FS7: FS7 Firmware v3 Format Tests

C

ChristopherA

Well-known member
I did some tests with my FS7 and the version 3 firmware today. I recorded in every format the camera can output, including those with the XDCA unit (when paired with an Odyssey 7Q+). I recorded three files for each format: reflective white at 61%, +2 stops, and -2 stops. This was done by setting correct exposure with the internal ND wheel at 1/4, then recording with the ND wheel clear, at 1/4, and at 1/16.


There seems to be a number of people asking questions about one format vs another. Hopefully this will provide a useful tool to answering those questions.


The total file size is 2.4 GB. In an effort not to shut down my website’s server, I’ll try to distribute this with bittorrent. I’ve not used bittorrent before, but this seems to work within my office. However, it did take about 15 minutes to make a connection and begin downloading.

https://www.dropbox.com/s/lu0hplq56218dyf/Sony FS7 Firmware v3 Camera Tests.torrent?dl=1
 
Just finished pulling my findings together.

4K RAW
When overexposed by 2 stops (i.e.- Camera is rated at ISO 500), the image isn’t that bad (other than the absurd file size). With correct exposure, the shadows pickup a significant amount of very ugly noise. At -2 Exposure, the image is covered in noise and completely unusable in any situation.


4K RAW > 4K ProRes
At all exposure levels, the shadows in 4K RAW > 4K ProRes are remarkably low noise compared to 4K RAW, but overall is pretty bad. The noise reduction in Davinci Resolve was able to remove most of it, but if you look at the flaring around the brighter exposed light, you can see that applying noise reduction destroys the flare.


One oddity. If you look at the yellow flowers, you can see some weird noise. This is present at all exposure levels. Not sure what is causing this noise, but it isn’t present in any of the other formats. This noise would prevent me from using the 4K RAW > 4K ProRes format.


4k HDMI > 4k ProRes
Appears to be a lower noise version of the 4K RAW > 4K ProRes. The shadows of the clip that is underexposed by 2 stops are dramatically cleaner.


In the over exposed clip, you can see some noise in the highlights on the yellow flowers. This is largely gone in the correctly exposed clip, and completely gone in the under exposed clip.


4K XAVC-I without internal Noise Reduction
It appears 4K XAVC-I has a little more noise than the 4k HDMI > ProRes path in all exposures, however it’s pretty close.


In addition, zooming on the flowers reveals that the 4K HDMI > ProRes contains more detail—which should be expected since ProRes has a higher data rate.


4K XAVC-L
On this still frame, the XAVC-L looks really good, especially given it’s file size. It appears to have similar noise and sharpness to the XAVC-I. However, remember that this format is only 8-bit, and I don’t have any areas with gradients in this test. Also, I’m not sure that this Long-GOP format will keep it’s sharpness if the scene has much movement in it. It is an interesting option for 4K interview material.


4k Conclusions
I still like the compressed XAVC-I for it’s remarkable quality and minimal data sizes. However, it’s clear from an image quality standpoint that the 4K HDMI is the winner.






4K RAW > HD
Similar problems to the 4K RAW > 4K Prores. Quite a bit of noise, weird sharpness artifacts. I would not use this.


XDCA ProRes and ProRes HQ
Looks very good. Low noise, good sharpness and detail. ProRes HQ had lower noise than ProRes, but it was a very minor difference.


SDI ProRes
Appears visually identical to the XDCA ProRes options.


HD XAVC-I
There is a slight softening compared to the HD ProRes options, however, it is only visible when heavily zoomed in. Appears to have less noise than the ProRes options at all exposure levels.


HD XAVC-L 50 and 35
The XAVC-L 50 is stunningly sharp and clean, given it’s small data rate. Remember that Long-GOP formats are best used when the majority of the scene is static, such as interviews. For shoots in which I was recording LOTs of interview material, I would seriously consider this format. You don’t get the benefits of 4k, but for an HD mastered project, this is the best balance of quality and file size.
At first glance, the XAVC-L 35 appears to be visually identical to the 50Mb version. However, I’d fear that this would fall apart with motion quicker than the 50Mb.


HD Conclusions
The best HD format will depend on your needs.
If I needed HD ProRes, I’d record to my Odyssey 7Q+ via SDI (not RAW). However, the only reason would be availability of media; ProRes files are large, and SSDs are quite a bit cheaper than XQD cards.
If I was shooting HD in a dynamic scene without needing to hand-off footage (to an editor demanding ProRes), I’d shoot XAVC-I. It’s sharp, clean, has small files, and won’t fall apart with motion.
If I was shooting interviews, and I had a lot of them to shoot, I would probably us the XAVC-L 50 format.




