Sony A1 : The One or Two Face?

I have been looking for a camera that would be a significant upgrade from my GH5M2 for some time and I have narrowed my options to two choices: the Sony A1 and the Canon R5. As the A1 underwater port system can use most of my glass I have recently acquired the A1.

Did I get the upgrade I was looking for? For photos I would say the answer has been an immediate yes due to the amazing autofocus and EVF of this camera and the burst rate. For me 15 fps is enough but the fact the A1 can trigger a flash with electronic shutter at 1/200 of a second is amazing. But I still like the mechanical shutter (unlike the Z9) and this goes to 1/400 which is also a first.

Let’s have a look at the A1 and where it stands at the end of 2022.

The Sony A1 was announced in January 2021 and was at that time the fastest, high resolution (>42 MP) full frame camera on the market. 

Capable of producing 20 fps continuous autofocus at staggering 50MP and 30 fps JPEG it still is the full frame camera producing the highest resolution at the highest frame rate as the Nikon Z9 has the same frame rate but lower sensor resolution.

As new cameras have come along we have seen some development especially on the video front and while the Sony A1 can produce ProRes RAW this is at half resolution (4230×2430) the Canon R5 is able to output 8192 x 4320.

Users have been curious on why the Canon R5 can do that while the A1 can’t and also why does the Z9 produce ProRes RAW internally at half resolution, so similar to the A1 and NRAW (that is likely not RAW) at full resolution?

Various tests on video show that both the Z9 and A1 outperform the Canon R5 in video on all formats.

Interestingly SNR improves 0.8 stops moving from 8K to 4K full frame which would not be possible if the camera was skipping pixels.

But of course neither the Z9 nor A1 can produce external ProRes RAW 8K and users have been screaming at Sony. 

Investigations

I was quite suspicious of the fact that the Z9 can only record internal raw and I have noted that Nikon has pushed back on RED lawsuit on RAW recording and therefore I believe NRAW is actually demosaiced.

I looked at dpreview studio scene and compared those cameras and in addition added also the Panasonic SR1 to check the image quality.

You can see how all cameras are affected by false color artefacts, the A1 and Z9 much more than others.

Moire is not an issue

The other suspicious fact is that the A1 and Z9 produce ProRes RAW at half resolution. How can the camera produce RAW at half resolution with no false colour artefacts if the 2×2 cell is made of different colours?

If you follow mobile phone technology, you are familiar with the super high resolution claim of certain phones, this article on Sony semiconductors web page provides an insight

The actual pixels are arranged in cells of 4 of the same colour and to produce the high resolution image the pixels are re-mosaiced, which in turn could produce artefacts. This technology has been mainstream for at least 4 years.

If you look at this video you can see that ProRes RAW video no longer produces false colour artefacts but is prone to moire as the camera does not have a low pass filter.

Dpreview studio scene provides some additional insight looking at video grabs.

No false color in 4k video

Moire in 4k video due to low resolution generating aliasing

My conclusion is that the A1 as well as the Z9 are cheating. Unlike the Canon R5 they are based on a quad-bayer sensor cell and therefore will not offer the same color resolution at 1:1 pixel of the canon R5.

It has already been proven that the A7S3 has a quad bayer cell.

Measures like DxOMark color depth do not look at color errors so this will not be spotted but I believe the remosaic of pixels of the same colour is the issue here that is showing in the dpreview studio scene.

There has been additional debate then on why the A1 defaults to APSC mode when producing 4K video, this is counterintuitive however the APSC image does not have moire nor false colour.

