Welcome to PhantomPilots.com

Sign up for a weekly email of the latest drone news & information

Petition 100Mbp/s bit rate for the Phantom 3 Professional

Discussion in 'Pro/Adv Discussion' started by Matthias Heinz, Oct 15, 2015.

  1. Matthias Heinz

    Joined:
    Oct 15, 2015
    Messages:
    2
    Likes Received:
    0
  2. EPT160

    Joined:
    Sep 30, 2015
    Messages:
    92
    Likes Received:
    15
    Location:
    Australia
    Hi Matthias,

    Do you know if it is possible given the internal bus speed of the P3? I'd be interested to know.

    I find the numbers below quoted on the web link a little hard to believe.

    How is this deduced? The quoted numbers are a high end bit rate for a pro camera. Let alone one that has to fly, navigate and transmit a preview.

    From the web site:
    "Having researched the hardware it seems the video encoder is capable of a minimum of 240Mbp/s."

    The codec is capable of this and even more for sure, but not necessarily the flying P3.

    Actually I don't think you will find a discernible difference in image quality going from 60 to 100 Mbps anyway. Remember the P3 does not include any sound in that 60Mbps pipe so it is probably the same as a regular 100 Mbps H.264 with a sound channel.
     
    #2 EPT160, Oct 15, 2015
    Last edited: Oct 15, 2015
  3. Matthias Heinz

    Joined:
    Oct 15, 2015
    Messages:
    2
    Likes Received:
    0
    Hi,

    the Ambarella A9 SoC can encode H.264 codec BP/MP/HP Level 5.1, you can find this in the Ambarelle datasheet.

    BP = BaselineProfile, MP=MainProfile, HP=HighProfile
    At Level 5.1 HP the bitrate is 300Mbps, at BP/MP it´s 240Mbps, the SoC is capable to encode Level 5.1 HP.

    You need a fast SD card and a fast reader, hopefully there are no other limitations inside the camera and the reader is fast enough.

    The quality would improve greatly, fewer artifacts and more details, especially when large areas of the image are changing.
     
  4. EPT160

    Joined:
    Sep 30, 2015
    Messages:
    92
    Likes Received:
    15
    Location:
    Australia
    I think you have nailed it here, quoting you above. No issues with codec capabilities or sensor response. But internal processing and write speeds are definitely the big issue.
     
  5. nickyb

    Joined:
    Sep 13, 2015
    Messages:
    391
    Likes Received:
    97
    Location:
    Northern Hemisphere
    Await the next Phantom maybe?
     
  6. ianwood

    ianwood Taco Wrangler
    Staff Member

    Joined:
    Jan 7, 2014
    Messages:
    4,916
    Likes Received:
    1,789
    Location:
    Lost Angeles
    Nice idea but it won't happen. The $2,000+ X5 camera for the Inspire is limited to 60Mbps. If there was more headroom, they would have used it. Once the A9 starts transcoding to output HDMI at 720p as well as write to SD at 4K, it won't have enough horsepower left to go past 60Mbps.

    If you want to petition DJI, it might be more achievable to demand 2.7K. The hardware supports it even though DJI claims it is a hardware limitation.
     
  7. nhoover

    Joined:
    Dec 11, 2013
    Messages:
    851
    Likes Received:
    182
    Location:
    Los Gatos, California
    You're suggesting that a typical audio channel is something like 40 Mbps?? More like 256kbps LOL.

    100mbps would be a nice improvement but the files would be correspondingly larger. I would love the option through and would occasionally use it.
     
  8. nickyb

    Joined:
    Sep 13, 2015
    Messages:
    391
    Likes Received:
    97
    Location:
    Northern Hemisphere
  9. Shammyh

    Joined:
    Apr 4, 2015
    Messages:
    325
    Likes Received:
    154
    Something similar to this was/is being accomplished for the P2V+. Search for the increased bitrate mod on the other side for details. I can say from personal experience that it makes a BIG difference.

