EOS 5D Mark IV Poor-Quality 4K

[wildwoodphoto]

I am struggling with the quality of video in my 5DiV.

I shot a clip on what I understand is the highest quality possible: in 4k 29.97P MJPG / 1/60sec / f10 / ISO 400 / Camera Standard / 5500k on a Laowa 14mm f4 zero-d. The memory card is a SanDisk Extreme Pro 95MB/s 64GB U3 10C XC1.

The result however does not look 4k. I have uploaded the clip to YouTube here: https://youtu.be/RB-qHm0pa-Y

What are the possible reasons for the quality? 

[Waddizzle]

Laowa?  The camera only records what the lens presents to it.

You have not described what it is about the image quality you find unacceptable.  I would not consider play back of videos uploaded to YouTube to be a good metric for 4K Video IQ.

[wildwoodphoto]

As I said, the result does not look 4k quality. And yes, a Laowa: excellent quality, lightweight, zero d 14mm. If there is another way to upload the video to demonstrate the quality, please do advise.

[rs-eos]

What exactly do you mean though by 4K quality? I don't have time right now, but this afternoon I'll try to post two still frames here (including some 100 or 200% cropped portions) comparing DCI 4K between an EOS 5D IV and an EOS C70.  Both will use a 50mm f/1.2L.

On the 5D IV, DCI 4K uses a portion of the sensor (approx 1.7x crop factor).  The sensor sites are still large (5.35µ).  But smaller than the C70 (6.4µ)

The 5D IV would also not have the dynamic range of the C70; especially if you don't have the version with CLog.

In terms of uploading video, it would be best to start off with a few still frames of the original source material (then use PNG since that would be a lossless format).   Or, if you could set up something on DropBox with the original footage, that would also be good.

Again, you'll have to explain what it is about footage you find is not 4K quality.

[wildwoodphoto]

Sorry for delayed reply.

The quality looks good in the LCD, but the quality looks poor when I download the clip from the SD card.

How can I share the original file with you for you to take a look?

[rs-eos]

Please re-read my earlier reply; I provided some details on what you could do to provide both still frames or the original source material.

Without seeing exactly what is going on, we cannot address the issues.  Stating "the quality looks poor" is just way too vague.   Are you seeing color banding? Aliasing? Low dynamic range? etc.?

[wildwoodphoto]

Here's a clip:https://www.dropbox.com/sh/t6zopl0vmgry7hk/AACb5yAk10EkIa4TI9oNKTsga?dl=0

[rs-eos]

The only thing I can pick up on is that the video seems a bit soft.   Which is odd for a 14mm lens at f/10.   Not sure where the focus point was during the entire clip, but near the end it appears to have been on the curtains.  So everything on the table is out-of-focus.

Dynamic range overall isn't the best, but if you don't have the CLog version of the 5D IV, that's to be expected.

I don't have your particular lens, but on my 5D IV (also without the CLog modification), I'll get some still images on here later using a 50mm f/1.2 and 135mm f/135.  No doubt the 135mm image will be sharper as that lens is quite a bit sharper than the 50.

[rs-eos]

I botched my first test using a 50mm f/1.2 on both the 5D IV and EOS C70.  Will re-run things properly tomorrow.

Having said that, the 5D IV footage was definitely softer.

[rs-eos]

I'm not seeing an issue with video from the 5D IV.   I used an EF 50mm f/1.2 L lens at f/1.8.  This lens at that aperture is still a bit soft and it was definitely the limiting factor in my testing.

Setup:

Common items: Single 5000 K light source (continuous), 5D IV and C70 both set to 5000 K.  Lens at f/1.8

Note: You'll see slight differences in where objects are within the frame.  This is due to the 5D IV 1.7x crop factor that is applied.  And the C70 is not exactly at 1.0x crop factor with the 0.71x adapter.  I did my best to frame both videos the same.

First, here's part of still frame from the DCI 4K footage on a 5D IV.   Settings were 23.976 fps, 1/50 shutter, f/1.8, ISO 1600.  Codec: MJPEG.  Rec-709.

5D IV MJPEG DCI-4K sample (cropped)

Next, here's the 50mm lens on my C70 with the 0x71x adapter.  I wanted to keep the aperture the exact same, but due to the 2x multiplier in light gathering, I lowered the ISO.  Settings were 24 fps, 1/48 shutter, f/1.8, ISO 800.  Codec: 12-bit RAW ST.  CLog2

C70 RAW ST DCI-4K sample (cropped)

When I injested the footage into Final Cut Pro, it really crushed the blacks for the 5D IV footage, so needed to apply a 14% increase in global exposure.   For the C70 footage, I applied the "Canon Log 2/Cinema Gamut" LUT for this test.  Usually I prefer the "ARRI Log C" LUT, but the Canon LUT more closely matched the 5D IV footage.  I did some global adjustments to exposure though: lowered the shadows and raised the highlights (mids left as-is) in order to match the Rec709 footage of the 5D IV.

Now then, when I use the RF 24-70 f/2.8 L IS lens with the C70, I will end up with a decent jump in sharpness (even when the EF 50mm is at f/2.8).  This isn't surprising.

Finally some observations:

Color.  The 5D IV footage (DCI 4K) using 8-bit 4:2:2.  The C70 was 12-bit RAW, but the minimum being 4:2:2 here is good enough for this test and most video.   8 bits though is not the best though on the 5D IV.   And if you're capturing in HD, it's worse as you'll now have 8-bit 4:2:0 footage.    Still, I don't see banding in this particular test.   The 5D IV footage though contains more red.   Both the 5D IV and C70 are effectively set up to capture "Faithful" images/video.   My guess is this is just the LUT that was applied in camera on the 5D.  The C70 produced more faithful colors to what my eye sees.

Noise.  A bit unfair since the 5D IV was set to use 1600 ISO vs the C70's 800 ISO.  I don't have a plain EF to RF adapter, so had to compensate for the 0.71x adapter.   Still, it's not bad.

Aliasing. I don't see any issues with aliasing with either camera.

Dynamic Range.  I chose a low dynamic range scene on purpose.  Since the 5D IV really cannot match the C70 at all regarding video dynamic range.

In summary, I'm just not seeing any issues with your footage beyond it being a bit soft.  So ensure your focus is locked on whatever it needs to be.   Perhaps is just that in video mode with this particular lens, that things are a bit soft as well?

If you're not shooting in 4K on the 5D IV, you may want to do so to get twice as much color info recorded (8-bit 4:2:2 vs 8-bit 4:2:0).  Especially if you need to do lots of adjustments in post.  The downside to 4K on the 5D IV is the absolutely massive files.  MJPEG is really not a good codec at all for working with video.

If you need to capture high dynamic range footage, the 5D IV (or any camera/camcorder) without Log will be quite a limiting factor.

So, again, I'll ask one more time what you mean by poor quality.  Color? Noise? Dynamic Range?  Some other factor?