Testing EPIC-M and ALEXA - suggestions?

[Clint Lealos]

I guess I just disagree. I don't think too many people will be shooting Epic at 3K unless they are doing high fps, and some real world comparisons sound more useful to me. Its like comparing a 5 speed car with a 6 speed car, but not using the 6th gear when test driving. I want to see the difference higher resolution makes, because that would be one factor in my desicion of which camera to use. To each his own though. I'm curious to see the results

 

[Eric Z]

Mark,
First of all, recording 3K on both cameras is apples-to-apples. You simply have to go by the lower resolution.
Secondly, at 3K you can use many lenses at your disposal, as EPIC's 5K requires a very high-quality glass to bring out its top performance.
I'm not sure Justin Mitchell has RPPs or Leicas, so any lens he may have that can fit both cameras should do just fine.

 

[IAN SUN]

Every system has tradeoffs. The cameras have similar sized sensors. One uses large pixels and benefits in terms of DR, the other has relatively smaller pixels producing greater resolution. These are imaging systems, you have to compare them as systems, for any analysis to make sense.

 

[Pietro Impagliazzo]

Every system has tradeoffs. The cameras have similar sized sensors. One uses large pixels and benefits in terms of DR, the other has relatively smaller pixels producing greater resolution. These are imaging systems, you have to compare them as systems, for any analysis to make sense.

Yes, and by purposely using Epic at lower resolution I can already see the skewed conclusions coming out of it: "Alexa is better than the Epic at the same resolution!" :biggrin5:

But yes, a test like this would be of use to someone. However it would be a little awkward since Epic would need to use wider lenses to achieve the same FOV. I'm not sure this is consistent, why would anyone compare just resolution, as Ian said? That's looking at just one point of the camera, but they are complete systems that should be used as a whole.

 

[Eric Z]

OK, Ian.

So Justin can shoot R3D at 5K FF, de-bayer to 4K, and compare that to Alexa's ProRes 4444 3K 16:9.
Then he can down-res the EPIC's footage to 3K ProRes in post, and compare that as well.

 

[Stephen Gentle]

OK, Ian.

So Justin can shoot R3D at 5K FF, de-bayer to 4K, and compare that to Alexa's ProRes 4444 3K 16:9.
Then he can down-res the EPIC's footage to 3K ProRes in post, and compare that as well.

Probably more likely to be downscaling to 2K and 1080p, because those are the resolutions they will probably distribute in.

I don't think that the apples-to-apples argument is useful. The test is to determine which camera system is better for them as a whole. They should see the cameras at their best (in the configuration that they would typically use - which, due to rental or ownership costs, may or may not include RAW recording on the Alexa).

 

[M Most]

I don't think that the apples-to-apples argument is useful. The test is to determine which camera system is better for them as a whole. They should see the cameras at their best (in the configuration that they would typically use - which, due to rental or ownership costs, may or may not include RAW recording on the Alexa).

I would think that any test should be project specific, and the end result should be judged as to what is most appropriate for the story being told and the creative vision for it. Every project has different needs, and each one has a visual pallette of its own. Comparative tests done in a vacuum are interesting for general knowledge, but they sometimes have little relevance to a specific project. In a visually creative world, one size does not fit all.

 

[Deanan]

OK, Ian.

So Justin can shoot R3D at 5K FF, de-bayer to 4K, and compare that to Alexa's ProRes 4444 3K 16:9.
Then he can down-res the EPIC's footage to 3K ProRes in post, and compare that as well.

There isn't a 3K ProRes mode on Alexa AFAIK.

IF you shot Epic at 3k, the FOV would be very different as well as the lens utilization.
Ideally you want the fov and lens to be the same which means slightly different camera
position in reality.

 

[Graeme Nattress]

I thought Alexa was 2880 across, which is 2.8k. Indeed, 1920x1080 is max ProRes output on Alexa. For comparison, shoot the mode you're going to shoot the cameras with in reality. If you try to shoot "3k" on both, you're trying to equalize a variable at the expense of vastly un-equalizing more important variables like FOV and framing, and you're not using the camera how you're going to be using it in reality - unless you're shooting 200fps on Epic, and then you can't manage that on Alexa so you're back to square one again.

Mike Most's suggestion: "I would think that any test should be project specific, and the end result should be judged as to what is most appropriate for the story being told and the creative vision for it." is spot on.

Graeme

 

[Oliver Peters]

I thought Alexa was 2880 across, which is 2.8k. Indeed, 1920x1080 is max ProRes output on Alexa.

Correct. I think 2880 is available in the RAW files. Definitely only 1920x1080 in ProRes. Also the option of shooting with a Rec 709 or Log-C profile.