    I've only taken a brief pass at decompiling the P3 firmware, but until they start encrypting it next release, it's theoretically possible to do the mod ourselves.

    What makes it more difficult is that the firmware for the P3 is far more complex, and all the respective module's firmwares are rolled into one file. The FPGA driving lightbridge is by far the most complex bit, but if we could isolate the camera firmware without messing up anything else, and manage to repackage the firmware update itself, we might be in business.
     
  10. apsphoto

    Joined:
    Nov 18, 2013
    Messages:
    165
    Likes Received:
    58
    Location:
    Orange County, CA
    Their interface to the SD card is the limiting factor, using a faster card does not decrease the save time for images, so the data path is at a low rate. I suspect that it would overload the data path to have a faster bit rate. There is a lot going on for the processor.

    Alan
     
  11. Shammyh

    Joined:
    Apr 4, 2015
    Messages:
    325
    Likes Received:
    154
    Has that been tested throughly? In my brief usage of the stock card versus my higher performance 64gb card, I thought photos seemed to save faster. I don't actually recall though, and I'm glad to test at some point, unless it's already been done.

    Either way, I'd say it's unlikely that sdcard write speed is the limiting factor. In a modern SoC, pushing 100 mbit to a storage medium shouldn't be that hard, that's basic ethernet speed, half duplex. I'd say the bottleneck is more likely in some manner of processing/scaling/decoding of the camera data and the secondary output to the lightbridge/comm module. Also, to keep prices down, it's possible dji simply didn't implement (or pay to license) certain aspects of the underling components.

    Then again, it's really all speculation without some more info on exactly what's in the camera module. Unfortunately, dji physically obscures the chips. The fcc internal photos should be becoming public soon though, and often they get pre-release hardware that doesn't have the opaque black stuff covering every chip.

    What we really need is a chipworx style teardown of the Phantom 3. :)
     
  12. Shammyh

    Joined:
    Apr 4, 2015
    Messages:
    325
    Likes Received:
    154
    In fact, the FCC photos are available (as of yesterday): Link Removed. Apparently the FCC does referring checking. Just click the link below, and click "int photos".

    Full application listing here: OET List Exhibits Report

    Unfortunately, they are too low resolution to be useful... shucks.
     
    #12 Shammyh, Oct 15, 2015
    Last edited: Oct 16, 2015
  13. Chris P Duck

    Joined:
    May 29, 2015
    Messages:
    297
    Likes Received:
    133
    Location:
    Chester, UK
    I wonder if this type of petition has ever worked?

    I doubt it
     
  14. Shammyh

    Joined:
    Apr 4, 2015
    Messages:
    325
    Likes Received:
    154
    Nope, never. I'd be surprised if they even have 1 guy working full time on the P3 series software dev. R&D team has already shifted to newer products no doubt.

    If we want more features or improvements, we're going to have to do it ourselves.
     
    Chris P Duck likes this.
  15. Meta4

    Meta4 Moderator
    Staff Member

    Joined:
    Jan 8, 2014
    Messages:
    7,729
    Likes Received:
    3,450
    While you are at it, ask DJI to make the P3 fly for 45 minutes .... or ask Santa for a pony.
     
  16. ianwood

    ianwood Taco Wrangler
    Staff Member

    Joined:
    Jan 7, 2014
    Messages:
    4,916
    Likes Received:
    1,789
    Location:
    Lost Angeles
    Guys, the restriction has nothing to do with the SD card. It's the Ambarella A9 SOC that is the limitation. It can only do 60Mbps when outputting and transcoding two streams at the same time. It's a hardware limitation. It's the same limitation on the GoPro Hero 4 black which uses exactly the same SOC. If GoPro could give us more than 60Mbps, don't you think they would?

    If you need more proof that it's a hardware limitation, take another example from GoPro. GoPro had to drop the HDMI output resolution to 480p when shooting at 2.7K or higher simply because the A9 didn't have enough horsepower to output 720p or higher at the same time. Again, hardware limitation.

