This is part 2 of my 2 part look at whether small cameras such as a Sony FX3 or A1 really can replace full size cinema cameras.

For this part of the article to make sense you will want to watch the YouTube clips that are linked here full screen at at the highest possible quality settings, Preferably 4K. Please don”t cheat, watch them in the order they are presented as I hope this will allow you to understand the points I am trying to make better.

Also, in the videos I have not put the different cameras that were tested side by side. You may ask why – well it’s because if you do watch a video online or a movie in a cinema you don’t see different cameras side by side on the same screen at the same time. A big point of all of this is that we are now at a place where the quality of even the smallest and cheapest  large sensor camera is likely going to be good enough to make a movie. It’s not necessarily a case of is camera A better than camera B, but the question is will the audience know or care which camera you used. There are 5 cameras and I have labelled them A through to E.

The footage presented here was captured during a workshop I did for Sony at Garage Studios in Dubai (if you need a studio space in Dubai they have some great low budget options). We weren’t doing carefully orchestrated  camera tests, but I did get the chance to quickly capture some side by side content.

So lets get into it.

THE FINAL GRADE:

In many regards I think this is the most important clip as this is how the audience would see the 5 cameras. It represents how they might look at the end of a production. I graded the cameras using ACES in DaVinci Resolve. 

Why ACES? Well, the whole point of ACES is to neutralise any specific camera “look”.  The ACES input transform takes the cameras footage and converts it to a neutral look that is meant to represent the scene as it actually was but with a film like highlight roll off added. From here the idea is that you can apply the same grade to almost any camera and the end result should look more or less the same. The look of different cameras is largely a result of differences in the electronic processing of the image in post production rather than large differences in the sensors. Most modern sensors capture a broadly similar range of colours with broadly similar dynamic range. So, provided you know the what recording levels represent what colour in the scene, it is pretty easy to make any camera look like any other, which is what ACES does.

The footage captured here was captured during a workshop, we weren’t specifically testing the different cameras in great depth. For the workshop the aim was to simply show how any of these cameras could work together. For simplicity and speed I manually set each camera to 5600K and as a result of the inevitable variations you get between different cameras, how each is calibrated and how each applies the white balance settings there were differences between in the  colour balance of each camera.

To neutralise these white balance differences the grading process started by using the colour chart to equalise the images from each camera using the “match” function in DaVinci Resolve. Then each camera has exactly the same grade applied – there are no grading differences, they are all graded in the same way.

Below are frame grabs from each camera with a slightly different grade to the video clips, again, they all look more or less the same.

The first thing to take away from all of this then is that you can make any camera look like pretty much any other and a chart such as the “color checker video” and software that can read the chart and correct the colours according to the chart makes it much easier to do this.

To allow for issues with the quality of YouTube’s encoding etc here is a 400% crop of the same clips:

What I am expecting is that most people won’t actually see a great deal of difference between any of the cameras. The cheapest camera is $6K and the most expensive $75K, yet it’s hard to tell which is which or see much difference between them. Things that do perhaps stand out initially in the zoomed in image are the softness/resolution differences between the 4K and 8K cameras, but in the first un cropped clip this difference is much harder to spot and I don’t think an audience would notice especially if the one camera is used on it’s own so the viewer has nothing to directly compare it with. It is possible that there are also small focus differences between each camera, I did try to ensure each was equally well focussed but small errors may have crept in.

WHAT HAPPENS IF WE LIFT THE SHADOWS?

OK, so lets pixel peep a bit more and artificially raise the shadows so that we can see what’s going on in the darker parts of the image.

There are differences, but again there isn’t a big difference between any of the cameras. You certainly couldn’t call them huge and in all likelihood, even if for some reason you needed to raise or lift the shadows by an unusually large amount as done here (about 2.5 stops) the difference between “best” and “worst” isn’t large enough for it to be a situation where any one of these cameras would be deemed unusable compared to the others.

SO WHY DO YOU WANT A BETTER CAMERA?

So, if we are struggling to tell the difference between a $6K camera and a $75K one why do you want a “better” camera? What are the differences and why might they matter?

When I graded the footage from these cameras in the workshop it was actually quite difficult to find a way to “break” the footage from any of them. For the majority of grading processes that I tried  they all held up really well and I’d be happy to work with any of them, even the cameras using the highly compressed internal recordings held up well. But there are differences, they are not all the same and some are easier to work with than the others. 

The two cheapest cameras were a Sony FX3 and a Sony A1. I recorded using their built in codecs, XAVC-SI in the FX3 and XAVC-HS in the A1. These are highly compressed 10 bit codecs. The other cameras were all recorded using their internal raw codecs which are either 16 bit linear or 12 bit log. At some time I really do need to do a proper comparison of the internal XAVC form the FX3 and the ProResRaw that can be recorded externally. But it is hard to do a fully meaningful test as to get the ProResRaw into Resolve requires transcoding and a lot of other awkward steps. From my own experience the difference in what you can do with XAVC v ProResRaw is very small.

