Bitrates for all settings?

[zakray]

Hi! I’m trying to find exact bitrates for each resolution+FPS combination on the M2Pro.

Would be happy to buy a coffee or a beer for anyone who can provide it

Alternatively, if anyone knows how much footage can fit on an SD card, I can calculate the bitrate from there (again, in each setting).

I’d also love to know the native ISO and the exact sensor size in millimeters if anyone has confirmed info on that. I’ve seen people say it’s ISO 100, but I wasn’t sure if people were confusing native ISO for just the setting with the least noise.

Thanks!
Zak

 

[Camino Ken]

I’ll take a crack at answering your questions but I must first confess that it has been multiple decades since I was knee deep in video compression technology.

The simple answer is the exact bit rate can’t be calculated because it is a function of the content being recorded. A video of a still scene will be significantly less than one with a lot of motion. This is because the compression algorithm mostly encodes differences from frame to frame and then periodically encodes a complete frame. Some good news is that for a given scene or video clip the data rate will track with increases on resolution and frame rate. It is important to point out that H.264 is different than H.265. H.265 is a more efficient compression but also carries the burden of 10bit vs 8 bit color depth so I suspect they might be close in bit rate.

One data point I can share with you is H.264 4K at 30 FPS is on average 12Mbyte/sec or 96Mb/s ( times 8 for bits vs bytes). I calculated this from a few videos I shot with my M2P.

Hopes this helps. Let me know if you have more questions.

 

[zakray]

Thanks Ken! I know it's impossible to get an exact bitrate, but I'd be happy with averages like the ones you posted. I'm working on app that can calculate storage requirements.

Any chance I could convince you to test all the settings on your M2P in exchange for a free copy of the app?

 

[Camino Ken]

Thanks Ken! I know it's impossible to get an exact bitrate, but I'd be happy with averages like the ones you posted. I'm working on app that can calculate storage requirements.

Any chance I could convince you to test all the settings on your M2P in exchange for a free copy of the app?

I suppose I could do the experiment but I don’t think I have a need for the app. My SD card is big enough that I never even think about available space. Just curious, what is your intended use case for the app?

I can probably get you the numbers in a couple of days. Does that work?

 

[zakray]

That'd be fine! Thank you so much!

The app is actually already available, it's a general camera app for filmmakers but people have been asking for more drones, so here I am!

CameraKit: Camera Tools

Everything you need to know about every camera worth using. "How much drive space will I need for this Alexa Mini LF shoot?” "What's the dynamic range of the Venice 2?” “How long will these batteries last me on a C300 Mk III?” "How big is this ProRes file going to be when I export?” If you've...

apps.apple.com

 

[Camino Ken]

That'd be fine! Thank you so much!

The app is actually already available, it's a general camera app for filmmakers but people have been asking for more drones, so here I am!

CameraKit: Camera Tools

Everything you need to know about every camera worth using. "How much drive space will I need for this Alexa Mini LF shoot?” "What's the dynamic range of the Venice 2?” “How long will these batteries last me on a C300 Mk III?” "How big is this ProRes file going to be when I export?” If you've...

apps.apple.com

The app looks interesting and it makes sense for some. Good luck with it.

if you don’t see something from me in two days send me a reminder.

 

[zakray]

Sounds good! Don't forget to test in both H.264 and H.265 please!

 

[Camino Ken]

Sounds good! Don't forget to test in both H.264 and H.265 please!

@zakray Unfortunately this may take a bit longer than I thought. I am busy today and tomorrow and then there is rain forecasted for this weekend so it may be early next week.

 

[zakray]

Totally fine, thanks again for the effort!

 

[Camino Ken]

Totally fine, thanks again for the effort!

@zakray

I was able to record six of the combinations yesterday and was able to confirm a few things.

H.264 and H.265 are virtually identical file sizes within 1% of each other even though H.265 has 4 times the amount of data. This is for both 4K, 30 as well as FHD. I was a little surprised by this.

Size versus frame rate tracks exactly. 60 FPS files are exactly two times bigger than 30 FPS files. I expected this.

The data does not track with ratios of total number of pixels like I thought it might. I want to look into this further. I plan to do a reference test with the drone motionless, focussed on a neutral surface in a few modes. This should give a good zero reference, with close to minimum file sizes.

i want to remind you that the biggest variable is content. The above tests were done with the drone hovering focused on a steady image for one-half minute and then slowing climbing focused on dense foliage (trees) for the second half. This should be a reasonable average.

I plan to run a second reference test to get an idea of what the maximum might be. My idea is to have the drone hover while rotating fairly rapidly so I get the maximum amount of pixel changes which should yield a file size close to the maximum.

If you are in a hurry I can give you what I have so far but I would prefer to wait until I have done the two reference tests to give me more confidence in my initial findings. Hopefully, I can complete this by the end of this week.

Any thoughts or suggestions?

Ken

 

[zakray]

H.264 and H.265 are virtually identical file sizes within 1% of each other even though H.265 has 4 times the amount of data.

I'm also a little surprised but considering H.265 is so much more efficient, perhaps I shouldn't be.

The data does not track with ratios of total number of pixels like I thought it might.

That lines up with what I've been seeing from other DJI drones... they seem to assign target bitrates of 45mbps, 60mbps, 100 mbps, etc, which don't necessarily scale up equally to the resolution. If they did, the 4K modes would produce a lot more data than they do.

i want to remind you that the biggest variable is content.

For sure, and because of that, I'm not really too bothered by what you're shooting during your tests. I'm making it clear in my app that the file sizes are just an estimate that can change based on what the user is shooting.

