HD 120 fps... Horrible !!!

[NoCoOutdoorGuys]

OK, I'll upload three clips. The one before, the jagged one, and the one after. Is there a way to tell what the frame rate of the video was? I know the jagged one is a 1080p file. But if I missed a setting or repeated one, it would be nice if I could confirm the frame rate.

Thanks.

 

[CAPTAINDRONE]

OK, I'll upload three clips. The one before, the jagged one, and the one after. Is there a way to tell what the frame rate of the video was? I know the jagged one is a 1080p file. But if I missed a setting or repeated one, it would be nice if I could confirm the frame rate.

Thanks.

Since you're using a Mac, just do the following:

Double click on one of your movie files. It will open in QUICKTIME. Now hold down the COMMAND button and press the letter "i".

A small window will open up and will show the resolution (beside the word FORMAT) and the Frames Per Second (beside FPS) and the BIT RATE (beside Data Rate).

 

[Tuna]

When you test the camera you have to test the entire integrated system. what if by somme fluke in the software the camera functions when bench tested but the moment it's airborne and being stabilized you start to get scaler issues? at that point do you just shoot when its on the ground and call it a day?

The point of testing is not to show that it works or doesn't work, it's to isolate the conditions that fail. If it works on the bench, but not in the sky, then that points to a different source of problems. If it doesn't work on the bench then the flight systems have nothing at all to do with the problems. We've seen that there are videos that show problems at 120fps, so no-one is arguing that there isn't something odd going on. The trick now is to isolate what that is so it can be fixed.

There is an awful lot of wild conjecture on this forum - it doesn't help people to understand their drones, or to see the difference between operator error, limitations of the machine and actual defects.

On the other thread (H Camera Test (now with video!!!)) I've posted an image that shows the issues are nothing to do with vibration or any of the flight systems. There's no evidence that any scaling is going on, but it does look like an encoding error that's messing up anti aliasing on the video stream. That's good news because it means the camera should be fully capable of 120fps, and the issues may be fixed by a firmware update.

 

[tinchob]

How is that? You have some strange ideas. part of what needs to be evaluated how well the gimbal is stabilizing and how that affects the image.

To properly test the camera you need to test in the environment that you will be using it. And seriously if you want to argue about this I have over 30 years experience shooting Movies and TV (and yes that includes testing gear lots of gear) and am currently the technical chairperson for one of the Socitey's.

I follow Captaindrone, if we talking about CAMERA by itself. Checking it will involve even camera and lens charts (like DSC Labs) placed at certain distances, well iluminated and all rock-solid stable.

Other stage could involve gymball and shutter/artifacts in-fly test of course, like you correctly aport here.

 

[tinchob]

The point of testing is not to show that it works or doesn't work, it's to isolate the conditions that fail. If it works on the bench, but not in the sky, then that points to a different source of problems. If it doesn't work on the bench then the flight systems have nothing at all to do with the problems. We've seen that there are videos that show problems at 120fps, so no-one is arguing that there isn't something odd going on. The trick now is to isolate what that is so it can be fixed.

There is an awful lot of wild conjecture on this forum - it doesn't help people to understand their drones, or to see the difference between operator error, limitations of the machine and actual defects.

On the other thread (H Camera Test (now with video!!!)) I've posted an image that shows the issues are nothing to do with vibration or any of the flight systems. There's no evidence that any scaling is going on, but it does look like an encoding error that's messing up anti aliasing on the video stream. That's good news because it means the camera should be fully capable of 120fps, and the issues may be fixed by a firmware update.

I´m with you. Today i´ve made a simple test to know if at 1080p 120fps the camera records actually 120 frames:

shoot-ingest-transcode to ProRes HQ (StreamClip)-import in FCP-place in the timeline-count the number of frames at 1 second intervals= 120

BG

 

[Tuna]

How do you think you're getting 1080P out of a native 4K sensor? That's right scaling.