One thing that happens with most highly compressed codecs such as H264 (XAVC-SI) or H265(XAVC-HS) is a loss of some very fine textural information and the image breaking up into blocks of data. But as I am showing these clips via YouTube in a compressed state I needed to find a way to illustrate the subtle differences that I see when looking at the original material. So, to show the difference between the different sensors and codecs within these camera I decided to pick a colour using the Resolve colour picker and then turn that colour into a completely different one, in this case pink.

What this allows you to see is how precisely the picked colour is recorded and it also shows up some of the macro block artefacts. Additionally it gives an indication on how fine the noise is and the textural qualities of the recording. In this case  the finer the pink “noise” the better, as this is an indication of smaller, finer textural differences in the image. These smaller textural details would be helpful if chroma keying or perhaps for some types of VFX work. It might (and say might because I’m not convinced it always will) allow you to push a very extreme grade a little bit further.

I would guess that by now you are starting to figure out which camera is which – The cameras are an FX3, A1, Burano, Venice 2 and an ArriLF.

In this test you should be able to identify the highly compressed cameras from the raw cameras. The pink areas from the raw cameras are finer and less blocky, this is a good representation of the benefit of less compression and a deeper bit depth.

But even here the difference isn’t vast. It certainly, absolutely, exists. But at the same time  you could push ANY of these cameras around in post production and if you’ve shot well none of them are going to fall apart. 

As a side note I will say that I find grading linear raw footage such as the 16 bit X-OCN from a Venice or Burano more intuitive compared to working with compressed Log. As a result I find it a bit easier to get to where I want to be with the X-OCN than the XAVC. But this doesn’t mean I can’t get to the same place with either.

RESOLUTION MATTERS.

Not only is compression important but so too is resolution. To some degree increasing the resolution can make up for a lesser bit depth.  As these camera all use bayer sensors the chroma resolution will be somewhat less than the luma resolution. A 4K sensor such as the one in the FX3 or the Arri LF will have much lower chroma resolution than the 8K A1, Burano or Venice 2. If we look at the raised shadows clip again we can see some interesting things going on the the girls hair.

If you look closely camera D has a bit of blocky chroma noise in the shadows. I suspect this might be because this is one of the 4K sensor cameras and the lower chroma resolution means the chroma noise is a bit larger.

I expect that by now you have an idea of which camera is which, but here is the big reveal: A is the FX3, B is the Venice 2, C is Burano, D is an Arri LF, and E is the Sony A1.

What can we conclude from all of this: 

There are differences between codecs. A better codec with a greater bit depth will give you  more textural information. It is not necessarily simply that raw will always be better than YUV/YCbCr but because of raws compression efficiency it is possible to have very low levels of compression and a deep bit depth. So, if you are able to record with a better codec or greater bit depth why not do so. There are some textural benefits and there will be fewer compression artefacts. BUT this doesn’t mean you can’t get a great result from XAVC or another compressed codec.

If using a bayer sensor than using a sensor with more “K” than the delivery resolution can bring textural benefits.

There are differences in the sensors, but these differences are not really as great as many might expect. In terms of DR they are all actually very close, close enough that in the real world it isn’t going to make a substantial difference. As far as your audience is concerned I doubt they would know or care. Of course we have all seen the tests where you greatly under expose a camera and then bring the footage back to normal, and these can show differences. But that’s not how we shoot things. If you are serious about getting the best image that you can, then you will light to get the contrast and exposure that you want. What isn’t in this test is rolling shutter, but generally I rarely see issues with rolling shutter these days. But if you are worried about RS, then the Venice 2 is excellent and the best of the group tested here.

Assuming you have shot well there is no reason why an audience should find the image quality from the $6K FX3 unacceptable, even on a big screen. And if you were to mix and FX3 with a Venice 2 or Burano, again if you have used each camera equally well I doubt the audience would spot the difference.

BACK TO THE BEGINNING:

So this brings me back to where I started in part 1. I believe this is the age of the small camera – or at least there is no reason why you can’t use a camera like an FX3 or an A1 to shoot a movie. While many of my readers I am sure will focus on the technical details of the image quality of camera A against camera B, in reality these days it’s much more about the ergonomics and feature set as well as lens and lighting choices.

A small camera allows you to be quick and nimble, but a bigger camera may give you a lot more monitoring options as well as other things such as genlock. And….. if you can – having a better codec doesn’t hurt. So there is no – one fits all – camera that will be the right tool for every job.