Hello Fellow Yuneec Pilot!
Join our free Yuneec community and remove this annoying banner!
Sign up

H Camera Test (now with video!!!)

Joined
May 13, 2016
Messages
187
Reaction score
33
Age
51
Ok, I just did a very unscientific test of my H's camera.

I recorded 20sec of footage for each quality level. The settings were natural, auto white balance, and auto exposure. I stopped and started recording between each 20secs. The 9th file down (I have to run thru the settings to figure out which setting it is) is all jagged. The rest are ok artifact-wize but I think they're washed out. That could be the natural setting.

I then played around with EV and white balance settings (didn't keep track there) and a little with shutter speed. I found that the image improved significantly with positive EV and I preferred it at the higher settings.

I'd upload footage but I'm on an old Mac with iMovie. So, if someone could suggest a way for me to combine the files without affecting (or effecting, I never know what one to use) the results I'll do it. I could upload a bunch of 20 second clips as well.
 
Here are the videos. These clips are consecutive settings just the frame rate varies. 1080p 120 fps is the bad setting.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
1440p 24fps
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
1080p 120fps
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
1080p 60 fps

I didn't even have to open the file to see the artifacts.
 
Last edited:
  • Like
Reactions: yuneec53
First let me say your camera lens appears very good in my opinion. Second your colors are perfect for post processing to achieve a good video. Third, your 1080p at 120 FPS is not at all what it should look like... there is a problem. As I noted in a previous post, this appears to be something that can be fixed with a firmware upgrade (I believe Yuneec knows about the problem).
 
  • Like
Reactions: NoCoOutdoorGuys
To my untrained eye,
The colors look great, but the focus is out.
Look at the church in the upper left corner, some of the windows on the houses.
I think this is what many of us are complaining about.
YES I know it is a small sensor camera, but it should be able to do MUCH better.
 
CAPTAIN,

Thanks for all the info you've provided in all of your posts. It's very helpful for us newbs. They only post processing I can do is the basic stuff in iMovie.
 
CAPTAIN,

Thanks for all the info you've provided in all of your posts. It's very helpful for us newbs. They only post processing I can do is the basic stuff in iMovie.
You're very welcome! I do most of my drone video editing in iMovie as well. I have Final Cut which is great for fixing and enhancing video quality. But when the quality of the video is pretty good straight out of the camera I just use iMovie as it's quick and very easy to edit a movie into something great.
 
  • Like
Reactions: NoCoOutdoorGuys
For these small thumbnail vids it all looks good to me. I'm assuming you guys enlarge them so the flaws show up. Just my untrained opinion.
 
I've taken a bit of a look. I've added an image below which is a 8 times zoom of a frame from the H's 120fps video (plant pots in our kitchen!)

I don't *think* this is a scaling problem. If you look at the vertical lines, they are smooth and pixel-for pixel appear correct. However, wherever there a horizontal edge, the video seems to get the anti-aliasing wrong, with the 'soft' part of the edge swapped over with the row above it. It almost looks like even and odd rows of the picture have been swapped. (See the top edge of the white plant pot - the anti-alias edge appears to float above the hard edge of the pot itself.)

If that's right, it's an encoding issue which should be corrected by a firmware update. We'll see.

 
That is scaling error the over lap is due to the rolling shutter which the scaler isn't correctly timing

How do you get rolling shutter errors on a still viewpoint? The camera isn't moving in the video. I get the impression you're just using words you've heard.
 
Far from using words I hear, I'm leaving that to you. Do you understand the difference between a global shutter and rolling shutter?

If the scaler is incorrectly written you can get temporal errors which look like your photo.

I started my career writing video codecs, so yes I understand. Rolling shutter artifacts only impact videos in which there is movement (as the pixels are captured in sequence, usually from top to bottom, instead of collectively). Most usually you see the results as 'tearing' when panning a camera at ninety degrees to the direction of the shutter travel. Temporal errors show up where there is a temporal component to the video stream - ie. when one frame differs from the next. The picture I posted had no temporal component - the camera was fixed and the scene contained no movement - so there shouldn't be any errors introduced due to changes in subsequent frames.