To get 1080P out of a 4K sensor you sample - that is, you only bother reading every other pixel - read one, skip one and so on. Because you're only reading a quarter of the pixels (skipping in two dimensions means 1/2 * 1/2 = 1/4) you use a lot less bandwidth. That means you can read frames more often which is why 4K at 30fps becomes 1080P at 120fps. It's the same total bandwidth (however many millions of pixels a second read from the sensor), but split up differently.

To go the other way (ie. if you had a 1080P video source) you would scale - that is for each input pixel you would create four output pixels. On 4K TVs there will be an upscaling mode which 'fakes' a 4K picture from 1080P video. Errors in that scaling could indeed cause interesting artifacts, but it's not what we're talking here. The H's sensor is native 4K so we don't need to scale.

The thing to notice is that switching between 4K and 1080 on the H doesn't change the field of view. That means the whole sensor is being used for both video modes. If it was scaling, then you would get an effective 2x zoom when you switched modes - and that's not what happens is it? Sampling 1/4 of the pixels gives you 4x the frame rate, which is why the peak frame rate at 1080p is 4x the peak frame rate at 4K.

 

[Eric]

What memory card are you using? The write speeds can affect the ability of the camera to process high bandwidth rates like 60fps.

I've not used 120 fps yet... next flight I'll give it a try.

I se the provided card (foresee 16 go)

 

[Eric]

Here are my tests without flying - raw mode :

 

[CAPTAINDRONE]

From a photography point of view, your image (in my opinion), looks good for post production work. You have just the right amount of uniform softness in your first few samples that post production software could handle it really well to create a nice image.

Of course, as per previous videos, your 120 FPS video image looks bad.

 

[tinchob]

Video 70 Mbps in a HD 1080 120fps
I decide to make this scene (dark) trying to keep less info for compute in the chain SENSOR-PROSESSOR. In a complex subject could require even more Mbps. But at this point we already have 70Mbps !!
microSD card used = stock (FORESEE 16GB )

Could better quality SD card = more Mbps ????

Cheers!

foto grabed from MPEG Streamclip ingested RAW direct straight from card.

 

[tinchob]

Only the bit rate was 25.74Mbps

SORRY SORRY SORRY, great mistake i've made!!!!! you r right 25.74Mbps. Therefore taking in mind that minimal acceptable quality in HD 720p at 60fps (the exact half) need arround 70-80Mbps (at this sensor size)....HERE WE A PROBLEM HOUSTON!

lets wait for an FW update for increase Mbps (if the hardware we have in TH accepts)

 

[Tuna]

SORRY SORRY SORRY, great mistake i've made!!!!! you r right 25.74Mbps. Therefore taking in mind that minimal acceptable quality in HD 720p at 60fps (the exact half) need arround 70-80Mbps (at this sensor size)....HERE WE A PROBLEM HOUSTON!

I don't know where you're getting these ideas from. Bitrate is nothing to do with sensor size.

Vimeo and YouTube recommend around 20Mbps for high quality 1080p video.

GoPro maxes out at 60Mbps for the Hero 4 Black, but it's variable, so in practise you probably won't see that sort of rate (the GoPro site says 'up to' - I'll try to confirm what that means in practise tomorrow.

Depending on the codec, you don't need anywhere near 70Mbps to get high quality video.

 

[tinchob]

The higher the bitrate the more better for video, no matter how the compression is implemented. Capture rates always will be higher than distribution rates and using youtube as a guideline is folly. Blu-ray data rates are in the 35-60Mbps for 1080p.

GoPro hero 4 black running protune provides a very consistent 58-60Mbps data rate and that's what I see in all my files.

As bitrate goes up so does quality no matter what the sensor resolution.

yes, i'm with you about make YT a guide in actual rates to compare. I mean actual as the camera produce itself.

Half an hr ago, wen editing some 4k 25fps clips direct from TH card, i saw in the FCP browser under the bitrate column: 102,7 Mbps !!
keep searching the light in this clue.....

 

[Tuna]

The higher the bitrate the more better for video, no matter how the compression is implemented. Capture rates always will be higher than distribution rates and using youtube as a guideline is folly. Blu-ray data rates are in the 35-60Mbps for 1080p.