Internal Noise Reduction
The internal noise reduction appears to be exceptional. When noise reduction is set to high, there is a small drop-off in sharpness, but it’s barely noticeable unless you’re zoomed in quite a bit. The noise reduction in Davinci Resolve is better, but if I needed to hand-off footage, I would not hesitate to use this setting at it’s highest option.


2k Full Scan RAW vs 2k Center Crop RAW
It’s looking like the Center Crop RAW is inferior to the Full Scan RAW, which is quite unexpected. My center crop is showing more noise as well as significant instances of chromatic aberration. The aberration could be due to zooming my lens out to attempt to match scene composition, but this will require more testing.
I’m not seeing Center Crop dramatically improve the image in any way over 2k Full Scan, which is not at all what I was expecting.




Further tests that need to be done
How does the 4K XAVC-L stand up to a scene with a lot of gradients?
How do 4K XAVC-L, HD XAVC-L 50, and HD XAVC-L 35 stand up to detailed scenes with a good amount of motion?
Does 2k Center Crop create more chromatic aberration if I don’t change the lens zoom?
 
Last edited:
Thanks for doing & posting this. Have not looked or tried downloading your footage yet. A couple of quick points/questions:
Assuming you meant -1 on the exposure at rating EI 1000 ? -2 would be EI500 - yes?
And on the 4k HDMI - 4 ProRes - assuming you recorded via the Ody 7Q+ ?

I have never used the ext unit or shot raw - but before the V3 firmware - my limited testing of UHD XAVC-I vs ProRes came up with same conclusion as you... pretty close.
 
markfpv said:
Thanks for doing & posting this. Have not looked or tried downloading your footage yet. A couple of quick points/questions:
Assuming you meant -1 on the exposure at rating EI 1000 ? -2 would be EI500 - yes?
Click to expand...

Nope, Typo. It was exposed up and down 2 stops from the base ISO of 2000.

markfpv said:
And on the 4k HDMI - 4 ProRes - assuming you recorded via the Ody 7Q+ ?
Click to expand...

Correct.

markfpv said:
I have never used the ext unit or shot raw - but before the V3 firmware - my limited testing of UHD XAVC-I vs ProRes came up with same conclusion as you... pretty close.
Click to expand...

Don't get me wrong, the 4K ProRes did have more detail. However, for the file size, I'll still probably shoot most things in XAVC-I.
 
ChristopherA said:
Don't get me wrong, the 4K ProRes did have more detail. However, for the file size, I'll still probably shoot most things in XAVC-I.
Click to expand...
Agreed. I only tested this once on paid gig.... shooting internal as well as externally to the Pix-E5... and there is a difference... but for convenience & file sizes... I do internal.
 
I think your findings were in line with my own - less scientific - ones.

As in the fS7 is a fantastic internal recording camera and almost every other way of recording from it is a compromise of sorts.
 
+1

These observations very much coincide with mine. Thanks for taking the time to run more formalized and controlled tests.

Ned
 
I came to the same conclusion late last year and sold my 7Q+ because of it. I saw no benefits from the extra weight, expense, and drive space. The jump to the 7Q was worth it with the FS700.
 
My conclusion wasn't that there was no need for ProRes, just that in most situations you wouldn't see the difference between ProRes and XAVC-I. If you are doing closeup or tabletop work, ProRes gives you a ton more detail. Take this shot of a bouquet of flowers I shot today. When I should it in XAVC-I, most of the fine detail on the petals was lost. Only in ProRes am I able to get the high-detail look I'm trying to achieve. Note that this shot has a LUT and a little color shifting in Resolve, but no noise reduction or sharpening.
3xAm2mE.jpg
 

Attachments

  • Flower_1.8.1.jpg
    Flower_1.8.1.jpg
    59.3 KB · Views: 0
One of my co-workers suggested that Sony may have changed something in the RAW and that Convergent Design needs to adapt their Odyssey firmware. The RAW certainly seems to be a step back from the previous camera firmware.

I'm attaching a color corrected picture and a frame of RAW from my testing. In this shot, I have white exposed for 61%.

Mitch Gross—can you comment? Has Sony materially changed RAW in a way that will require a firmware update from you guys?

https://www.dropbox.com/s/7y5kqhrsxfhaqbm/4K RAW Correct Exposure.dng?dl=1

GgecMX2.jpg
 
Regarding the NR settings: if you don't want to loose detail, did you find you needed to stick to "low NR" or could you go to "mid NR" and still maintain fine detail?
 
digitalinnovations said:
Regarding the NR settings: if you don't want to loose detail, did you find you needed to stick to "low NR" or could you go to "mid NR" and still maintain fine detail?
Click to expand...

There was a slight decrease in sharpness when you enable NR, but there was no significant difference between the Low and High setting. However, the high setting did significantly decrease noise over the Low setting.
 
You must log in or register to reply here.
Back
Top