MicroSD card testing for photography, 5 shot AEB

[ctp]

There are multiple threads discussing MicroSD cards and recording video. It is well established that the maximum data produced by recording video is 60Mbps or 7.5MB/s (all 4K and UHD is encoded at 60Mbps. Pretty much any Class 10 or UHS-I is capable of these speeds, (Class 10 and UHS-I are defined by a minimum sequential write speed of 10MB/s. That's been hashed to death, walk away.

Where write speed really matters is in multiple exposure pictures. A 5 shot AEB (Auto Exposure Bracketed) J+R (JPEG+RAW) photo generates roughly 150MB of data that must be recorded to the card. Write speed here is crucial because the camera locks up until it completes writing the card (the spinning blue circle around the shutter button) and prevents you from taking another photo until processing completes and the data is written to the card.

What I want to attempt is to shoot 5AEB pictures (on the bench) of a high speed timer (to measure time difference between exposures) and also measure the time we get the blue spinning circle. I will also examine the file write data on the card to measure the time lapse between writing the first and last file.

My first attempt failed (mostly in a good way) because my finger obscured the subsecond readout on the timer, which is good because it suggests that the images are captured extremely quickly and copied to the buffer. It also failed in a bad way because there is little consistency in the "blue circle", with some photos taking 8 seconds and others taking 23 seconds (same card). The third failure was using the shutter on the remote made the "5shot" sound, but only reocrded 1 photo to the card (on my second test, first recorded just fine).

For reference these are the cards I have handy measured using CrystalDiskMark 4.0.3 x64 (all figures in MB/s)

CrystalDiskMark 4.0.3 x64 - Test Sequential Q32T1, 3 Pass, 1GB
Sandisk Pixtor (Extreme Plus) 80MB/s 64G
Read 90.58
Write 64

Lexar Professional 633X 16GB
Read 87.82
Write 24.14

PNY SDXC U1 90MB/s 64GB
Read 87.78
Write 55.38

Crucial M500 1TB SSD (just for fun!)
Read 518.3
Write 439

I will attempt to pickup a Sandisk Extrteme Pro later today (Best Buy price match at $54.99) and add that to the testing.

Why post now? I need suggestions for the best ways to capture valid test data that will measure any difference in camera performance on the bench that will be affected by the write speed and see if there is a performance difference between using a 24MB/s (stock), 64MB/s (Extreme Plus), or 90MB/s card (Extreme Pro).

Thank you in advance for your suggestions.

 

[jadebox]

Perhaps you could photograph a stopwatch and take two sets of photos in a row. Reviewing the time captured in the photos would reveal the delay between the individual photos as well as between the sets.

Of course, your reaction time and speed pressing the shutter button will have someveffect. So, it isn't s perfect test method.

-- Roger

 

[loganboyd]

http://www.phantompilots.com/thread...mats-and-bitrate-data-rate.42039/#post-393019

 

[ctp]

http://www.phantompilots.com/thread...mats-and-bitrate-data-rate.42039/#post-393019

Logan, that's I specifically said this is not about video, but about 5AEB where 150MB of data is being generated in 1-2 seconds and the camera locks up while transferring the data to the card. I want to develop a repeatable test that allows us to gauge if a faster card will enable quicker second shots (5AEB) or if there is an internal limitation in the camera for how quickly data can be moved from the buffer to the card.

 

[ctp]

Perhaps you could photograph a stopwatch and take two sets of photos in a row. Reviewing the time captured in the photos would reveal the delay between the individual photos as well as between the sets.

Of course, your reaction time and speed pressing the shutter button will have someveffect. So, it isn't s perfect test method.

I like this. While the camera is writing it won't let you take a follow up shot. In theory if I just keep hitting the shutter button it will take the next picture as soon as it is possible. If that works well I can try some 7shot (non-AEB) photos, which should generate around 210MB and give us a better test.