If you are in a hurry I can give you what I have so far but I would prefer to wait

Totally fine with me, thanks again for putting in so much effort!

 

[Camino Ken]

Hi @zakray,

I was able to finish the testing that I had planned and have compiled the results below. Another huge surprise, it turns out that the data rate has nothing to do with content. All tests with different content show no difference in file size within the margin of error. I think the reason for this is that is takes a fairly large amount of processing power to do this type of compression for content changes in real time so all pixels are handled the same and there is no intra-frame compression. The good news is that any exposure changes will have no effect.

Most of the data tracks well with frame rates and resolution with a few exceptions at the highest and lowest data rates. My best theories for these exceptions are at the lower data rates there is some minimum overhead (i.e. minimum size with no data) and at the high end it seems like more compression is happening when the data rate approaches the maximum of around 100MBits/sec.

Below is a table of the different tests I ran with the sizes in MBytes (not bits) abbreviated MB. I also included the ratio compared to the largest.

Let me know if you have any questions

Ken


All 1min duration (+/- 0.5 sec)


Average file size. 1/2 pointed at ground and 1/2 slow accent

4K, 30, H264 765.5 MB 1.00

4K, 30, H265 760.8 MB 0.99

2.7K, 30, H264 380 MB 0.49

HD, 30, H264 265.6 MB 0.35

HD, 60, H264 532.3 MB 0.70

HD, 60, H265 532.9 MB 0.70


Maximum file size. Hovering with a rapid rate yaw

HD, 30, H264, 266.5

HD, 60, H264, 532.2



Minimum file sizes. Flat white wall with stationary drone on a table

4K, 30, H264, 759.2 MB 1.00

4K, 30, H265, 761.7 MB 1.00

2.7K, 30, H264, 380.8 MB 0.50

2.7K, 60, H264, 684.8 MB 0.90

HD, 30, H264, 266.3 MB 0.35

HD, 60, H264, 531.4 MB 0.70

HD, 120, H264, 758.8 MB 1.00


‘Number of pixel for each resolution

4K 8.294 Mp 1.00

2.7K 4.064 Mp 0.49

HD 2.074 Mp 0.25

 

[WithTheBirds]

Hi! I’m trying to find exact bitrates for each resolution+FPS combination on the M2Pro.

Would be happy to buy a coffee or a beer for anyone who can provide it

Alternatively, if anyone knows how much footage can fit on an SD card, I can calculate the bitrate from there (again, in each setting).

I’d also love to know the native ISO and the exact sensor size in millimeters if anyone has confirmed info on that. I’ve seen people say it’s ISO 100, but I wasn’t sure if people were confusing native ISO for just the setting with the least noise.

Thanks!
Zak

Do you own a M2?

If you stick a blank SD card in one simply scroll through the video settings for each one and note the depicted available minutes remaining to record.

It’s a basic calc to ascertain bit rates for each setting from that point.

If you would like to include a feature in your app to calculate the number of photos you might expect to record for a given card size the DJI GO app already included this feature also.

 

[Camino Ken]

Do you own a M2?

If you stick a blank SD card in one simply scroll through the video settings for each one and note the depicted available minutes remaining to record.

It’s a basic calc to ascertain bit rates for each setting from that point.

If you would like to include a feature in your app to calculate the number of photos you might expect to record for a given card size the DJI GO app already included this feature also.

Yes I own a M2P. I was doing these tests as a favor for another member and I was also curious as to how the MPEG compression algorithm functions. I’m not sure your approach would have been any faster or easier or given me the insight I was looking for but I do appreciate your suggestion.

edit: I also did not trust the accuracy of the algorithm used to calculate remaining time.

 

[WithTheBirds]

Yes I own a M2P. I was doing these tests as a favor for another member and I was also curious as to how the MPEG compression algorithm functions. I’m not sure your approach would have been any faster or easier or given me the insight I was looking for but I do appreciate your suggestion.

Sorry Ken, there might be some confusion here- I was pretty sure you used an M2 for the test, my question was directed primarily to the OP.....

 

[Camino Ken]

Sorry Ken, there might be some confusion here- I was pretty sure you used an M2 for the test, my question was directed primarily to the OP.....

I’m pretty sure he doesn’t. I think that’s why he asked for help.

 

[zakray]

I’m pretty sure he doesn’t. I think that’s why he asked for help.

Yes, to be clear I don't own an M2 and was looking for the data so that I could add it to my calculator app.

Ken, this is exactly what I needed, and I thank you again for it. It's not surprising to me that the 4K resolutions are less than 4X the data of the HD resolutions, most camera I've studied do it that way too. Especially considering the smaller processing power of the drone.

Anyways, thanks again, and that offer for a beer via PayPal is still open if you want it!

 

[Camino Ken]

Yes, to be clear I don't own an M2 and was looking for the data so that I could add it to my calculator app.

Ken, this is exactly what I needed, and I thank you again for it. It's not surprising to me that the 4K resolutions are less than 4X the data of the HD resolutions, most camera I've studied do it that way too. Especially considering the smaller processing power of the drone.

Anyways, thanks again, and that offer for a beer via PayPal is still open if you want it!

Let’s just call it a virtual beer ....... and thanks.

I am hoping you can fill in the missing pieces of data from what is here. Just let me know if you have any questions or need further help.

 

[zakray]

The missing frame rates you mean?

 

[Camino Ken]

The missing frame rates you mean?

Yes, also some missing combinations of resolution and frame rates. You should be able to easily calculate them from the data. If you need help let me know.