ARRI Alexa Classic & Mini LF Lab Test: Rolling Shutter, Dynamic Range and Latitude

Lab Test

by Gunther MachuApril 21^st, 2021

For quite a long time now, we were planning to get our hands on the ARRI ALEXA Classic and also ALEXA Mini LF to give them a spin in our lab. Well, the time has finally come! Curious to hear how the ARRI ALEXA Classic and Mini LF fare in our standard lab test? Then read on …

First of all, a big thank you to AV Pro, a Vienna-based rental house, who loaned us the ARRI Mini LF including SIGMA Cine primes! Secondly, another big shout out goes to Konstantin “Costa” Konstantinou as well for lending us his ARRI Alexa Classic camera!

Thirdly, I’d also like to mention NVIDIA, who gave us a RTX 3090 graphics card to test various cameras with – thank you!

Now that we speak of it… just as a quick reference: the RTX3090 cuts through the Mini LF’s open gate 4480×2520 ARRIRAW files on a 4480×2520 DaVinci Resolve 17.1 timeline like a hot knife cuts through butter. It runs at about 33% load, using about 6.6GB of memory. An absolute joy to use!

You might ask why this article is about two cameras instead of one. Well, my colleague Sebastian already tested the ALEXA Classic many moons ago, so I thought it would be good to just quickly retest this classic ARRI camera to make sure all results fall into the same place before moving over on to the ALEXA Mini LF.

This article is organized as followed: we will first have a look at the rolling shutter and dynamic range of both ARRI cameras, and then we will take a detailed look at the latitude of the ALEXA Mini LF.

Have a look how we test dynamic range using our XYLA21 stepchart and IMATEST here.

Rolling Shutter of the ARRI ALEXA Classic

The ALEXA Classic boasts a Super 35 image sensor, and allows to shoot 2048×1152 (16:9) in ProRes 4444 (ARRIRAW was not available on this camera).

The rolling shutter shows a really good value of 7.2 ms.

Rolling Shutter of the ALEXA Mini LF

With its full frame senor, the ALEXA Mini LF shows slightly higher rolling shutter values than the ALEXA Classic, 7.4 ms. This is the best result that we ever measured for a full frame sensor – the best so far was the Sony a7S III which showed 8.7ms (at 1.0x crop full frame).

Dynamic Range of the ARRI ALEXA Classic in ProRes 4444 at ISO800

Looking at the waveform plot shooting the Xyla 21 stepchart shows a strong 14 stops that can be identified above the noise floor:

Taking the Xyla21 stepchart files into IMATEST, we get the following result: the ALEXA Classic shows a strong 13.8 stops of dynamic range at a signal to noise ratio of 2, and 14.8 stops at SNR = 1. Also, 18 zones are identified by IMATEST. Those are the highest values we ever measured in our lab.

Looking at the middle graph, it is easy to see that even above the 14.8 stops (SNR = 1, blue curve) about two stops more are identified by IMATEST within the noise floor. Those can certainly be utilized using advanced noise reduction techniques in post.

The 13.8 stops at SNR = 2 are within 0.2 stops of the 14 that were measured by us more than 7 years ago. So all good on this front (thats the beauty of a standardized test setup), and we can move now to the ARRI ALEXA Mini LF.

Dynamic Range of the ARRI ALEXA Mini LF at ISO800

Let’s have a look first at the waveform plot, shot in ProRes 4444XQ, using Log C in open gate mode (4480×2520) at ISO800:

Very similar to the ALEXA Classic, a solid 14 stops are identified above the noise floor. A 15th and even a 16th stop are visible as well.

Looking at the IMATEST results below, for ProRes 4444XQ (open gate 4480×2520) we get 13.5 stops at SNR = 2, and 14.7 stops at SNR = 1. 18.8 zones are identified, and, about three more stops are identified by IMATEST within the noise floor – looking at the middle graph, above the blue line to the left (SNR = 1, showing 14.7).

Wow. Again, those are the best results we’ve ever measured in the lab for a full frame sensor. Quite interestingly, it’s a tiny bit less (at SNR = 2) than for the ALEXA Classic – it seems that the internal signal processing was changed a bit going even more towards a “raw” signal with less noise filtering. However, more zones are identified and 3 additional stops are buried in the noise floor. Hence, the Mini LF fares similar if not better than the Classic in the dynamic range department – as can be seen in the waveform plot, as well.

