the YouTube video that you refer to (
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
)
I refer to several issues I have with the P4P and I have attached one of the four files of the flight in question. In it you will see the blown highlights in the sky and also the flickering/pulsing artifact that occurs every 30 frames -- flickering most obvious from about 3:30 onward, while flying over the pond. So, let me start at the beginning.


1. The camera settings were D-Log, with a custom setting of -2/-2/-2 for sharpness, contrast and saturation. The exposure was controlled manually with 30fps, f/6.3, ISO100 and a shutter speed determined partly by the Go app histogram and partly by the observation of the video feed.

2. It is always difficult to see the Go app screen when working in bright daylight even with the brightness turned all the way up on my Samsung tablet. When using D-Log in particular the washed out screen is even more washed out due to the DR compression of D-Log making it even more difficult to see the video feed on the Go app. This makes confirming critical focus and exposure very difficult.

3. I set the exposure first by looking at the video feed and then checking the histogram. The histogram looked very low so I adjusted the exposure up a bit and then the feed looked over exposed so I reduced it half way between those points. At this point the peak value in the histogram was less than 75 indicating no clipping and that the exposure was pretty far away from clipping the highlights. The min value was about 20 indication there was no crushing of the blacks.

4. I reviewed the video in Adobe PP CC and found the problems mentioned in the video with terribly blown highlights even the the Go app histogram did not show that the exposure was any where near clipping. I also noted the flickering/pulsing problem and verified it occurs every 30 frames. You can single step the video and when going from frame 29 to frame 0, every second, you see the Lumetri scope indicate a jump in the values.

5. This flickering problem is not limited to the new P4P and also exists with the Inspire 1 Pro with X5 camera. I own both the P4P and the I1 Pro and this flickering problem occurs with both cameras. I have found that the flickering issue becomes more noticeable after doing contrast stretching in post and the more you stretch the contrast the more visible the flickering. D-Log requires a large amount of contrast stretching and seems to produce the greatest amount of visible flickering in the final output.


So, my feeling is that D-Log is not properly handling contrast and that data is being lost. However, even if this turns out not to be the case the operational problem is that it is so hard to determine proper exposure and critical focus when using D-Log due to the washed out screen. Also, the flickering effect is made worse when using D-Log due to the contrast stretching amplifying the artifact.

You will note that I have posted two more videos with the P4P that are not as critical and demonstrate that even with the image processing problems it's still possible to get good video. In this video (
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
) I shoot using D-Cinelike and was able to get much better exposure because the histogram was more accurate and it was easier to see the video feed. My last video (
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
) was also shot using D-Cinelike with a custom setting of -1/0,-1 and also using the new firmware from 09Dec2016. Exposure was once again manual with f/5.6, ISO100 and a shutter speed that was 1/1000 throughout.

For me the biggest problem is the flicker issue but I would also like to see D-Log improved both from a functioning standpoint and an operational one. The Go app histogram needs to be useful to help set the exposure because the video feed, being so washed out, makes visual adjustment of exposure very difficult.


Thank you,

Brian Stirling

Glad to hear that DJI has reached out. Looking forward to seeing an update to this thread after they have responded. :)
 
I posted a discount code from an authorized dealer on my Facebook group and dji contacted the dealer telling me to remove the post. They had a screenshot of the whole thing, I guess it went all the way up to the VP
 
Once they verify the problem it will no doubt go on some kind of punchlist. Whether anything happens will depend on

1) Whether it can be fixed in firmware and if so,
2) What kind of priority they assign it.

The latter may be the determining factor, as in they say 'hey, he's right' but have also received very few complaints so it gets a lower priority than other issues, and when fixing firmware bugs a low priority sometimes means never. So if this is a problem for you you should open a ticket so DJI knows it is really customer impacting. Personally I can't see it even if I look for it (I can detect it in a waveform display but not in the image) but clearly others can.
 
Last edited:
I've done a lot of work with China suppliers. That email is too long and confusing. They'll understand about 20% of that.

Emails in the future need to be short and precise and to the point. English is a very hard language to understand, especially if it's not your first language.


Probably not going to be Chinese engineers looking at it -- at least at first.


Brian
 
  • Like
Reactions: Kman
Personally I can't see it even if I look for it (I can detect it in a waveform display but not in the image) but clearly others can.

