First Day with Dragon

[Florian Stadler]

I got my Epic x Dragon back last night and with giddy excitement I took it home with a 32mm TLS rehoused Cooke S2 on the front. Since I had stripped the camera to the brain for the upgrade, I realized how small and light and nimble the camera can be without any of the usual gack. I shot some beautiful stuff with my kids hopping around at 5KHD 96fps in natural window light. I was thoroughly impressed. Nice natural image right out of the gate.

Today i found some time to do some testing. My Epic has been taking a back seat to my Alexa with Gemini Raw recorder so I wanted to see where it stood with the new sensor.

I shot Dragon at 6KHD 5:1 compression and Alexa 2.8 K Arriraw. I rated both cameras at 640ASA and put them side to side. I used RPP's and discovered that a 25 on the Alexa almost matches a 35 on the Dragon, the Dragon with the 35 being only slightly tighter (!). I lined up a matching shot that included full sun, open shade and covered dark shade. And sky. Both cameras were thoroughly unimpressed by my stress test and handled it beautifully. Highlights appeared to go just about at the same point on either camera according to the onboard tools.

I needed something more challenging for any of the cameras to show a weakness. I covered the skylights inside my office and opened the door, shooting from inside out. My incident meter read 1 2/3 at the camera, 4 1/3 at the door, 44in the sun and 8 in the covered shade (at 640). I exposed each camera at 5.6, 8, 11 and 16. Both cameras held every highlight, including a white appear held into the direct sun at T8!!

I then took the footage into Resolve to look at it in an UHD timeline.
The Dragon footage was very close to Alexa, except that I had to rate Alexa at half the ISO than the Dragon to get a match at the same exposure. I found Dragoncolor and Redcolor 3 to be a fairly close match to Arriraw with a Rec709 LUT. The only 2 big differences I found was that Dragon has more resolution and is marginally sharper and more detailed in UHD and Alexa has a little more to give in the shadows. To make the right choice for the right job has gotten just that much more difficult...

Here is both cameras at T8 in underexposure T 1.0 2/3 read at 640, developed at 800 & 1600ASA. They are UHD jpeg stills out of Resolve. Arriraw with standard LogCtoRec709 LUT and Dragon with Dragoncolor/Redcolor3 gamma and ASA changed in RAW tab.

http://florianstadler.com/Arriraw800.jpg

http://florianstadler.com/arriraw1600.jpg

http://florianstadler.com/dragon800.jpg

http://florianstadler.com/dragon1600.jpg

Here is 3 parts of the shot in RAW 5 frames each have at it...
https://www.hightail.com/dl?phi_action=app/orchestrateDownload&rurl=https%3A%2F%2Fwww.hightail.com%2Ftransfer.php%3Faction%3Dbatch_download%26send_id%3D2544297594%26email%3D17f0c843951a88cdff3d5767adf5abe3

 

[Doug Beatty]

Wow- Alexa at 800 seems to be a full stop more sensitive than Dragon in the shadows. I'm confused though, which did you choose: RC3 or Dragoncolor? And which gamma space?

I think a fairer test would be to set the Dragon to Dragoncolor and RLF and then apply the same Rec709 LUT to both Alexa and Dragon footage in Resolve.

 

[sergio arguello]

Wow- Alexa at 800 seems to be a full stop more sensitive than Dragon in the shadows. I'm confused though, which did you choose: RC3 or Dragoncolor? And which gamma space?

I think a fairer test would be to set the Dragon to Dragoncolor and RLF and then apply the same Rec709 LUT to both Alexa and Dragon footage in Resolve.

yes try that

 

[Rohan B]

hmmmmm

 

[luigivaltulini]

Florian,


the first thing if you want to post something, please let put the Raw files for download, maybe 5 frames R3d and 5 frames ARRIRAW frame, so we can also use Add Dragon.
Otherwise it's just chaos ...


thanks

 

[Florian Stadler]

which did you choose: RC3 or Dragoncolor? And which gamma space?

I think a fairer test would be to set the Dragon to Dragoncolor and RLF and then apply the same Rec709 LUT to both Alexa and Dragon footage in Resolve.

Sorry for the confusion I chose Dragoncolor and Redgamma3. Redgamma4 was too contrasty in this case.

I did try RLF and the Alexa LUT and the results are very similar as the end result. You do end up with a lower ASA in the metadata that way but you have to take saturation down quite a bit mess with the color balance and change lift gamma gain to get a match. I thought it fairer to use the manufacturers recommended development without any change to color balance.

I will try to post some r3d snapshots and Arriraw frames but prepping and uploading those takes time.

 

[luigivaltulini]

I will try to post some r3d snapshots and Arriraw frames but prepping and uploading those takes time.

great, expect to see the raw files!

 

[Ramesh Jai]

Why were you blocking the view behind the door in both instances of the Dragon shots? I'm asking because we could see behind the door in both the Alexa shots. Thanks.

