The A1 and A7S3 share the same codecs and format for (4K) video. The cameras have been around a few years however I could not find a proper analysis of the codecs. You can find the usual YouTube doing some qualitative analysis mostly lacking rigour and based on static charts which is not really sufficient.

I did this analysis on my A1 but the results hold for the A7S3.

So I decided to give it my own attention and you will get my findings in this post. I do not shoot 8K and for this format the choices are limited anyway to XAVC HS so not much to go about.

XAVC Basics

Sony has like many other brands their own codec for in camera video recording and their version is called XAVC.

You can find some more detailed information on Sony Pro Site.

Take into account that the documentation does not consider the newest XAVC HS based on HEVC coding.

In general terms you have two options:

• Intra frame -> each frame is self contained and does not depend on other frames, the clip is a collection of compressed images
• Group of Pictures (GOP) -> frames are a combination of 1 reference intraframe for each group of pictures and multiple P and B frames which means respectively Predicted and Bi-directional predicted.

More on GOP on Wikipedia.

GOP video is temporally compressed and more efficient in terms of space however it takes more computing effort to decode and encode.

Depending on the number of reference Intra frame inside a GOP you can have more or less prediction errors which manifest as noise or artifacts. So if you want to make sure that you don’t get motion artifacts you need to use an intraframe which means using more space.

A1 Codec Options

As mentioned XAVC gives you both options and the A1/A7S3 can record Intra video up to 240,250,300,600 Mbps depending on the frame rate Sony approach is 10mbps per frame.

GOP is more compressed and has 140,200,280 Mbps respectively for 24/25/30, 50/60, 100/120 frames per second. Please note higher frame rate of 100/120 fps is only supported on GOP formats as it would otherwise reach 1.2 Gbps.

Higher compression does not mean at all lower quality, analysing all codecs using ffprobe you can see that the individual Intra frames within a GOP structure are larger than the Intra frames of an All-Intra clip at higher bit rate. However due to the nature of GOP there will be potentially motion prediction errors and more noise.

A1 Sample Video

I ran some tests of a music box both static and in motion using Slog3. I warmly recommend using Slog3 for all your video as it the log compression achieves less bit utilisation and makes the video compression even more efficient.

From what I can see (there is a small focus error on the XAVC S moving part) there are minimal difference in motion handling and minimal differences on noise.

I checked each clip with Neat video using the same reference area. XAVC S and HS return the same amount of noise.

While All intra for some reason has less noise. I believe this is because the prediction introduce some small error themselves.

Considering those shots are taken at ISO 4000 there is not much to complain. ISO 4000 in Slog3 is the point where dual conversion gain goes into High gain. If you use a standard profile this would be 500. I do not recommend shooting the A1 or A7S3 in any other picture profile than slog3.

Visual Differences between Codecs

I frankly cannot see much of a difference between those codecs, I have also shot a few sequences in APSC in equivalence so at ISO 2000 and f/3.2 instead of f/4.5 and I could not see much difference except APSC has a bit more noise as expected.

Analysis

The following table shows some key data points and my observation based on a series of 50 fps video test shots with the A1.

You can see that the key difference between the codecs is the bitrate, the edit effort (how taxing is on your machine) and the noise.

What codec you will use depends mostly on practical considerations as the difference in image quality are intangible.

The HS format which is based on HEVC has a single reference frame each second which means this codec is really for static scenes as most frame are predicted. XAVC HS biggest limitation is the lack of 25/30 fps options it is a good choice for 24 fps and it does not completely stall your machine in the editing.

The S variation is good in terms of space saving as well as generally being effective for motion however it is the hardest one to edit. If space is a problem and you are happy to convert or edit natively XAVC S is a good option. If you need something that can go on for long and manages almost all situations the XAVC S should be your choice however unless you have a workstation that supports accelerated H264 10 bits 422 (and there are not many at all) you will need to convert to an intermediate codec spending more time in post processing.

Finally the S-I All intra is the one that has the least noise and is more demanding on your memory card but is the easiest one to edit and play despite the higher bit rate. If you shoot wildlife, actions, or use gimbals and your default is 50/60 frames per second, the All intra codec XAVC S-I is the best option but it does not support higher frame rate so you would need to switch to another codec for those situations and this should be the XAVC S as it manages motion better.

I have not done a test with a gimbal but those scenario where a lot of pixel move in the frame is much more taking than my music box test.

Conclusion

Although the choice of codecs is not as rich as Panasonic cameras and there are no formats other than 16:9 there are no quality issues with the A1 and A7S3 codecs as long as you shoot in Slog3.

Obviously there are options for external recording although this is not the best option for run and gun and that would give you ProRes 422 HQ or ProRes RAW with huge files easier to edit. In the future I will do a comparison between the internal intra codec and ProRes 422 as well as ProRes 422 HQ so stay tuned.

If you have any question leave a comment and I will follow up.

When you buy a new camera some time after the initial release date you are lucky to find all sorts of videos of other geeks like you that have been testing.

I was referred to this video that shows quite a few quirks of the Sony A1. Gerald has since confirmed the A1 video is binned not line skipped.

You can’t always rely on third party so here are a number of geek tests.

The first question was do you need to shoot ASPC or full frame. APSC is scaled while full frame is binned which means the first may look better than the second but will have more noise.

Pixel Binning

As usual there is no official documentation of the camera inner workings but this diagram should help explain a few things.

On the upstream the potential flow for a classic bayer filter camera to accomplish binning (it is a guess). On the bottom what a mobile phone may be doing.

In the first case binning results in a reduction of resolution and potential artifacts. In the second only a reduction of resolution.

Now one of the question is what if we just crop the sensor in APSC and then scale down will it look better?

I have done exactly this test and the answer seems to be no.

In theory the binning should look much worse but what I have seen is that moire kicks off for both at around 2x the focal length of the other. So if APSC gives moire at 50mm full frame will give it at 100mm. The full frame moire is more severe when it occurs but in most cases you cannot tell the APSC and binned UHD apart in conditions of good light.

other side by side of a dpreview sample shot cropped again there is moire in their situation on the full frame but in effect the image quality is identical.

So personally am not going to bother with APSC unless I need more magnification and I am not in low light. For all the rest I will use full frame binned.

ProRes RAW

The A1 also provides its pixel binning mode in ProRes RAW and it is identical to the internal recording when the lens corrections are off. ProRes RAW does not have a concept of lens profile so you get all the lovely defect of your lens. To my horror e-mount leses have many defects, all are distorted and have significant amount of CA.

In short unless you have a DSLR adapted lens with zero defects those aberrations are troublesome so with native glass am skipping ProRes RAW altogether.

4K with External Recorders

If you have a Ninja consistent to what Gerald Undone says you can get a scaled down version of 8K setting your HDMI output to Auto or 2160. Auto generates 4k60fps while 2160 gives you the same frame rate of your 8K.

Interestingly if you do not record to card the HDMI output goes back to what you get in internal recording. So in short you need to record 8K to card which means eventually overheating. It is unclear what subsampling is being output howeverthe image does look a bit cleaner.

8K vs 4K

There is no doubt that the 8K mode although only available up to 30 fps is superior however editing the 580 mbps HEVC files is not that easy.

I personally shoot in 4K so I am set on the full frame binned 4K but if you have the hardware to process and the screen to watch 8K is the way forward. Gerald Undone trick of the HDMI 4K while shooting also works but be careful with overheating.

Coming Soon

Next article will break down the codecs available with the A1.