GoPro hero 4 black running protune provides a very consistent 58-60Mbps data rate and that's what I see in all my files.

As bitrate goes up so does quality no matter what the sensor resolution.

That's very misleading, and it's only encouraging people to think there is something wrong when the Yuneec uses a lower bitrate.

It very much depends on the codec. With GoPro, you can have a completely still scene and it will still use a nearly constant bitrate close to the device specification which means it's generating unnecessary data. If nothing is happening, it still uses the same high bitrate.

As an example, here's the bitrate graph for a GoPro Session video. The first half of the scene was busy - lots of movement, high contrast. The second half of the scene the lens was covered - completely black. You can see that the GoPro bitrate is more or less constant, even where there is absolutely no detail to encode.The graph stays flat.



And here is the same for the Typhoon camera (shot for slightly longer). Same scene - lots of movement, high contrast for the first half, then the lens covered for the second half. You can see that the Yuneec codec is much more variable, with the bit rate dropping hugely when there is nothing to encode. There is no loss of detail here, just smaller file sizes for the Yuneec camera. Note that even when the Yuneec is encoding a completely still scene, it's still above the base data rate for Blu-ray video.

 

[skyhawk75]

Good point Captaindrone, I agree with you totally, it is now 2016.
Me to, mostly having the drone sitting on a table when comparing different camera settings. Impossible to have the same conditions otherwise and make a good/true comparison.
120 fps is totally not usable, terrible quality. As soon I switch to 120 fps the image on the ST16 is zoomed in. Both still images and movies are recorded with max 720p the most. Very pixelated.

 

[Tuna]

No it's not misleading. not at all.

You really are being obtuse aren't you?

I'll say it one more time and then leave this to you.

You cannot compare a professional broadcast camera to a consumer device - of course the broadcast kit is better, what do you expect when you spend twenty times as much?

You cannot use absolute bit rates to measure the quality of a video stream *especially* when you have a variable bit rate encoder.

There is nothing inherently wrong with a chipset that is designed for small file sizes. That doesn't mean you're loosing any quality.

Different codecs behave differently. If I used a five year old codecs to encode a 4K video stream, it would produce a huge file. Does that mean it's better quality? No. You know that, so stop pretending otherwise.

ALL devices under $2000 as far as I'm aware compress their video. Live with it.

I've not seen any evidence yet of image quality issues with Typhoon video that are down to bitrate. It's a complete red herring.

As you're a professional, you could perhaps provide evidence that would help Yuneec improve their firmware? You know, instead of shouting about how a very expensive broadcast camera is better than a fixed lens camera with a sensor a fraction of the size?

 

[DaveIn NH]

Are you running manual settings or automatic?

If you're in full manual mode and flying with the wrong camera settings, the image is going to look dark and muddy.

If you change the camera over to full auto mode, it will determine the best settings, depending on the conditions.

 

[DaveIn NH]

My 1080p 60 looks like crap.

Are you running the camera on full auto mode, or are you running it in manual mode? It looks to me that you're running it in manual mode. The dark image and noise is caused by not enough light getting to the sensor. If you don't adjust your shutter speed, you're going to get a dark image. If you switch to automatic mode, I'll bet your image will improve.

 

[NoCoOutdoorGuys]

Are you running the camera on full auto mode, or are you running it in manual mode? It looks to me that you're running it in manual mode. The dark image and noise is caused by not enough light getting to the sensor. If you don't adjust your shutter speed, you're going to get a dark image. If you switch to automatic mode, I'll bet your image will improve.

Full auto

 

[DroneRanger]

Good point Captaindrone, I agree with you totally, it is now 2016.
Me to, mostly having the drone sitting on a table when comparing different camera settings. Impossible to have the same conditions otherwise and make a good/true comparison.
120 fps is totally not usable, terrible quality. As soon I switch to 120 fps the image on the ST16 is zoomed in. Both still images and movies are recorded with max 720p the most. Very pixelated.

My results as well. Completely unusable footage @ 120fps. Aliasing is horrible.