    So you can hack the firmware all you like but unless you overclock the A9 (good luck with cooling), you won't be getting much. If you want to hack something realistic, hack the 2.7K mode that is already there.
     
  17. shartlza

    Joined:
    Oct 21, 2013
    Messages:
    304
    Likes Received:
    5
    Location:
    Orange County, CA
    It was not really a petition, but back when the P2 was launched there was a firmware bug that was causing the P2 to not fly strait and hooking off when the craft was coming to a stop in areas with large magnetic declination. @ianwood started a campaign called "Project Drunken Sparrow" to get the issues fixed. After a month or so DJI release some beta firmware to a select few for testing to fix the compass declination error and after a few months they released a firmware update that got rid of the issue. So a petition could work, but there has to be a lot of people involved. The thing is, the difference between having the 100 Mbp/s bit rate vs. the compass declination issues is that the 100 Mbp/s bit rate is a nice to have and the compass declination issues was a flight characteristic that could cause crashes and was a major hassle to line up shots which was what the P2 was made for. Just my .02 cents.
     
  18. Mako79

    Joined:
    May 16, 2014
    Messages:
    3,413
    Likes Received:
    567
    Location:
    Sydney Australia
    Even if it could be done, there my be some physical limits also. The Pros all have a small fan attached on the gimbal board. What if increasing the Mbs meant increasing processing which therefore increases heat and therefore shorten the life.
     
  19. apsphoto

    Joined:
    Nov 18, 2013
    Messages:
    165
    Likes Received:
    58
    Location:
    Orange County, CA
    Yes it has been extensively tested, the write speed is a fixed rate. Not uncommon at all with cameras, many cameras use a slow write speed. Lots of things are possible but with the current software/hardware the write speed is limited. That is not the only limitation of the system, the video processor is another one as noted above.

    Alan
     
  20. Shammyh

    Joined:
    Apr 4, 2015
    Messages:
    325
    Likes Received:
    154
    Understand the other limitations, though given that the GoPro forces 480p secondary out versus the 720p (alleged at least) on the P3, there must be some differences. That and each implementation of the same chip tends to have slight differences.

    On further researching... if you look at the tear-down of a GP Hero 4 Black, versus the Phantom 3 Pro camera, you'll see that while they do appear to both contain an Ambarella A9 SoC, the associated supporting silicon is different. There's always some package for RAM (often combined with some NAND), but there also appear to be at least 2 additional large packages on the Phantom 3 that are not included in the GoPro. Possible DJI is cheap, and rather than uses more expensive PoP, they're using more individual chips... but also possible the differences between the two are deeper than it appears.

    As far as the sdcard performance, any links to where this has been thoroughly tested? I did a little searching, but couldn't find anyone who posted actual data on the P3 Pro/Adv. I'm out of my home country at the moment, without my birds, but can test it myself when back.

    Unless I can find that it's positively been done before, I propose to do:

    10-15 photos on each type of card, average it, and see the results.

    Testing w/ JPEG vs RAW + JPEG and some of the burst/HDR shots as well.

    I have two different Class 10 from different manufacturers, 1x UHS-1, and 1x UHS-3 cards. While I wouldn't bet money, I expect they won't all perform the same. This might be due the camera or might be due to the manufacturer of each MicroSD, but I'll do some quick stats and see if indeed higher speed cards decrease time between shots with any statistical significance.

    My understanding is that "time between shots" is dependent on how quickly the Ambarella can flush the connected NAND to the MicroSD interface. Perhaps it's fixed... or perhaps it's more opportunistic.

    The 60 mbit/s video limitation could be something further "up the chain", as @ianwood suggests, and not dependent on the ability of the SoC to flush the NAND to the MicroSD. The internal ARM cores in the Ambarella are variable clocked (though it may not be variable from actual model to model), so perhaps there's some headroom to be found. Wild speculation at best.
     
    #20 Shammyh, Oct 16, 2015
    Last edited: Oct 16, 2015