I'm not sure what you're referring to when you say 'scaler'. There's no evidence in the video file or that image that the output is scaled, though it will undoubtedly have been sampled from the larger sensor. That's typically how a higher frame rate is achieved, as the 4K @ 30fps bitrate can be progressively sampled to produce four times as many frames with one quarter the resolution.

As for your comments on the other thread, I don't believe the H was ever promoted as having a fixed bit rate. Being a professional you'll understand that a 100Mbps bitrate spec for a variable rate compressor defines only the peak (and admittedly theoretical) throughput of the system. In typical usage you see far lower rates and without detailed information about the codec, they're only going to be indicative.
 
As for the data rate please show me ANYTHING that has exceeded 50Mbps

I replied to your scaling comments on the other thread. Let's talk about bandwidth here.

For a variable rate encoder, the bandwidth will change according to how 'complex' the scene is that is being encoded. Complexity isn't necessarily about how much detail there is in a picture, but how much it changes over time. So I could show you the most detailed picture in the world, but if it is completely still then for a video I only have to send it to you once and for each subsequent frame I say 'same as last one' - bandwidth is almost zero. Video codecs are designed so that some changes are easier to compress than others. So we might say that sky turning a little more blue takes low bandwidth, but if it turned bright red we'd have to send a little more data.

People on here have complained that the chip in the H is 'designed for a security camera' which focuses on getting a picture down a network connection. That's not necessarily a bad thing. As a rule better compression takes much more calculation - codecs these days use some pretty freaky maths to compress the video stream. So if your processor is designed to be powerful enough you can take the same stream and compress it more without so much loss of quality. Better compression does not mean worse picture - all video streams are compressed (a completely uncompressed video stream is Gigabits a second, not Megabits!). However, the codec chosen and the power of the processor will decide how much the compression affects the quality of the video stream. Over the years, codecs have got better and better, and the maths has got more and more complex, and the chips needed to handle the maths have got smarter and smarter.

Now, for the H, as it's a variable rate encoder, scenes that are easy to compress will take lower bandwidth. Lots of open sky will mean smaller sizes and I've taken videos that rarely go above 20Mbps. I assume that the 100Mbps figure quoted is the maximum throughput for the processor - the point at which it would have to skip frames to keep up. As it is, you'd probably have to rig up some sort of test scene to push the video that high. It looks like typical flying footage suits the codec much better and yes, I've not seen anything that exceeds 50Mbps - but that number isn't a measure of quality. a 50Mbps stream isn't automatically worse than a 60Mbps stream. You can only decide by looking at the actual output and seeing how noticeable the compression artifacts are.

It doesn't help that the out of the box automatic settings for the camera give slightly unpredictable white point and exposure. They're not as smart as we see in some other drones. However, that's nothing to do with bandwidth or the video quality. For pro video you'd want manual control over exposure, fixed white point and the 'raw' profile to give flat, predictable video that can be properly controlled in post.
 
Great stuff Tuna.

I work in TV and will use my H to film aerial footage that, contrary to what some people here seem to think, will be fine for broadcast!

Neither this camera or the GoPro actually meet the BBC's camera requirements as their sensors aren't of sufficient size, however, they will treat the footage as non HD, and, as long as it is less than 25% of the programme duration they will allow its use.

I don't think the compression artefacts are too excessive but I do see there is quite a severe red caste on all the clips I've watched, which interestingly seems to disappear on footage where the lens has been changed.

I edit and grade my raw footage on Avid Symphony and overall I'm very happy with the results.
 
  • Like
Reactions: BrianG and yuneec53
I replied to your scaling comments on the other thread. Let's talk about bandwidth here.

For a variable rate encoder, the bandwidth will change according to how 'complex' the scene is that is being encoded. Complexity isn't necessarily about how much detail there is in a picture, but how much it changes over time. So I could show you the most detailed picture in the world, but if it is completely still then for a video I only have to send it to you once and for each subsequent frame I say 'same as last one' - bandwidth is almost zero. Video codecs are designed so that some changes are easier to compress than others. So we might say that sky turning a little more blue takes low bandwidth, but if it turned bright red we'd have to send a little more data.