 

[Jacobo Martinez]

There is something wrong going on in exposure, it just does not seem right that one of the cameras is one full stop different. One of the 2 cameras is not properly calibrated to use with the incident meter.
Did you see this difference in the monitor? One stop is noticeably different.

 

[Florian Stadler]

Why were you blocking the view behind the door in both instances of the Dragon shots? I'm asking because we could see behind the door in both the Alexa shots. Thanks.

In the actual shot I walk to the door, open and close it, walk outside into the full sun and then into the open and close shade. This is the only instance where you can see a big difference between the 2 at 5 stops under exposure.

 

[Florian Stadler]

There is something wrong going on in exposure, it just does not seem right that one of the cameras is one full stop different. One of the 2 cameras is not properly calibrated to use with the incident meter.
Did you see this difference in the monitor? One stop is noticeably different.

I was surprised at the discrepancy too. I rolled at the same time with the same brand lens at matched stops. The same amount of light hit each sensor. I can see this stop difference in all scenarios (the 2 cameras match closely if I adjust the metadata ASA by a little under a stop between the 2. In anything but the most extreme underexposure situation this won't make a difference.

Maybe the takeaway is to rate Dragon at 400, rather than 800, if underexposure is a concern? Maybe that's why Mark Toia likes to rate the camera at 250?

I'm still evaluating the results and trying to make sense of it.

 

[luigivaltulini]

with jpeg you can not do much ..
very destructive ... but with the raw would be much better.
but I do not see huge differences .. I'm sure Dragon with ADD can make beautiful images.
should have been done the same shot.
But thanks again for testing Florian.

[/URL][/IMG]

[/URL][/IMG]

 

[Florian Stadler]

jpg from UHD Resolve still grab is just an easy, quick and valid way to publish on the web as I don't work in RCX. I will have to isolate the timecode, translate it to absolute timecode and find the individual Arriraw frame as well as go into RCX to make r3d snapshots to provide the RAW files. No the shots were filmed at the same time with the camera side by side as I described.

 

[Doug Beatty]

with jpeg you can not do much ..
very destructive ... but with the raw would be much better.
but I do not see huge differences .. I'm sure Dragon with ADD can make beautiful images.
should have been done the same shot.
But thanks again for testing Florian.

[/URL][/IMG]

[/URL][/IMG]

I agree you can do more with the Raw images out of Dragon, as the artifacting on the JPG image you corrected looks really bad in this example.

What troubles me is the fact that I would bring up the ISO on a DragonColor image to 1600 to get an IRE on the scopes that falls more closely in line with Alexa at 800.

What makes more sense if you are concerned with noise is, as Florian suggests, to rate Dragon at 400 and expose accordingly, but then you are opening up 2/3 of a stop to compensate which is not always possible or desirable.

I did try RLF and the Alexa LUT and the results are very similar as the end result. You do end up with a lower ASA in the metadata that way but you have to take saturation down quite a bit mess with the color balance and change lift gamma gain to get a match. I thought it fairer to use the manufacturers recommended development without any change to color balance

This is something that has always bothered me. Red is always tweaking their color science, while Arri seems to have found a look they are pleased with.

Why isn't there a RED sponsored log->709 LUT for each of the RED color sciences to match the standard Arri log->709 LUT? Or do they exist and I just haven't found them yet? If they don't exist, then every test is going to be a flawed apples to oranges experiment.

 

[Florian Stadler]

Here are 3 equal parts of the shot in r3d and Arriraw, each one is 5 frames long.

https://www.hightail.com/download/ZUczS3dnYTJoeVlYRHRVag

 

[Matt Ryan]

Pushed the low end a bit further:

It's well known that the Alexa is about a stop brighter when compared at the same ISO. Keep that in mind when you shoot and give the Dragon 1 more stop, then compare.

 

[Thomas Koch]

I'm confused. I thought the Dragon was supposed to be ISO 2000 compared to 800 on the epic for similar noise levels. Is that wrong? Is there any reason to spend the $10K for this upgrade? All I care about is shadow noise at high ISOs (800-2000).

 

[Matt Ryan]

I'm confused. I thought the Dragon was supposed to be ISO 2000 compared to 800 on the epic for similar noise levels. Is that wrong? Is there any reason to spend the $10K for this upgrade? All I care about is shadow noise at high ISOs (800-2000).

Noise level is different from exposure level. This shot was exposed for the highlights outside, not for lowlight. If you shoot for low levels and open up you will not get this amount of noise because you are not closing down to expose for the highlights. Does that make sense? You can not look at this shot and say well the Dragon isn't good in low light. Yes, it is worth the upgrade cost by far.

 

[Thomas Koch]

copy. Now, only if they would send that e-mail.

 

[Florian Stadler]

Yes, worth the upgrade, the images are lovely. I purposely tried to BREAK the sensor to learn where it's breaking point is. The lens is set to T8.0 and my meter reads 1.0 2/3 @ 640 ISO, 5 stops underexposed at 800ISO!!! That being said the Alexa holds almost another stop...