Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

"simple and ridiculous" dragon vs Mx test..

Status
Not open for further replies.
luigivaltulini said:
I completely agree with you.
For now my only expectation is a direct competitor to Alexa, the F65 in my part does not exist, there is only Alexa, and some F55.
And honestly the only test I did between Mx and DRagon and see how much difference exists under tungsten light, and here I can confirm that DRagon is greatly amazing than MX. Everything else is better skintone by the dynamic range of colors, resolution, everything is much richer.
Just having the carbon with the new OLPF do testing with Alexa, this to me is the most important thing.
Click to expand...

Agreed.
Alexa/f55/C500 are the main contenders in most markets. Having the Phantom 4k and the F65, the KineRAW and the BlackMagic 4k (if it shows up) in there, does not hurt though.
 
Denis Buhot said:
... you're right, but do not miss the point here, please. No one expects Red failing to improve their sensor. The question is that we are very different people here, cost sensitive in different ways. For people like me, the question is whether the improvement is worth the amount of money. Besides, the iso thing is only one side of the possible and already appreciated improvements, and the point is trying to check on that particular question, because it is of vital importance for people shooting with no additional light source available. One very quick, simple test does not stop anyone to conduct any other he wants.
Click to expand...


To me it was, once I got to shoot it.
I think that is a pretty repeatable experience.

Not sure if tests conducted by others will convince anyone as much as that.

It is in all ways a new and different camera.

Which it should be.
 
Gunleik Groven said:
Agree. David.

Gonna participate in a F65/Alexa/F55/C300/Phantom Flex/Dragon/+++ thingy in February.

I do not think the Epic is even in there. (I do not pick the cameras)
Click to expand...

Gunleik,
where you are to take the test?
you can only come to see?
 
A lot has been written about the difficulty of testing and the likeliness of misleading results. All true. However, here is a suggestion for a very simple test. Ridiculously simple. Virtually no time asked from any of you. It would not adress all our needs, but it would adress quite a few of mine and presumably of others who work in uncontrolled lighting environment. How about the same shot, in dim enough light so that the histogram would be correct, show no noise bar, on both cameras, when not viewed raw of course. MX would be set at iso 800, Dragon would use a combination of high frame rate + shutter speed that would require iso 2000 to get a similarly centered histogram. And both results would be posted for us to compare.

Ok, I know... that test is a very basic, awkward one, scientifically questionable. Not totally despicable from a matter of fact standpoint. Anyone with my needs would see its usefulness, I guess. And no one would ask from that test results or conclusions that it obviously cannot provide.

So... any of you having your hands on both cameras ?
Click to expand...

+1 Simple but badly needed!
 
I get Denis point and I think it is a valid one. Regardless of the fact that I know that Dragon is absolutely amazing. He just wants to know the difference in noise side by side in low light between MX and DRAGON. I used to want that comparison, but now I do not worry about it anymore after Gunleik's tests. But it is nice to see different tests underexposed to see how the noise kicks in and compare them.

Happy Sunday! :)
 
David Mullen ASC said:
Once the conversions to Dragon have taken place at the rental houses, and the Dragon color science is done.
Click to expand...

I am honesty not sure anymore what to expect from the new color science at this point , at NAB they said they were just tweaking it ... and it was 7moths ago ... they shipped quite few cameras and shooting big movies without that color science ...so must not be that much relevance i guess ...
 
Gabriele Turchi said:
I am honesty not sure anymore what to expect from the new color science at this point , at NAB they said they were just tweaking it ... and it was 7moths ago ... they shipped quite few cameras and shooting big movies without that color science ...so must not be that much relevance i guess ...
Click to expand...


I think two things are confused.

1. How the sensor sees colors (which is there and hugely different from MX)
2. Software optimized for that.

New colors are there.
New optimization is not.

Still a ton better than MX.

Actually (as stated a few times before ) THAT was my reason for changing my mind from "no" to "yes" on the Dragon.
Not 6k or DR but fidelity in colors.
 
not saying that is not better than MX , for what is have tried on my CC system looks way better

just saying that people keep talking about this long waited new color science ( for post) but the fact that they prefer to have camera shooting and finishing out there without the color science being available means that maybe the difference will not be that big (i mean between using Redcolor 3 and the possible new "dragon color" )

i guess will see ...
 
Gabriele Turchi said:
not saying that is not better than MX , for what is have tried on my CC system looks way better

just saying that people keep talking about this long waited new color science ( for post) but the fact that they prefer to have camera shooting and finishing out there without the color science being available means that maybe the difference will not be that big (i mean between using Redcolor 3 and the possible new "dragon color" )

i guess will see ...
Click to expand...

Possibly true...

I still think I see where it has potential, and I think I saw some of exactly that in Jarreds OLPF images. But they cannot be taken as a test. I know that. But for lowlight performance, if those were an indication, I think the difference can be significant.

But of course, I do not KNOW that. I just see the potential.
 
David Mullen ASC said:
At some point, how much more reassurance do people need that the Dragon sensor is an improvement over the (4-years-old?) MX sensor? Does anyone expect a test that shows that there is no improvement?

Personally, I'd rather see a test comparing it to the Sony F65, which is the Dragon's nearest competitor. Or the Alexa, which is the current dynamic range king. Once the conversions to Dragon have taken place at the rental houses, and the Dragon color science is done, and the new OLPF is installed, I don't see renting an MX-sensor Epic anyway so it's mainly just curiosity at this point to see it compared to the Dragon.
Click to expand...

I don't see renting MX anymore either, but I think I do see some indy owner ops, working their own clients for music video and so on independently, keeping their old MX Epics as MX, and continuing to work with that, and continuing to get vastly better results than their competitors shooting with DSLRs.
 
