Photo buffering bug found? Can others test please?

Joined
Jan 2, 2017
Messages
23
Reaction score
6
Posted to DJI forum, but no one has attempted to replicate issue yet. Would appreciate others trying please...
PROBLEM EXPERIENCE

Wasted a whole afternoon shooting a commercial job, high res panoramas (to get big files) of 8 images x 5 AEB (tricky llighting) = 40 frames each panorama.

Downloaded the files and discovered numerous AEB sets missing? A wasted 300km round trip and 1/2 day of work. Very annoying.

TEST
Ran tests in studio shooting 5 frame AEB's as fast as possible for over 100 frames (20+ AEB sets) and discovered a problem.

RESULTS (5 frame AEB groups. 'x' = missing frame)
11111 - 22222 - xxxxx - 44444 - xxxxx - xxxxx - 77777 - xxxxx - xxxxx - 1010101010 - xxxxx - xxxxx - 1313131313 - xxxxx - xxxxx - xxxxx - 1717171717 - xxxxx - xxxxx - xxxxx - 2121212121 - xxxxx - xxxxx - xxxxx etc etc*

CONCLUSION
P4P DUMPS full sets of buffered AEB files in favour of NEWER AEB sets? This is a software failure
This is highly problematic as the controller gives audible feedback when frames are taken (click x 5 etc) but does not indicate when the buffer is full. Cameras mustl stop taking more images until there is buffer space (or files are written), but DJI chucks all of them out?! There is no way, that I know, to tell if you have images written to card memory except to wait a very conservative time between shots, otherwise you will have missing frames. Not good.

REQUESTS
1) Can others please try replicating this test and report results?
2) Can DJI please check that this bug? I can't imagine 5 shot RAW AEB buffering is commonly tested, and may have been overlooked?

*The card was a Sandisk Pro 64gb U3, but similar results with other cards. Image files 2:3 format RAWs.
** Before you suggest it, no, using jpegs instead of RAW wasn't an option.
 
What are your image sizes? If you're bracketing at 5 fps and your image sizes are greater than 20MB then you may be running into a situation where (1) the SD bus controller is unable to keep up no matter how fast your SD card is, and (2) you're back to back shots are hitting the physical limit of the P4P's image buffer.

As I understand it, DJI specs the SD bus at 100 mb/s translating to 12.5 MB/s. It also specs burst frames up to 14 fps.

If you're shooting 20 back to back brackets of 5 - 20 MB frames @ 5 fps, the P4P would have to buffer an aggregate overflow of 1.7 GB of image data. At 100 mb/s (12.5 MB/s) it would take over 2 minutes to clear the buffer after 20 sets assuming unlimited buffering.

But the buffer has a limit which I'm estimating at ~250 MB based on the maximum spec'd "burst rate" of 14 fps.

If my assumptions are correct, you'll capture and fill the image buffer at 2-3 sets of 5 frames. Everything for the next couple of seconds will be lost until the buffer is available again. Rinse and repeat.

So perhaps the work around is slow your rate of yaw to 1/2 or 1/3 to allow sufficient interval between brackets for the buffer to clear?
 
What are your image sizes?

About 40Mb

So perhaps the work around is slow your rate of yaw to 1/2 or 1/3 to allow sufficient interval between brackets for the buffer to clear?

Yes, I understand what you're saying, but no other camera I have used ever allows already taken frames to be dumped ahead of new frames.

All digital cameras:
1) operator initiates photo 2) IF files can be written to card or buffer THEN take photo, ELSE wait until files can be written (or buffer frees space)
P4P situation is:
1) operator initiates photo 2) IF files can be written to card or buffer THEN take photo, ELSE dump all previous unwritten files and take photo This is an oversight/bug and should be a simple fix.
 
Various - the ones in the test centred around 1/20th sec - doesn't really matter until you get to long exposures I'd think?

It might not matter, but good to know just to replicate the exact same situation. A smart camera would spend the 1/20 of a second writing to the card
 
About 40Mb



Yes, I understand what you're saying, but no other camera I have used ever allows already taken frames to be dumped ahead of new frames.

All digital cameras:
1) operator initiates photo 2) IF files can be written to card or buffer THEN take photo, ELSE wait until files can be written (or buffer frees space)
P4P situation is:
1) operator initiates photo 2) IF files can be written to card or buffer THEN take photo, ELSE dump all previous unwritten files and take photo This is an oversight/bug and should be a simple fix.

Possibly so. Makes me wonder if the issue could be handled in DJI's telemetric handshake for shutter control or whether communication latencies generally preclude it.

In other words, rather than introduce latency between a shutter press from the controller or smart device and delaying shutter click feedback until the frames are written, DJI gives you shutter click feedback immediately from the controlling device while performing its file handling overhead independently from the controller. Thus every time you press the shutter button, you get a false sense of security that an image is actually taken when it's not.

In the interim, have you figured out a workaround? Could you set up an app like Litchi or Pano to index the aircraft rotation at a rate consistent with frame capture & writes?
 
  • Like
Reactions: GadgetGuy
There is no way, that I know, to tell if you have images written to card memory except to wait a very conservative time between shots, otherwise you will have missing frames. Not good.

There is. There's a spinning wheel animation around the shutter button on the Go4 app while the buffer is being written to the card.

