Question about recording video with VFR

[kiri]

Hi

I am trying to record a clip of my 3d printer while its printing with a high framerate (120fps) with my phone. I am doing this so i can play it back frame by frame in order to determine if the movement on the axis is smooth and consistent.

The problem is i found out that my phone, sony xperia 1 iv (and most other phones) record in VFR instead of CFR. While i know what VFR means i dont understand it enouph to determine if this will affect my tests or not.

When i play the video frame by frame i can see that between some frames the printer made bigger movements than it did between other frames which shouldnt be happening. The printer is set to move at a constant speed. But now i dont know if this is the result of a problem with the printer or an effect produced because of the VFR.

So i guess my question is does VFR affect how often the camera records frames or does it only affect how long each frame is displayed during playback?
If its the second then when i am viewing the video frame by frame manually the results would be the same regardless if the video is in VFR or CFR, am i right?

Thanks in advance for any input on this

[poisondeathray]

Hi

I am trying to record a clip of my 3d printer while its printing with a high framerate (120fps) with my phone. I am doing this so i can play it back frame by frame in order to determine if the movement on the axis is smooth and consistent.

The problem is i found out that my phone, sony xperia 1 iv (and most other phones) record in VFR instead of CFR. While i know what VFR means i dont understand it enouph to determine if this will affect my tests or not.

When i play the video frame by frame i can see that between some frames the printer made bigger movements than it did between other frames which shouldnt be happening. The printer is set to move at a constant speed. But now i dont know if this is the result of a problem with the printer or an effect produced because of the VFR.

So i guess my question is does VFR affect how often the camera records frames or does it only affect how long each frame is displayed during playback?
If its the second then when i am viewing the video frame by frame manually the results would be the same regardless if the video is in VFR or CFR, am i right?

Thanks in advance for any input on this

Phone recording VFR means frames are dropped. So you have less than 120 FPS . Certain frames are displayed for longer periods of time to keep sync

Standard VFR to CFR conversions insert duplicate frames where the frames were originally dropped (not recorded) to make up the 120FPS CFR - so there will be no real difference in real data when you go frame by frame.

You can interpolate during those periods of dropped frame to make an educated guess - but you cannot determine anything definitively with your recording because you have missing data

[kiri]

I was afraid of this!

Just to sure that i understand you, based on what you are saying the reason that i see bigger movements between some frames in the video is probably because the phone dropped a frame (or more) in between those frames while it was recording so in reality more time has passed between those 2 frames than when compared with others?

[poisondeathray]

Just to sure that i understand you, based on what you are saying the reason that i see bigger movements between some frames in the video is probably because the phone dropped a frame (or more) in between those frames while it was recording so in reality more time has passed between those 2 frames than when compared with others?

Yes - and you can look at the timestamps to verify this

The VFR timestamps display that last recorded frame longer to make up for what would have been a recorded frame. That keeps everything in AV sync

So if you go frame by frame in a VFR recording, it appears like a jump in time . A VFR to CFR recording (using standard methods) inserts duplicate frames instead of that jump when you go frame by frame. The inserted duplicates are placeholders for what should have been the real frames that were dropped

[pandy]

Search for some timer able to display time (ticks) with ms accuracy - place it in camera view - later you can use it as reference stamp so you will be able to quantitatively describe time accuracy. Without this you can do only guess.

[kiri]

...A VFR to CFR recording (using standard methods) inserts duplicate frames instead of that jump when you go frame by frame. The inserted duplicates are placeholders for what should have been the real frames that were dropped

Now that i think about it, its not as bad as i initially thought. I can still definitively determine if the printer moves at a constant speed and for the most part i can check for any jerkiness in its movement.

I convert my video to CFR so that it inserts the duplicate frames and then i go frame by frame and every time i spot a duplicate frame i will know that the printer extruder should have moved exactly as much as it did in the previous frame. When i get to the next non duplicate frame i will still be able to tell if it moved as much as it should have or not.

This will work as long as the phone always records exactly 120fps and all the missing frames are replaced by duplicates.

Media info shows the following about my recording:
Frame rate : 119.963 FPS
Minimum frame rate : 118.421 FPS
Maximum frame rate : 120.000 FPS
Original frame rate : 120.000 FPS
So this looks OK.

But i should mention that a previous recording that i made (not the one that i am using now) had this:
Frame rate : 119.813 FPS
Minimum frame rate : 30.000 FPS
Maximum frame rate : 121.294 FPS
Original frame rate : 120.000 FPS

That 121.294 kinda worries me, i understand dropping frames but why such a huge difference between min/max and would it ever go above 120?

Either way i believe you covered me as far as info on VFR recordings goes. Thank you very much!

I will do some tests and see what comes out.
If you think of anything else that might be relevant here let me know.

[kiri]

Search for some timer able to display time (ticks) with ms accuracy - place it in camera view - later you can use it as reference stamp so you will be able to quantitatively describe time accuracy. Without this you can do only guess.

Thanks for the idea. If nothing else works i might do it. The main problem with it is that i dont think i can find anything like that locally so i will have to order it and wait about 2 weeks before i can continue that's why i leave it as a last resort.

[poisondeathray]

Now that i think about it, its not as bad as i initially thought. I can still definitively determine if the printer moves at a constant speed and for the most part i can check for any jerkiness in its movement.

I convert my video to CFR so that it inserts the duplicate frames and then i go frame by frame and every time i spot a duplicate frame i will know that the printer extruder should have moved exactly as much as it did in the previous frame. When i get to the next non duplicate frame i will still be able to tell if it moved as much as it should have or not.

For single drop pattern it should be ok .

Yes you can make some assumptions , and for certain types of motion they might be valid. A mechanical printer is probably ok.

But non linear motion types , start/stop, organic motions, acceleration/deceleration, sports, nature etc.. - that might be problematic , especially when you have multiple consecutive drops

This will work as long as the phone always records exactly 120fps and all the missing frames are replaced by duplicates.

Media info shows the following about my recording:
Frame rate : 119.963 FPS
Minimum frame rate : 118.421 FPS
Maximum frame rate : 120.000 FPS
Original frame rate : 120.000 FPS
So this looks OK.

Yes, this is nearly CFR with just a few drops, probably never more than strings of 1 drop. The min value is important

But i should mention that a previous recording that i made (not the one that i am using now) had this:
Frame rate : 119.813 FPS
Minimum frame rate : 30.000 FPS
Maximum frame rate : 121.294 FPS
Original frame rate : 120.000 FPS

That 121.294 kinda worries me, i understand dropping frames but why such a huge difference between min/max and would it ever go above 120?

The min value is what should worry you, 30 FPS means there are sections with multiple consec drops. Strings of multiple drops ware worrisome - do you assume constant linear motion ? It might be a valid assumption for printer motion

The max value is an overshoot artifact - the timestamp delta between some frames is slightly less than 8.333ms. It's not super accurate, sometimes a recording will behave like this to "catch up" after a string of drops