People on here have complained that the chip in the H is 'designed for a security camera' which focuses on getting a picture down a network connection. That's not necessarily a bad thing. As a rule better compression takes much more calculation - codecs these days use some pretty freaky maths to compress the video stream. So if your processor is designed to be powerful enough you can take the same stream and compress it more without so much loss of quality. Better compression does not mean worse picture - all video streams are compressed (a completely uncompressed video stream is Gigabits a second, not Megabits!). However, the codec chosen and the power of the processor will decide how much the compression affects the quality of the video stream. Over the years, codecs have got better and better, and the maths has got more and more complex, and the chips needed to handle the maths have got smarter and smarter.

Now, for the H, as it's a variable rate encoder, scenes that are easy to compress will take lower bandwidth. Lots of open sky will mean smaller sizes and I've taken videos that rarely go above 20Mbps. I assume that the 100Mbps figure quoted is the maximum throughput for the processor - the point at which it would have to skip frames to keep up. As it is, you'd probably have to rig up some sort of test scene to push the video that high. It looks like typical flying footage suits the codec much better and yes, I've not seen anything that exceeds 50Mbps - but that number isn't a measure of quality. a 50Mbps stream isn't automatically worse than a 60Mbps stream. You can only decide by looking at the actual output and seeing how noticeable the compression artifacts are.

It doesn't help that the out of the box automatic settings for the camera give slightly unpredictable white point and exposure. They're not as smart as we see in some other drones. However, that's nothing to do with bandwidth or the video quality. For pro video you'd want manual control over exposure, fixed white point and the 'raw' profile to give flat, predictable video that can be properly controlled in post.
Ok everyone. I vote this post as Post O'The Month. I mean, this is why I wade through threads and read forums. Wow! :)
 
I think I understood one sentence in the whole exchange. Your right John.hey how's bout some definitions so we can understand you can get on the same page.
 
Last edited:
Great stuff Tuna.

I work in TV and will use my H to film aerial footage that, contrary to what some people here seem to think, will be fine for broadcast!

Neither this camera or the GoPro actually meet the BBC's camera requirements as their sensors aren't of sufficient size, however, they will treat the footage as non HD, and, as long as it is less than 25% of the programme duration they will allow its use.

I don't think the compression artefacts are too excessive but I do see there is quite a severe red caste on all the clips I've watched, which interestingly seems to disappear on footage where the lens has been changed.

I edit and grade my raw footage on Avid Symphony and overall I'm very happy with the results.

Thanks. The technical stuff is relatively straightforward for me - what I struggle with is the creative side - turning a very average flight into something worth watching. I'm still getting to grips with controlling the camera well enough for lens quality issues to even matter! I'm using Lightworks to edit and grade, and play with this, GoPro and Canon equipment. I think I'm getting better with each video, but know that they'll make the professionals wince. The quality of my equipment is the least of my worries. :)

I guess for the BBC they'll be wanting a big old sensor so the pixel noise is kept down. That's probably the biggest issue with these 'prosumer' drones - all the jargon about 4K and 100Mbs makes it sound like this is better than broadcast quality when in fact there is a hard limit on the quality of image you're going to get out a drone that's light and cheap enough for amateurs to play with. It's like the early days of digital cameras where each one had higher megapixel counts than the last - the numbers don't mean much but people buy them anyway.
 
Snip... It's like the early days of digital cameras where each one had higher megapixel counts than the last - the numbers don't mean much but people buy them anyway.
Good point. I used to get better pictures from a 4 megapixel camera I had (and still have it in a draw somewhere) than I now get from a 14 megapixel I now have...but my! The prices have come down!

Interesting note on these cameras: The 4 meg camera would only work with the best lithium batteries, nothing else would cut it, but the camera I now use will run on any old s**t batteries.

I'll get my coat now.
 
I was photographer on set with a BBC documentary a few years ago and they were using Canon 5D MKII cameras which only output 1080p at 38Mb/s.

Regardless, 50 seems to be the limit in the current H, half the advertised rate. No one has seen anything higher than that, and it is going to be us significantly poorer quality than 100 when you realise that the encoding and hardware are constant.

The S2 processor is designed to keep storage space requirements low, not for high quality encoding. That's what you need in security applications, forget broadcast quality, it's set up for minimum file size.
 

New Posts

Members online

Forum statistics

Threads
21,162
Messages
244,087
Members
27,882
Latest member
Speedracer2008