I have posted this very clear example of the pulsing issue in another thread. This is the only scene from my footage so far of which I would define this as a clear "issue"

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

(segment 1: h264 extreme curve, 2: h265 extreme curve, 3: h264 basic adjustment, 4: h265 basic adjustment)

In this example, it only shows up in the h265 sample, which I assume is more or less random (it is also observable in h264 footage but in a lesser degree).

You can clearly see that the pulse/flickering is a side effect of the picture gradually degrading after every i frame. Also, only after considerable adjustments is this observable outside the scopes. This is especially good news when they (hopefully) soon give us a working dlog which will give us plenty of tones also in the shadows which will certainly help also this issue as well as most banding problems.

The dlog should be fixable with a better calibrated lut or algorithm during initial toning from raw data (avoiding the clear compressing of tonal range) which should be a easy fix. The compression artifacts should be possible to mitigate atleast somewhat by changing some of the compression-parameters (also probably fixable in firmware).

I find it very reassuring that they are now evaluating both issues of which I think the dlog one the first we will see resolved (since this seems to already have been recently fixed for p3)
 
Last edited:
That's great.. DJI has
Got an important year and I am sure they want to hold the upgoing line. I don't believe there is a long term bussinessmodel that would suggest not to listen to your customer.. they're learning! Thumbs up

Sent from my SM-G935F using PhantomPilots mobile app
 
Did anything come back from this. Any word if they are looking to fix the problem?


Not yet, and I won't hold me breath either. So, I'll keep my fingers crossed that something comes of it but usually a request like that is a PR effort. We'll see...


Brian
 
Well, DJI just replied with there evaluation of the video I sent them -- here's there response:

Hi Raptorman0909: We have analyzed your original footage and confirmed the video itself does not contain pulsating (flickering) frames. At the present stage, ultra high definition H.265 playback and editing are poorly supported by consumer level computers, and it is necessary to rule out the performance of the computer as a possible cause of the flickering. Therefore, we extracted the frames one by one from the video stream and directly viewed the waveform/histogram of the decoded frame sequence (procedures attached), thus eliminating the real-time H.265 decoding process that demands enormous computation power. The frame sequence is free from the pulsating effect. In your YouTube clip, Premiere is decoding the H.265 video at a frame rate as high as 30 fps. However, we found that decoding 4K H.265 video at 100 Mbps is performed at about 3 fps on a Mac Pro using FFmpeg. Without dedicated hardware decoder for UHD H.265, it is rather difficult for Premiere to decode the video at 30 fps, which is also verified at DJI. Therefore, our guess considering the aforementioned experiments is that the decoder might choose to reduce the image quality in order to keep the decoding speed, which results in a flickering video. Our method is independent of the speed of the computer and proves the data contained in the video file is in good condition. As to the “diagonal washing”, we do not observe it. Sincerely, DJI R&D


OK, I guess I didn't expect much more than that but I'll make a couple points here:

1. I do not have a basic consumer level PC, in fact, my PC is a pretty high end PC with i7-5820K CPU that's over clocked as well as a 980 Ti GPU that is also over clocked -- definitely not a run of the mill PC.

2. The flickering effect was noted on my other DJI drone, the Inspire 1 Pro, and is not present in the video of any other video camera I own -- camera including: Sony HDR-CX500, Panasoni GH2 and G85, Nikon D800E and D7200. So, if this is a PP issue it's pretty amazing that it only effect the DJI cameras.

3. The washing effect IS apparently a PP artifact and I was later able to identify how and when this washing effect happens. When replaying the video within PP at less than "full" resolution I see this effect, but when "full" resolution is selected I do not see the effect. I'm not sure what the cause for that is but this one might well be a PP issue.

So, I'm not sure where we go from here but it looks like DJI has determined the problem is not at there end.


Brian
 
Brian,
Did they say anything about the IRE limitation with D-Log?


Sent from my iPhone using Tapatalk
 
Brian,
Did they say anything about the IRE limitation with D-Log?


Sent from my iPhone using Tapatalk


No, they said what I posted. Honestly, I didn't expect them to admit a problem so although it would have been nice if they'd come back with a solution or advise that might have been about the most I could have expected. So, it is what it is...


Brian
 

Members online

No members online now.

Forum statistics

Threads
143,066
Messages
1,467,358
Members
104,936
Latest member
hirehackers