Can you pick the camera and lens ? Venice / DXL2 / Alexa 65 and lenses

J

John Brawley

Well-known member
As part of pre for my about to be dropped on Hulu series The Great, I did a bunch of large format camera tests with some of the lenses I was interested in looking at.

Cameras are DXL2, Alexa 65 and Venice.

Lenses were

Cooke S7
Primo 70's
Arri DNA
TLS Vega
Panavision H series
Zeiss Super Speeds
Canon K35


Here's the little blind compilation of them to look at.
https://johnbrawley.wordpress.com/2020/04/12/dealers-choice-venice-dxl2-alexa-65/

I always like testing this way for the stakeholders on the show to look at. The vimeo that links through the above link is also downloadable if you want to scrutinise.

I also have made it possible to download some of the camera original files. Be warned though it's 80Gb ! Thanks to my friends at Digital Pigeon who graciously make it possible to host a massive collection of files like this.

You can source the camera original files though this link. If you go this way they are labeled by camera and lens so it's cheating but I forgive you. Show me what you can grade !
https://johnbrawley.wordpress.com/2020/04/14/dxl2-venice-alexa-65-files/

Feel free to post what you see and what you like or if you have any questions....

Enjoy.

JB
 
Last edited:
Lazy or efficient..? they all look fine to me :)

Only things I noticed were..
-with the CU of the candle sometimes she was too out of focus - I wanted more DOF
-her skin was not perfect and that was probably brought out by good kit (not to dis' her)
 
Fantastic test JB! Thank you for sharing. Looking forward to jumping into this!
 
Actually, just for fun, going to throw in my guesses based on 30seconds of watching on my cellphone in a bright room. As, i am sure that is part of the idea in blind comparisons. Sure they’re different, but how much effort does it take to tell them apart? I already spent over 45min on that Red blind test, when it came out. Gotta switch it up. Though, i’ll definitely spent some time with your wonderfully crafted test too! Thanks so much for sharing.

A: DXL2
B: Arri
C: Venice
 
I have no experience with most of those lenses, so i couldn’t even begin to guess. What i intend to do is study your test and kind of break down my observations on them completely blind. So, less of a guess and more about picking out characterisitcs and discussing those. As i think hearing your feedback on everything will eventually be the most valuable.
 
I've been waiting for a new monitor, but the process is taking longer than I though. was going to get some basic calibrating done and download the file etc.


Surprised no one else is looking at these lenses! helpful any time, but very helpful during lock-down.

Thanks again!
 
Great tests JB, thanks for sharing!

Here are my thoughts based on the blind test:

Cameras:

Camera A: Harsh + Clips unpleasantly
Camera B: Clips okay, bit blocky in the shadows, highlights are lovely
Camera C: Clips beautifully, most uniform and pleasing colour, smoothest noise


Camera/Lens Combos:

Cam B Lens C:
Close = +
Wide = +
(nice vintage look, I like the catseye, almost anamorphic bokeh)

Cam C Lens F:
Close = +
Wide = +
Flares = ++

Cam C Lens G:
Close = ++
Wide = -
Flares = ++
(too fringey/busy OOF in the wide shots, close-ups are lovely though)

Cam C Lens A:
Close = +++
Wide = +++
(smooth, lovely, and has NOTHING that distracts from the subject)

Cam A Lens A:
Close = -
Wide = -
(too pink, harsh clipping - the camera's response ruins the lens' evaluation here)

Cam C Lens E:
Close = +
Wide = ++
(pleasantly vintage edges in OOF)

Cam B Lens D:
Close = -
Wide = -
(too much CA/fringing, but nice smooth OOF)
 
Last edited:
Camera A - Was crap in comparsion to B and C.

I prefer C. I think both B and C would be OK for me but the A camera was a no go. If my older bias is correct then A is RED/DXL2 and B Venice and C Arri.

Will look later more for the actual lenses, not that I have worked with any of them except the Super-speeds so I can't tell the others.

Anyways, great test! Really love having the light ramping up from candlelight to "bright sun" to really show how they roll between gradations, etc...
 
I'd suggest everyone take a look at the camera original files JB's provided, because this test does demonstrate something you have to always be very careful with - your post workflow.