OK. Cool.

Interesting take on the analysis, but it gets some information through.

(Interesting is not bad. Just a different approach than I would have had! :) )

Even on youtube a lot of the juice and colorprecission comes through. The luma differences are somewhat more brutal, but you get that picture, too.


Thx.

I think for the ever recurring ISO question. it would be nice to start off at exposure 2 or three and use ISO to normalize levels between all exposures. Should give people some of those answers.

But THX for posting!

(that would mean:

set 1st exposure at 100, next at 200, next at 400, next at 800, next at 1600, next at 3200, next at 6400 and the last at 12800)

Levels should then be pretty similar in all the exposures, and you could look at the noise at different ISO
 
That is kinda what confuses me, and makes me think tests have little value for many. (Sorry for being arrogant), but if you do not know how to interpret the data, they are of little use.

this (and the Espen test) contains exactly the info people on this and other threads are requesting. There is one HUGE faultsource, and that is ME.

But they are there, the raws are there. Download them, normalize with ISO and the answers are there.

/rant off

But it illustrates the point that tests are eaily rigged and misinterpreted. (Besides they only give you info about what you are testing for).

As stated earlier:
Technically I have no problem "proving" that "any" iso equals "any" iso, and that 12800 can be noisefree whil 50 can be noisy.

But to get the info it seems like people want. There are files there to check with.

I will probably take them down when the new OLPF and colorscience is present. That will render these tests irrelevant.

The fun of testing...
 
It's a bottomless pit. Testing for others, and testing in general. The results mean next to nothing if you don't know what exactly was adjusted and why...don't get hung up on it.

Have to go do something else right now so can't reply in more depth.
 
Gunleik I think it would be great for many people for you to explain how you can get an image noise free @12800 ISO or 6400 ISO when people struggle above 1000ISO to get a clean image. I know a lot has to do with Lighting and Exposure, but it would be great to see your approach. I think people want the noise comparison for the most part at Underexposed images, because that is when the sensor kicks in with noise.

I am just interpreting what some folks are requesting...At this point I am cleared at all this tests and saving the bucks for DRAGON to upgrade this year...Until then...I ll enjoy the Dragons of others... hehehehe :)
 
Thing is simply that ISO is not a parameter when looking for noise profile.

And I get why people think that is hard to understand. But all ISO does is to normalize the recorded levels to a gamma-profile.
Early on, it was not like that. But it has been like that for a few years now.

To confuse matters more, all gammas but linearlight and RedLog seems to be based on RedLogFilm these days, which means that they by design "hold the top". (Even the "Legacy" gammas)
This is great for all practical applications and I love it, AND it means that it is possible to rise ISO to silly levels, take an image and "grade it back with a simple curve".

Yup. That is true :) That is how the curves are supposed to work, but it tells you exactly 0 about noiselevels at X exposure.

Thus simple ISO tests with eyballing of levels are

a) easy to rig to get whatever "proof" you want
b) are essentially mildly confusing...
c) UNLESS you use ISO to normalize a controlled exposure.

Simply said:

If you shoot an image at say 800, and want to look at noise "at 1600" as people say they do, changing the ISO gives you zero info. Even changing the ISO and using a waveform is pretty risky for faultsources.
You need to stop down the exposure in a controlled fashion.

And the ways we do that can all contain faultsources.

I did these tests with the aperture-scale on the lens and with shutter, because I do not like the image-contamination by doing it with ND's and fixed ND/shutter. (and I was not looking for IR info)
But that is a possible source for fault.
As is any other approach.

If you change the external lightlevels with a lightmeter instead, that is another source for faults, as you inevitably change the contrast in the image you record, and thus the lowlight info from one exposure to the next is not relevant.

So.
Best one can do is:
a) do it as controlled as you possibly can.
b) be open about your methodology, so people can repeat the test if they question the results.
c) try to give people a hint on how to extract relevant information out of the test, if that is what is desired.
d) admit fault if fault is done.

Now.
To take the example asked for here as "simple and ridiculous"

It is the basic test for checking DR. It is not at all ridiculous, but you must know how to do it and how to interpret the data.

And you must accept that all tests are by definition "faulty and narrow at best" and each test just looks at a certain part of the equation.

If I wanted to look at how shutterspeeds affect noise, I would need to normalize lights, NOT with ISO, but with lightsource. Aperture could be used, as long as DOF does not confuse the matter. A lightmeter and external source would be the better option for THAT test.

The test described would tell you very little about lowlight exposures, but a lot about the shutterspeeds effect on noise.

Problem with all tests, are that imagechains/image-creation by nature are complex. And you have to be conscious about what you test for and how you test it and make it repeatable to have any - even small - relevance.

And despite doing tests, I just understand more and more that they are not relevant for many.

I do want to check colorfidelity and saturation levels per channel through exposure-range when I get the OLPF and the new color-science, and I have an idea on how to do that. Because that interests me and tells me a bit about what light I should maybe use if I get into issues at certain exposurelevels, if I want colors to "pop".

But I think I will not publish such a test, because it seems all people see is "noise" or lack thereoff, no matter what one test for.

As to ISO tests, they are not very relevant without a controlled exposure parameter in the equation.

They tell you exactly nothing about noise creation, as on the REDs "X ISO = Y ISO" as to noise.

What ISO does do, is to Normalize an exposure of X to the levels of the exposure of Y, while still keeping the top intact.

And if that sounds complex... Well... That is because imagecreation and signalchains are pretty complex, even on this level.

And thus it cooks down to "do I like the Dragon better than the Epic". To me the answer is: Yup. By a wide margin.
As to "why", I can explain a bit, and the rest is:
Well I just like it. You should try one.
 
Status
Not open for further replies.
Back
Top