Monday, February 24, 2014

Alexa v.s. Sony F55 on image quality

Today I'm posting a new camera test between the Alexa and the Sony F55. I have a client who loves the look of the Alexa, which is understandable because it is such a fabulous camera! But I wondered if I could achieve very similar image quality on the Sony F55 for a lower cost. But this has been kind of hard to prove since it's very hard to tell unless you have the cameras side by side. I know there are already a number of side-by-side tests out there with these two cameras, like the Zacuto Camera Shootout, which I highly recommend. But there is nothing like doing your own test, because you can actually test the footage with your own parameters and with your own post workflow, and really see the attributes of the cameras.

I was fortunate enough to have help from Bryan Mayo, owner of MP&E Dallas, and his staff, who provided the Alexa and a couple of Cooke S4i lenses, and Ted Barnett, owner of Lone Wolf Media, who provided the Sony F55. I also had fantastic help from the talented actors at Linda McAlister Talent Agency who came out to the test shoot to be on camera.

I wanted to set up three lighting scenarios. The first one is a high contrast scene with very little fill light, only a bounce card.

This was to see how much noise in the shadows was present and also see some example of dynamic range in a high contrast ratio situation. I used a tungsten Kino Diva for a key and a mostly daylight balanced (3 tubes 5500k, 1 tube 3200k) Diva kicker and then balanced the cameras to 3200k. Both cameras shot in Log mode. Log-C for the Alexa and SLog-2 for the F55. I didn't care about matching exact T-stop because I would have needed a handful of ND filters for Alexa, since it doesn't have roll-in ND on the camera the way the F55 does. So you will see a wider depth of field with the Alexa in the first scene.

We also shot the scenes with S-log-3 but noticed considerable noise with this relatively new log from Sony, so I am not sharing that result here. I wanted to see the optimum quality result of each camera.

The second scene was a closer key/fill ratio but with the same two sources in place. Also brighter background. It was interesting to see where colors stayed the same and where they did not. Notice the difference in the intensity of the blue kicker on the F55 in both scenes 1 and 2.

Because my client's post house prefers to work with ProRes files, I sent the F55 image to a Ki Pro Mini via the SDI cable with no LUT, so the Sony F55 ProRes had the same SLog-2 attributes applied. I was recording ProRes 422 in the Ki Pro, and ProRes 422HQ in the Alexa. Captured a higher data rate than the F55's native XAVC codec, which I believe was around 80 mbps at 23.976.
Part of my test was to color correct using basic native plug-ins in Premiere Pro CS6. Certainly superior results could be achieved using DaVinci Resolve or even Speedgrade, but part of my client's concern for the F55 was that it might require expensive color grading time to achieve Alexa level imagery. I think it's good to know if log footage can be corrected with basic native plug-ins for speed and reduced post expense. So these stills were grabbed from color corrected video within Premiere Pro CC, but then for optimum results I used Speedgrade to grade the final video that plays on YouTube. The link is at the bottom of this blog.
The third scene was the only exterior. I chose to shoot this in the shade of the MP&E building near the end of the day. This meant the shade was probably more around 7000K or higher, much cooler than the 5600K preset white balance limit of both cameras in Log mode. So seeing how well the footage was brought back toward warm became part of the test.
As with log files, color correction was mostly a matter of increasing contrast, and except for the exterior, I didn't want to influence the natural color. Within Premiere and Speedgrade I had to use some Color Balance on the exterior shots to shift the image off the cooler cast the shade created. My aim was to match the cameras in contrast and saturation without changing the sensor's natural rendering of skin. And it's worth mentioning that I used an HP Dream Color monitor in Rec.709 mode to achieve a grade that would more closely meet broadcast standards. What you might see on your monitor may vary.

In Conclusion: I was rather pleased to see the F55 hold it's own. Using the exact same lens and lighting situation really showed me that the F55 can have a comparable image to the Alexa under favorable lighting conditions. Both cameras show off their dynamic range with the edge leaning slightly toward the Alexa. For instance, I could see easier into the fill side of the first scene. And the Alexa was more gentle, recording the 5000K kicker than the F55. In fact I feel the F55 tends to grab hold of a skin tone and then exaggerate it so faces tend to get spotchy in places compared to the smoother overall skin rendering from the Alexa.

The Alexa leaned consistently toward a yellowish hue but was slightly less noisy in the shadows. Of course, I've left some quality on the table with both cameras. I could have shot ProRes 444 HQ with the Alexa and SR444 with the F55. The SR being more limitedly accepted in NLEs at the moment, gives the edge to Alexa, but there are always trade-offs.

There is great value in the Sony F55. It is quite capable of producing a nice image and has lots of "look" flexibility. 4K capable, global shutter, s-gamut, multiple codec options, more lens options, lighter, less power hungry and best of all, much less expensive to own, rent, or insure. But in the end the Alexa still produced the best, more cinematic image, which is my preference all the time. So I still had to put my money down on the up-coming Amira (the baby Alexa) which I hope to take delivery by September 2014.

Here's the link to the YouTube video of the camera test: https://www.youtube.com/watch?v=rQM1WPyQ60o&feature=youtu.be

Feel free to leave your thoughts. Thanks for your interest.

3 comments:

  1. Actually I think your test is (as most are) flawed in different aspects.
    I would really prefer people to match not only the exact lens/stop but also crop to the same aspect as well as match even the angle of view (compensate for different sensor width with camera positioning). I think it is important because lenses perform VERY differently with wide open apertures and you can easily see here that that is swallowing up most of f55s 4k capability.
    I also think that using a lower codec is unfair because xcvc is a L-GOP codec and therefore more efficient than ProRes (I-Frame only). So I would have liked to see at least same data rate (I assumed you shot HD so same is fine as opposed to higher data rate for higher resolution).
    Watching the test I feel the ALEXA performed better in the end (skin and hair detail) but the lens is a lot softer on f55. Also uploading to YT the ALEXA should perform better after compression because it has a noisier sensor which sometimes translates into getting compressed nicer (a bit contradictory it seems but adding some noise on sony stuff sometimes really increases the quality on YT).
    Altogether I'm pretty torn on which camera is superior (leaning f55 but alexa sure is a mainstay for the big boys so most DPs like it).
    It's really hard to find camera tests that are truly flawless, I guess I will have to test for myself. That said I really liked this one and I think it really gives you some idea. Also I think it's really nicely matched.

    ReplyDelete
    Replies
    1. Thanks, Simon, for taking a look at this old test. You are correct that it is flawed, as many tests are. I did use the same lens on both cameras, but couldn't match T-stop due to lack of ND for the Alexa, but did shoot within the sweeter spot of the lens for both cameras. Neither were wide open. I did record ProRes from the F55 via SDI to KiPro, but just 422, not 422HQ like the Alexa had. Can't remember why we didn't use 422HQ in the KiPro. You are right that we should have used the same datat rate. Using the same lens was more important to me than matching crop factor perfectly, and seeing what the crop difference, was part of the test as well. Both sensors are close to the same size. Not sure of the pixel count though. There are so many variables when doing tests, but so little time. My test told me what I wanted to know and certainly shooters should run their own tests for what they want to know.
      I can tell you since that test, I have purchased and been using the Arri Amira, and happy I did. I got to shoot under same lighting and white balance conditions with the F55 alongside on a Dallas Cowboys game and it was interesting to see more differences. The F55 skin tone was more red and saturated while the Amira was more natural. Different lenses were used but I don't think color rendering was lens dependent. I'm pretty convinced of the Arri quality over F55, but not to say the F55 can't produce good results. And the Arri is easier to use. Thanks, again, Simon for your comments. Alan

      Delete