Also, I take AEB shots with my P4P and Mavic all the time and subsequent shutter presses are NOT allowed until the buffer is clear. It won't even allow a half-press on the shutter button to focus the shot or adjustment of camera settings until the buffer is clear.
 
Possibly so. Makes me wonder if the issue could be handled in DJI's telemetric handshake for shutter control or whether communication latencies generally preclude it.

In other words, rather than introduce latency between a shutter press from the controller or smart device and delaying shutter click feedback until the frames are written, DJI gives you shutter click feedback immediately from the controlling device while performing its file handling overhead independently from the controller. Thus every time you press the shutter button, you get a false sense of security that an image is actually taken when it's not.

In the interim, have you figured out a workaround? Could you set up an app like Litchi or Pano to index the aircraft rotation at a rate consistent with frame capture & writes?
I suspect you're right about the controller versus whats happening overhead, but other things seem to lock off when the camera is working like focus from memory. I don't have a satisfactory work around yet other than guessing a wait time. Others speculate that the shutter button animation is a work around, but it's another thing to have to watch. I've tried Litchi panos but the overlaps they use are minimal and in my testing made for poor stitching (possibly due to the strong lens profile changes). I've had good results controlling overlap myself.
 
There is. There's a spinning wheel animation around the shutter button on the Go4 app while the buffer is being written to the card.

Also, I take AEB shots with my P4P and Mavic all the time and subsequent shutter presses are NOT allowed until the buffer is clear. It won't even allow a half-press on the shutter button to focus the shot or adjustment of camera settings until the buffer is clear.
Would love you to test what you say there that your P4P does not allow subsequent presses until the buffer clears, as this is directly opposite to what I've systematically tested. My issue is clear that in testing buffering DOES NOT stop subsequent shutter presses when it should
 
Would love you to test what you say there that your P4P does not allow subsequent presses until the buffer clears, as this is directly opposite to what I've systematically tested. My issue is clear that in testing buffering DOES NOT stop subsequent shutter presses when it should

It happens all the time. When trying to take quick follow up shots or even just adjust camera settings I cannot until the buffer is cleared. Pressing the shutter button does nothing until the wheel stops spinning.
 
Rob8888, what revision firmware/software are you using?

Just wondering if maybe you are not on a different revision, or if you have a bad flash if others are not finding what you are.

Maybe try refreshing the latest version firmware in the AC if your are actually running the latest, or switch to the latest if you are not?
 
  • Like
Reactions: AyeYo
Rob8888, what revision firmware/software are you using?

Just wondering if maybe you are not on a different revision, or if you have a bad flash if others are not finding what you are.

Maybe try refreshing the latest version firmware in the AC if your are actually running the latest, or switch to the latest if you are not?
Thanks Brian.

Firstly, yes, I have the latest firmware installed. Secondly, one potentially important issue after reading comments above is that I'm using the controller button to fire the camera, not the app button. I notice last night that the app locks up until images are written, so potentially this is happening only when I use the controller shutter (right hand front side)?

->Perhaps the app limits taking new photos while the buffer is full, but the controller button does not? Anyone fancy trying 5xAEB shooting flat out for say 10 sets with the controller button and see if they have dropped sets on the sd card like I am?
 
Thanks Brian.

Firstly, yes, I have the latest firmware installed. Secondly, one potentially important issue after reading comments above is that I'm using the controller button to fire the camera, not the app button. I notice last night that the app locks up until images are written, so potentially this is happening only when I use the controller shutter (right hand front side)?

->Perhaps the app limits taking new photos while the buffer is full, but the controller button does not? Anyone fancy trying 5xAEB shooting flat out for say 10 sets with the controller button and see if they have dropped sets on the sd card like I am?

Nope, I've never used the app button to take a photo. Are you getting the spinning wheel over the shutter icon after taking an AEB shot?
 
Hello. I have exactly the same problem. I am trying to take 3 AEB raw photos to panorama and after 10-12 photos buffer is full and i need to wait. I also lost more than a half photos until I noticed this problem. This is really bad. I tried all 3 mavic units that I have and also bunch of different cards also most expensive one with same results. I was able to track this problem. Last complete panorama that i have is with firmware 0700. After that I was starting to loose more and more photos. They have to change something with mavic write speed as raw files have still the same size.
 
And with new 1000 firmware it is slower then 0900 so it is getting worse and worse with new firmwares.
 
And with new 1000 firmware it is slower then 0900 so it is getting worse and worse with new firmwares.
You may not appreciate my answer but I'd suggest simply shooting jpg files.
Shooting raw files when you want to combine multiple AEB shots and then blend them in a panorama is probably overkill.
I often shoot very large panoramas using jpg files and get very good results.
 
  • Like
Reactions: Pharm
Shooting some eclipse photos the other day I was shooting RAW + jpg and had the same buffer issue with just single shot photos. The audio "shutter click" is missing when you try to take a quick second shot until the spinny screen button indicator stops. I was not using the multi-shot feature just hand pressed sometimes 3-4 frames quickly. I have no idea though if it was dropping frames or even if it's possible to find out
 
At the time being write speed to cards is not that fast to shoot 3 or 5 raw photos at short intervals. I use FPV Camera to shoot panoramas (JPEGs). You can set rows and column quantity as you wish. Stitch with PTgui.
 

Members online

Forum statistics

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