Camera A (which I think everyone will agree, looks awful in this test) is the DXL2. I was a little shocked by how unusable it looks, because I've seen the camera in action, and it can make stunning images. However, pushed through the "default" settings in Resolve, it looks godawful.

However, simply adjust the raw settings to Red's IPP2 colour, widegamut colour space, and Log3G10 gamma curve, then drop Light Iron's DXL LUT on top of it, and you get precisely the gorgeous images you'd expect:

tyMvBsN.jpg


wEdmv2A.jpg
 
Another interesting little discovery. I decided to look into the cameras' latitude a bit, and tried a 1-stop and a 2-stop push on each of them, with the minimum amount of noise reduction needed to get them presentable again (i.e. the workflow I'd use if I needed to recover some more info in the shadows before that key light comes up).

And to my considerable surprise (give the Alexa's enormous sensor and Sony's newer sensor tech), it's actually the DXL2 that seems to hold up best in a 2-stop push (I'm guessing the slightly finer details from the 8k sensor, compared to 6k in the other two, is what gives it the edge). Now all three do extremely well (just to be clear), but I'd give the Panavision the edge, followed really closely by the Alexa65, and then the Sony's a fair bit off - it's considerably harder to recover from a 2-stop push compared to the others (as there's fixed-pattern noise when dig right deep).
 
Thank you John for the effort to do the test and your willingness to share the results.

Have not looked at 'what's what' yet. It's pretty hard to keep track of all the configurations, but I found it interesting to see that the character of each camera (or it's performance/color science) punched through well enough despite all the various lenses.

All remarks strictly from the video without looking at, or regrading, original files. Looking at the downloaded 'original Vimeo file' on an uncalibrated iMac 5K.

I liked camera C in this test. I find. its shadows to look pretty naturally subdued and I like the response as the light ramps up. I like camera C best both in the shadows and the highlights. It seems to have a natural reaction to light and leaves no signs of 'process'.

Camera B comes in pretty saturated in the shadows. The darker images register as a mismatch with me somehow. Not sure murky is the right words with darker but saturated images, but I'm not a fan. Grade comes into play, obviously. But I'm guessing we're looking at some sort of manufacturer standard conversion here and not graded to taste?
I still don't like camera B much better as the light ramps up. The transition on the face seems 'flattish' and I don't love the highlights, but it should be said that there is no direct harshness or strange leftover saturation. The highlights look pretty clean.

Camera A also looks too saturated in the shadows. As the light raps up, images take on a harshness and look busy. A few of the clips leaves me with a feeling of too much saturation in the highlights and leaves me with that taste of 'process' that make the camera system 'opaque' instead of transparent.

Not done with the lens evaluation, but I notice that I'm I'm not liking when lenses add too much to the images, as I find it distracts. Some might call it character or whatever. I know some people just can't get enough of flares, halos and whatnot. I think it can be beautiful when a lens renders a light flare in a camera move—and it can be all kinds of colors for all I care—but when way off axis lights "wave hello", I find it too much.

Again, starting fresh with the original files, I guess one could adjust them to one's own particular preferences. They are obviously capable camera systems.

Eric Coughlin said:
And after all that testing you decided to go with the Blackmagic Pocket Camera 4K.
Click to expand...

Pretty much. But the P6K.

Nothing about this test makes me even a wee bit sad or longing for more in terms of pure image quality. I'm not comparing the P6K to these cameras, but I do find that the delta you get in IQ does not warrant getting in debt in tens of thousands of $$. Upcoming filmmakers today are pretty lucky when it comes to options for < $5K.
 
Grug said:
Another interesting little discovery. I decided to look into the cameras' latitude a bit, and tried a 1-stop and a 2-stop push on each of them, with the minimum amount of noise reduction needed to get them presentable again (i.e. the workflow I'd use if I needed to recover some more info in the shadows before that key light comes up).

And to my considerable surprise (give the Alexa's enormous sensor and Sony's newer sensor tech), it's actually the DXL2 that seems to hold up best in a 2-stop push (I'm guessing the slightly finer details from the 8k sensor, compared to 6k in the other two, is what gives it the edge). Now all three do extremely well (just to be clear), but I'd give the Panavision the edge, followed really closely by the Alexa65, and then the Sony's a fair bit off - it's considerably harder to recover from a 2-stop push compared to the others (as there's fixed-pattern noise when dig right deep).
Click to expand...