- oliver

 

[James Brundige]

I
Mike Most's suggestion: "I would think that any test should be project specific, and the end result should be judged as to what is most appropriate for the story being told and the creative vision for it." is spot on.

Graeme

That would make this test very useful for the workflow tested, and not useful for most others. I agree with the point that each camera can only be judged in comparison to an intended distribution.

 

[Eric Z]

...1920x1080 is max ProRes output on Alexa.

In that case, what's the point of this comparison? The EPIC's going to win hands-down if both are going to be evaluated at their max settings.

Let's look at it this way: Suppose the end distribution format is going to be 1920x1080 ProRes 422 HQ. Otherwise, there's no point in considering the Alexa at all.
So the Alexa can do this natively.

What's better for the EPIC?
1. Shoot at 5K 16:9, correct shadows & highlights in RCX, and then down-res to 1920x1080 in post (plus convert to ProRes 422 HQ).
2. Shoot directly at 1920x1080, correct shadows & highlights in RCX, and convert to ProRes 422 HQ.

 

[Graeme Nattress]

Eric, "max settings" as you put it are how the cameras are used. If you don't test them as you'd use them then your results are invalid. For best results on Epic shoot the highest resolution that your frame rate choices allow for, grade from REDColor2/REDLogFilm (or RedGamma2 for a more TV workflow), then downsample to your output format.

Graeme

 

[Eric Z]

Justin,

What are your thoughts on all of this resolution/format/settings debate?
You must have thought of that by now.

Have you received the update package yet?

Any updates sent to you from the ARRI camp?

 

[Deanan]

In that case, what's the point of this comparison? The EPIC's going to win hands-down if both are going to be evaluated at their max settings.

Let's look at it this way: Suppose the end distribution format is going to be 1920x1080 ProRes 422 HQ. Otherwise, there's no point in considering the Alexa at all.
So the Alexa can do this natively.

What's better for the EPIC?
1. Shoot at 5K 16:9, correct shadows & highlights in RCX, and then down-res to 1920x1080 in post (plus convert to ProRes 422 HQ).
2. Shoot directly at 1920x1080, correct shadows & highlights in RCX, and convert to ProRes 422 HQ.

#1 is preferred.
#2 is like trying to compare 35mm and 16mm film but cropping out a 16mm chunk out of the 35mm.

 

[Eric Z]

Graeme,

Thanks for your input.
I totally respect anything you say with regard to manipulating footage.

 

[Eric Z]

#1 is preferred.
#2 is like trying to compare 35mm and 16mm film but cropping out a 16mm chunk out of the 35mm.

So does that mean that the EPIC is designed to use its inner pixels when setting a resolution less than 5K FF?
The outer-most pixels are simply discarded?

 

[Graeme Nattress]

Eric, yes - the lower resolutions are "crop" modes where a smaller area of the sensor is read out faster, hence achieving higher max frames-per-second. That's why you should try to shoot the highest resolution that offers the fps you need. If you just need 24fps, shoot full rez.

Graeme

 

[Jeff Kilgroe]

As Graeme said, EPIC (and other RED cameras) achieve lower resolution by using a cropped region of the sensor, not scaling the image. The primary reason to shoot a cropped region of the sensor on EPIC (or RED One) is for the faster frame rates. Some other minor reasons come into play, such as lens choice and coverage. When shooting RAW on the Alexa, it's the same way, full sensor ARRIRAW is 2880x1620. Their onboard ProRes is scaled.

For this test, for it to be of real comparative value, I think they need to record ARRIRAW and compare to REDCODE RAW. I personally won't be interested and won't see a whole lot of value in the comparison if they're recording ProRes 422 on the Alexa.

If a lower resolution or sensor crop is to be done on EPIC, I think the only one that should apply is perhaps a setting of 4880x2700. This makes the sensor area roughly the same dimensions as the Alexa sensor with a 16:9 aspect. And that is perfectly valid, not crippling the camera. Lots of people will be shooting EPIC in this mode as it's the largest 16:9 resolution. Full 5K is 5120x2700, or only 240 pixels wider at the same height. Not a big difference either way.

 

[Dan Kanes]

Let's assume they are using sxs prores 4:4:4:4 on the Alexa.

This is how 98% of the Alexa jobs I have done shoot.

So, for an apples to apples comparison I would do:

Prores 4444 logC on Alexa
5k 16:9 on epic. Full debayer to 1920x1080 prores 4444
Using rc2/redlogfilm. No other adjustments.

Examine log results
Grade log results
Compare

This is what I would call a super basic episodic workflow
And it is where Alexa is supposedly king of the hill right now: episodic TV
Commercials have been doing same.