If we carry on with the assumption that what I have written here is correct, we can have a look at the required bandwidth to read the sensor and produce video output at various resolution and frame rates

HVDepthFPSBandwidthRGB BandwidthSubsampled
86404860123015.11654445.34963222.674816
57603240126013.43692840.31078426.873856
4320243012607.55827222.67481615.116544
384021601212011.94393635.83180823.887872
28801620121206.71846420.15539213.436928
Bandwidth Gbps for various video resolutions

Considering a readout at 12 bits we can see that the highest bandwidth is for the 8K and the 4k APSC mode as the other modes have less pixels even the 120 fps does not get to that bandwidth however required a faster sensor scan and is cropped

When the raw data goes in the image pipeline it is converted into RGB signal and here we can see that after subsampling the ASPC format has the highest data volume due to the 422 subsampling.

This in turn produces the least artefacts in fact it is quite resistant to moire as anti aliasing can be performed in camera using different techniques. So this is why APSC footage from the A1 is smoother but not necessarily sharper in fact the opposite.

A different current of thought may say no it is a full resolution classic bayer filter array which is then binned for 4k video however

  1. Such technology does not exist is not advertised there are no patents
  2. The remosaic of quad bayer sensor has been mainstream in mobile phones for years now and is done on chip

So my take is Sony is just leveraging mobile phone technology for the IMX610 in the A1 but I am open to the challenge.

For clarity as some readers seem not to understand I believe the camera has a total of 50 megapixels arranged in a quad bayer cell and goes to 12.5 in video 4K full frame combining pixels in 2×2 cells. There are phones on the market with 108 megapixels so this is nothing new.

The A1 is produced on the Exmor RS line which has been developed for mobile technology so no suprise the same investement is leveraged for cameras.

Many commercially available phones already implement the same features of the A1 see for example the specs of the Xiaomi12

Video Format Choice

The other question is then what to shoot now that we know or think we know the inner workings of the camera?

  1. 8K suffers from similar false colour artefacts of still images and 8k displays are rare it is only available in 4:2:0 subsampling due to bandwidth issues.
  2. APSC is cropped while the image has no defect this mode does not have a benefit on other cameras like the Panasonic GH6, it also does not support 120fps. Many other cameras offer cropped APSC 4k footage: you do not need an A1 if you want APSC video.
  3. UHD has moire in certain situations due to the lower resolution being out resolved by the lenses used and the lack of anti aliasing filter however it does offer the highest dynamic range and no false colour artefacts

My approach is to use UHD and if I have moire, use APSC. Moire is visible in the EVF so you can then mitigate it by switching to APSC only when required.

I have done a full analysis of the codecs and frames which I will post in a later article.

The other consideration is that I did not get a full frame camera to shoot it in APSC and in fact the A1 APSC also looks the same as my Panasonic GH5M2 and offers minimal benefit of DR and SNR due to the smaller size of the cropped area.

Now that I know (or think I know!) What may be behind the A1 limitations, am I disappointed? Actually I am not. I did not buy this camera for 8K, I have no ability to edit or display 8k but I wanted an upgrade to 4k and I can say the A1 holds video footage at 12800 ISO in slog3. I have yet to see any moire and prores raw 4k@60 is amazing quality, surely there is distortion and chromatic aberrations and vignetting but especially underwater or topside long lens this is not an issue. I am a bit disappointed by the codecs on card especially as HEVC does not have a 30 fps mode however overall the camera delivers an extremely pleasing image quality in 4K with outstanding clean colours using slog3/cine.gamut. If there is one thing that is weak is the IBIS.

If you are a purist and want the best image quality in full sensor should you look at the canon R5? This is where it gets interesting. I believe the R5 has a cleaner image however Canon is behind in terms of sensor technology so at the end when you look at real life images in terms of IQ and SNR I do not see the Canon taking an edge. What the Canon is better at is ergonomics, menu systems but not ultimately image quality despite all the things discussed here.

Conclusion

Despite all the cheating the A1 remains an amazing camera, it is small, it has many lens options and has the best underwater ports option and I do not regret my choice in fact I look forward to using this underwater. And finally, all of this just made me reflect on what a great camera the Panasonic GH5M2 is and I will keep it for some time until I am happy with all use cases.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s