 

[loganboyd]

Logan, that's I specifically said this is not about video, but about 5AEB where 150MB of data is being generated in 1-2 seconds and the camera locks up while transferring the data to the card. I want to develop a repeatable test that allows us to gauge if a faster card will enable quicker second shots (5AEB) or if there is an internal limitation in the camera for how quickly data can be moved from the buffer to the card.

Okay, but your throughput numbers don't show the block size used and are much different than my average write speeds for the same card type (16GB vs 64GB).
I wasn't making a point about video. The multi shot RAW bursts whether AEB or rapid fire will push the most data to the card.

 

[loganboyd]

My sample tests with a 5 Shot AEB produced 5 DNG files just over 116MB combined. let's say 1gigabit of data being written at my test speeds of 45MB/s or 360mbps will still take about 3 seconds to write to disk. You're showing 24Mb/s so it would take more like 5-6 seconds to write to disk for the Lexar 633x cards to write out a 5 Shot Raw AEB.
I actually show a 12 second gap between the 1st and last photo in a 5 shot AEB based on the image time stamps on the card from one of the tests I did.
That means it took about 12 seconds to write out the last 4 of the 5 photos taken. those 4 photos were a combined 96MB or 768megabits divided by 12 seconds is only 64mbps or 8MB/s throughput which is REALLY slow!
I'd be surprised if the limiting factor in all this is the SD Card write speed. Assuming you are using any of the top 3 or 4 SD Cards.

 

[jadebox]

Oh, yeah ... or you could just look at the timestamps on the files.

Please ignore the typos in my previous message. I am using my phone and it has decided to periodically type random things completely unrelated to my intent.

-- Roger

 

[ctp]

Okay, but your throughput numbers don't show the block size used and are much different than my average write speeds for the same card type (16GB vs 64GB).
I wasn't making a point about video. The multi shot RAW bursts whether AEB or rapid fire will push the most data to the card.

It's just a 1GB sequential read/write, I didn't run the other tests because the different block sizes are not relevant to sequential writes

Okay, but your throughput numbers don't show the block size used and are much different than my average write speeds for the same card type (16GB vs 64GB).
I wasn't making a point about video. The multi shot RAW bursts whether AEB or rapid fire will push the most data to the card.

Yeah - I didn't post all of the test results because what we really care about is the sequential write speed and I was trying to keep the chart simple.

For reference, these are the full results from CrystalDiskMark 4.0.3a x64 using an Anker USB3 card reader for the Lexar 633X 16GB card that came with my P3Pro:

-----------------------------------------------------------------------
CrystalDiskMark 4.0.3 x64 (C) 2007-2015 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 88.874 MB/s
Sequential Write (Q= 32,T= 1) : 24.194 MB/s
Random Read 4KiB (Q= 32,T= 1) : 5.626 MB/s [ 1373.5 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 1.137 MB/s [ 277.6 IOPS]
Sequential Read (T= 1) : 72.981 MB/s
Sequential Write (T= 1) : 23.697 MB/s
Random Read 4KiB (Q= 1,T= 1) : 4.230 MB/s [ 1032.7 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 1.036 MB/s [ 252.9 IOPS]

Test : 1024 MiB [G: 72.6% (10.8/14.9 GiB)] (x3)
Date : 2015/06/08 17:33:54
OS : Windows 10 Professional [10.0 Build 10130] (x64)

 

[ctp]

My sample tests with a 5 Shot AEB produced 5 DNG files just over 116MB combined. let's say 1gigabit of data being written at my test speeds of 45MB/s or 360mbps will still take about 3 seconds to write to disk. You're showing 24Mb/s so it would take more like 5-6 seconds to write to disk for the Lexar 633x cards to write out a 5 Shot Raw AEB.
I actually show a 12 second gap between the 1st and last photo in a 5 shot AEB based on the image time stamps on the card from one of the tests I did.
That means it took about 12 seconds to write out the last 4 of the 5 photos taken. those 4 photos were a combined 96MB or 768megabits divided by 12 seconds is only 64mbps or 8MB/s throughput which is REALLY slow!
I'd be surprised if the limiting factor in all this is the SD Card write speed. Assuming you are using any of the top 3 or 4 SD Cards.