Shooting the Xyla21 chart using ARRIRAW open gate (4480×2520), almost exactly the same results are obtained using IMATEST:

The above results for both the ARRI ALEXA Classic and Mini LF show why ARRI has been the reference for digital cameras for more than 10 years. Not a single camera tested so far in our lab has come close to these dynamic range results.

Latitude Test with the ARRI ALEXA Mini LF at ISO800 in ARRIRAW

Latitude is the capability of a camera to retain details and colors when over- or underexposed and pushed back to base exposure. Some time ago, we have chosen an arbitrary value of 60% luma value (in the waveform) for our subjects’ faces in our standard studio scene. This CineD base exposure should help our readers to get a reference point for all the cameras tested, regardless how they distribute the code values and which LOG mode is used.

Depending on their LOG curve and code value distribution, some cameras are already quite light at 60% luma, others are darker. As an additional information, I will also give you the 18% reference grey value.

For Log C of the ALEXA Mini LF, the 18% grey comes at a code value of 400 (in a 10 bit color space). That is 3 stops below our base exposure scene.

For this test we used a SIGMA Cine Prime, 85mm T1.5. Using a shutter angle of 356° (360° setting is not possible), an aperture value of T1.5 and ISO800 and my colleague Johnnie’s face is at the cusp of clipping with max. studio light output. Brought back to our base exposure in DaVinci Resolve 17.1, using the camera raw controls and a 3D LUT (ARRI Log C to Rec709) the following image is obtained:

As can be seen above, nothing clips, also the left side of the color checker is completely intact.

From there we successively underexpose the scene by adjusting the shutter angle to 180°, 90°, 45°, 22.5° and 11.25° and closing the iris from T1.5 to T2, T2.8 and so on until T8. Those shots are then all brought back again to our CineD base exposure scene.

Now, lets jump ahead to the 3 stops underexposed scene (from the CineD base exposure scene), brought back. Here, Johnnie’s face maxes out at the 18% grey value (400 code value in 10bit). Hence, from 18% grey you have almost 8 stops (-3 to +5) of exposure latitude until the face clips (actually, a bit less as we went from 356° shutter angle to 180° and from T1.5 to T2.0 when reducing the exposure – hence about 7.8 stops to be more precise).

Impressive. And it doesn’t stop there. Looking at the 3 stops under shot, it is a bit noisy, but it is a very fine, pleasing noise:

Now, lets go further to 4 stops under, brought back:

The noise becomes more prominent at 4 stops under, brought back. The image still looks good, but if you feel you need it, noise reduction would clean up the image nicely.

At 5 stops under, brought back, noise really starts to kick in heavily:

At this point noise reduction is required. But because the noise is finely dispersed, noise reduction works quite well, as seen on the image below. I have to mention though, that shadow information starts to clip at 5 stops below our CineD base exposure – which corresponds well to the waveform shown earlier (as the shadows are another 5 stops below the face, we are 15 stops below clipping).

I had to use quite heavy noise reduction (in relation to other cameras) to get a good clean up. For your reference, here are the noise reduction settings in DaVinci Resolve 17.1:

At 6 stops below our CineD base exposure, shadow information is starting to fade, and noise heavily degrades the image. Nevertheless, no horizontal or vertical lines appear, as can be seen with so many other cameras. Astonishing.

Using heavy noise reduction still leads to a somewhat usable image, but I would say the point of no return has been reached:

Quite impressive, to say the least. For our standard studio scene, the ARRI ALEXA Mini LF shows an exposure latitude of about 10 stops (5 over, 5 under). The best cameras so far in our lab have showed 8 stops of exposure latitude (the Panasonic S5, and the Canon C300 MK III for example). All that without nasty horizontal / vertical lines, and no significant color shifts.

Summary

Not surprisingly, the ARRI ALEXA Classic and Mini LF show the best lab test results ever measured by us in all disciplines: rolling shutter, dynamic range and exposure latitude. That was to be expected of course, but considering that the ARRI ALEXA Classic sensor design is already more than 10 years old, I find it impressive that ARRI’s line of cameras still represents the benchmark in this industry for a digital cinema camera.

Have you shot with an ARRI ALEXA camera before? What are your thoughts on this timeless piece of equipment? Let us know in the comments below.