Great you brought this up !

I liked DXL a lot but I did find in my “stress” test the only way to get the most from the image was to animate the grade.

So in “under” when she’s walking with the candle you can get it looking nice without a correction but as soon as you go a bit over (when the light sums up) then you end up with a not so nice image.

All the grades were designed to be a middle grade for me. I don’t work with DITs and often my footage has to go through a dailies process that may or may not have someone making those corrections. So my “grade” is an attempt to emulate that one size fits all.

I found the DXL was the hardest to create a single look from. Maybe that means there’s so much more to work with it can’t all be contained ? Who knows.

And I’m unsure of the p4k sledging. It wasn’t tested and wasn’t even used in the show ?

JB
 
So here you go.

In my test the cameras were


A - DXL2
B - Alexa 65
C - Sony Venice @ 500 ISO
D - Sony Venice @ 2500 ISO (Not shown I think)

ARRI FSND's were used out front to match exposures.


Lenses were

A - Primo 70
B - Cooke S7
C - Arri DNA
D - TLS Vega
E - Panavision H Series
F - K35
D - Zeiss SUperspeed.

Interesting that many seemed to like Venice overall most. As Greg pointed out, the "look" on this more of less the stock 709 look with a minor tweak (a single node in Resolve) and rendered out.

In my own evaluation I did a lot more pushing and pulling of the files to see where they could be taken.

I found that the DXL2 wasn't as nice under candle light and very low WB lighting. Once it got into daylight white balanced lighting it becomes spectacular.

But I was shooting a show that would have a lot of candle light and flames.

DXL2 colour in those scenarios were kind of yellow skewed in a way I didn't like very much. Yes, I COULD correct that out of the DXL2, but it was an extra step. Panavision also offered a slightly different OLPF for me that was IR tweaked that's possible better under firelight but it was a bit late by the time I heard about that.

The Sony was really nice. I found that it had, like I remember on the F55 a kind of glossy in the blacks silky feel. I tried both ISO's and found that although the 2500 ISO was nice, that it didn't offer me much of an advantage in shooting candle light, as it has LESS DR than at 500.

The thing is, though there are low light levels, shooting by candle light, the candle flames themselves are also in shot. The HDR delivery requirement of this meant I wanted to try and PRESERVE colour in the flames as they approach clipping.

So by having in effect a top down exposure based on the flame brightness, less DR doesn't help in that situation and neither does 2500 ISO. Candles in shot put's a pin in where your exposure is anyway and then DR becomes the most important thing, not crazy high ISOs.

I think I over all prefered the sony colour and look, however, the noise once lifted, like I also recall from the F55 isn't very nice. So as long as the blacks are black you're fine, but as soon as you start draggin exposure in near blacks up from there, the noise that comes up isn't as "nice looking" as the Alexa noise.

I think that's Alexa's secret weapon. The noise on that sensor is never really offensive once you start seeing it and I think is also why you get so much more apparent DR than their own spec says.

So imagining a flame lit scenario, where I'm imagining digging into the blacks a lot, the Alexa made the most sense there. In terms of grade, the Sony and the ALexa just "fell in" according to the colourist, who felt the DXL could be made to match and get there, but it always seemed to need a bit more wrangling. (This is a very experienced senior colorist as a major past facility in London)

JB
 
John Brawley said:
The thing is, though there are low light levels, shooting by candle light, the candle flames themselves are also in shot. The HDR delivery requirement of this meant I wanted to try and PRESERVE colour in the flames as they approach clipping.
JB
Click to expand...
This is what I thought you did, like, Masterfully well! The exposure on the Sony was hanging in their just at the invisible cusp. Seemed like you completely nailed the contrast ratio in that scene, as if the candle was a breathe of light more and it looks like it would have gone over. and stopping down, even the slightest would have been just casing the need to dig a little more into the noise. I really appreciate those kinds of details in a camera and lens test. Well done.
 
Should have also added,

From a noise point of view and low light, the DXL2 was very good as well.

The DXL2 was also quite good in low light. The DIT I had with me wasn't sure what the native iso "should" be either and it took a while to get some consensus :-)


JB
 
You must log in or register to reply here.
Back
Top