Just looking at the files from the PNY U1 64GB card and I'm seeing a consistent 23 second gap between 1st and 5th on a 5AEB J+R of 140MB, which corresponds to the spinning blue circle I observed. The PNY tested at 55MB/s on sequential writes, so that performance is abysmal and points to a significant bottleneck in the camera between the buffer and the card.

 

[ctp]

Just finished 2 rounds of testing with the SanDisk Extreme Plus 80MB/s 64B card. Across 6 5AEB photos I saw
Time Stamp for file creation 12-16 second gap between 1st and 5th
Time observed on photos of 1/100 second digital stopwatch 0.2-0.3 second gap (average 25/100 second)
Time between successive 5AEB photos (blue spinning circle) 19-21 seconds (measured by photos of digital stopwatch, taking new phnoto as soon as blue circle stops spinning

Average file group (5 J+R) size is 139MB which works out to a write speed of 8.5-11.5MB/s on a card that tested at 60MB/s sequential write.

Conclusion - there is a huge bottleneck between the camera buffer and the card that severely limits writing data speed. There is no value in purchasing cards faster than Class 10 because that appears to be the maximum speed the camera is capable of writing.

Question -Is it possible that the DJI camera only has a SD bus chip that can't utilize the UHS-I protocols for faster writing?

 

[neslex]

As stated the bottleneck is within the cameras on board processor, not the microSD card. I believe dji reps have stated that the p3p has a different processor to the p3a in order to handle 4k footage, which might indicate its even slower on the p3a when processing raw multi step images.

 

[ctp]

As stated the bottleneck is within the cameras on board processor, not the microSD card. I believe dji reps have stated that the p3p has a different processor to the p3a in order to handle 4k footage, which might indicate its even slower on the p3a when processing raw multi step images.

Interesting that they confirmed that. I guess I need to go digging in the official forum? Do you have any links?

I should have clarified that all my testing was on the p3p.

 

[stevemedwin]

I took a different approach to quantifying the delay between pressing the shutter and when it is available to take another picture. I ran a test on my P3A comparing this delay using three different microSD cards. It's not an exhaustive test (see pix for details) but the results show a significant difference between the SanDisk Ultra and SanDisk ExtremePro. Interesting that the effective speed is an order of magnitude lower than the "up to 95M/sec" speed advertised on the package for the ExtremePro. My recommendation is to get the fastest card you can afford rather than the largest capacity card (which only needs to match your available flight time using all your batteries).

 

[mertyas]

I like this. While the camera is writing it won't let you take a follow up shot. In theory if I just keep hitting the shutter button it will take the next picture as soon as it is possible. If that works well I can try some 7shot (non-AEB) photos, which should generate around 210MB and give us a better test.

Derde

 

[ctp]

I took a different approach to quantifying the delay between pressing the shutter and when it is available to take another picture. I ran a test on my P3A comparing this delay using three different microSD cards. It's not an exhaustive test (see pix for details) but the results show a significant difference between the SanDisk Ultra and SanDisk ExtremePro. Interesting that the effective speed is an order of magnitude lower than the "up to 95M/sec" speed advertised on the package for the ExtremePro. My recommendation is to get the fastest card you can afford rather than the largest capacity card (which only needs to match your available flight time using all your batteries).View attachment 25839

That is some interesting test data and it's really good to see some results from the P3A to contrast with the P3P. To integrate this with other results though, we need some control data:

What Firmware and application version are you using?
Can you provide more specific model information or part numbers for each mSD card?
Can you test the in a PC with a standard tool such as CrystalDiskMark (free)?

The only card I have not specifically tested with is the SanDisk Extreme Pro because all of my testing has suggested that the bottleneck is the SD bus on the camera, not the card. If that has changed in the latest firmware I will get an Extreme Pro card for further testing.