Welcome to our community

Be a part of something great, join today!

Register 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 :)

EPIC-ALEXA-BMC overexposure...

Status
Not open for further replies.
Björn Benckert said:
It's nice doing tests and not using the full potential of the camera. As he is not really turning on HDRx... is he.

I hate that kind of comparison.

The oppsite would be to bring up the iso of the alexa to 2500 and have a complety burned out image that could not be fixed in post and not mention that the camera can actually shoot 800 ASA baked that looks twice as good.

HDRx would blow the alexa out of the water for this shot.

It's the same when people take the epic and shoot 2k and say that epic is far softer than the alexa since even epic 2k is far softer than HD on alexa.... Then when trying to tell them, to take the 5k image sharpening it and then down res it to HD and then compare they just mumble something stupid that post sharpening is not whats in the test.

But again, the more people out there that do not understand the better it keeps the completion on a nice level..
Click to expand...

the more i read your posts Bjorn, The more i like you. I need to come to Sweden to buy you dinner or a beer.

david
 
There is definitely too much....:sifone: involved with some of these posts.

:)

The Alexa does not have 16 stops of DR. The Alexa does not have internal HDRx, or AUTO HDRx.
 
Martin Stevens said:
There is definitely too much....:sifone: involved with some of these posts.

The Alexa does not have 16 stops of DR. The Alexa does not have internal HDRx, or AUTO HDRx.
Click to expand...

No?, explain this: (from Arri's website)

"DGA simultaneously provides two separate read-out paths from each pixel with different amplification. The first path contains the regular, highly amplified signal. The second path contains a signal with lower amplification, to capture the information that is clipped in the first path. Both paths feed into the camera’s A/D converters, delivering a 14 bit image for each path. These ima-ges are then combined into a single 16 bit high dynamic range image. This method enhances low light performance and pre-vents the highlights from being clipped, thereby significantly extending the dynamic range of the image.
ALEXA’s sensor design provides 32 pairs of outputs. Each channel is divided into a high amplification (gain) path (H) and a low gain path (L), resulting in 64 channels arriving at the 14 bit A/D converters.
In the final images, the shadow areas are re-constructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful luminance information in all 16 bits."
 
Last edited:
Jeff Kilgroe said:
LOL. Uh, nope. Even ARRI won't make that claim. Alexa shows about 13.5 to 14 stops. About right at 13 usable in ideal circumstances.



I don't know, that seems right from what I've seen and the little bit I've played with the BMCC. I think it could do better than the 9.5-10 stops if BM could get their sensor under control. It does a terrible job at highlight rolloff and bright area tempering. They have the exposure limit black-out like the original M sensor (and most all CMOS sensors) have where you get the black dot in the middle of the sun or other white-hot objects. Simple to fix in post, yet most people never do for some odd reason.



That's complete rubbish. 13.5 stops yes and we've seen several charts to choose it. Under ideal conditions (as it always has to be for any camera), we're getting 12 to 13 stops out of EPIC. If they were having trouble at over 10 stops, with the EPIC or Scarlet, then they're doing something wrong. Rating the camera wrong or improperly post-processing their RAW. We were getting 9.5 to 10.5 stops of usable range on the RED One with original M sensor and it was rated at 11 to 11.5...

The sad part is so many people do get lost along the way when working with RED and processing the RAW, understanding exposure, etc.. It's super simple, but unless they put in the effort to learn the best practices, they won't get the results. Same thing with shooting ARRI RAW, but with ARRI RAW we have the ability to apply the LUT from the simultaneously-recorded ProRes or DNxHD proxy as a place to start if we don't want to begin with drab and compressed-looking RAW.

When it comes to usable, real-world DR, the EPIC and Alexa are nearly identical. Only your colorist will know for sure. And both are right there with the best film stocks, IMO. Film still produces superior reaction in the highlights and better flesh-tones right out of the camera. ARRI has done a phenomenal job with their default looks to mimic some of what film does and to preserve flesh tones. Both the EPIC and Alexa have superior dark detail response compared to film.



With all the extra DR afforded by Dragon, I think we're going to see a lot more poorly-graded footage. It's not the magic bullet that people want it to be.

Dragon should give us 16+ usable stops out of a registering range of close to 18. At least that's what I'm expecting based on what's been said and shown thus far. This should surpass what the best film stocks are currently delivering. At least in terms of range. If the noise response is as good as claimed, or as good as what we're seeing with the monochrome EPIC, then we're in for a treat there as well.
Click to expand...

Ciao Jeff,
I understand, I downloaded the file from the site of Ryan, one must admit that there is too much difference in your test I do not think they are only 1 or 1.5 stops.
How can we explain this situation? obviously people see these tests and of course at first glance says that RED be bad, but why '?
If we look on the Internet are just bad test on the Red One and The RED EPIC.
We must always fight .... I have to be honest when I saw this test I had the impression that you wanted to put a bad light on the EPIC.
So 'be in real life, I think all of us have found reduser often argue with people and explain how the Epic, the usual comments about skin tone, latitude, fan noise, too difficult to edit because' not in ProRes, etc. etc. in most of these tests in network certainly does not help.
Be all very boring ..
I own 2 RED Epic and I am very happy, normal shot, turn the crank and shot-to-shot 120 fps or 300 fps, I can take pictures, small, compact, fast, I love my Epic for ease 'with which I can use it. Why nobody talks about this or does not talk much?
Maybe then again maybe things will change with Dragon ..

It would be nice intervention by Graeme ..

excuse for my bad english :(
 
Pablo Moreno said:
No?, explain this: (from Arri's website)

"DGA simultaneously provides two separate read-out paths from each pixel with different amplification. The first path contains the regular, highly amplified signal. The second path contains a signal with lower amplification, to capture the information that is clipped in the first path. Both paths feed into the camera’s A/D converters, delivering a 14 bit image for each path. These ima-ges are then combined into a single 16 bit high dynamic range image. This method enhances low light performance and pre-vents the highlights from being clipped, thereby significantly extending the dynamic range of the image.
ALEXA’s sensor design provides 32 pairs of outputs. Each channel is divided into a high amplification (gain) path (H) and a low gain path (L), resulting in 64 channels arriving at the 14 bit A/D converters.
In the final images, the shadow areas are re-constructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful luminance information in all 16 bits."
Click to expand...

This, to me, is not HDRx or AUTO HDRx.

:)
 
Pablo Moreno said:

In the final images, the shadow areas are re-constructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful luminance information in all 16 bits."
Click to expand...

Read again, because, As far as I know, there is not such a thing as frame A - frame X, in Alexa's workflow, you get only one final frame, as I just quoted..

So for me, that is Auto HDR.

EDIT:

Anyway, is not a secret for me, that Arri is probably using this sensor, or similar from the same manufacturer, just like these guys here..
 
Last edited:
Pablo Moreno said:
No?, explain this: (from Arri's website)

"DGA simultaneously provides two separate read-out paths from each pixel with different amplification. The first path contains the regular, highly amplified signal. The second path contains a signal with lower amplification, to capture the information that is clipped in the first path. Both paths feed into the camera’s A/D converters, delivering a 14 bit image for each path. These ima-ges are then combined into a single 16 bit high dynamic range image. This method enhances low light performance and pre-vents the highlights from being clipped, thereby significantly extending the dynamic range of the image.
ALEXA’s sensor design provides 32 pairs of outputs. Each channel is divided into a high amplification (gain) path (H) and a low gain path (L), resulting in 64 channels arriving at the 14 bit A/D converters.
In the final images, the shadow areas are re-constructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful luminance information in all 16 bits."
Click to expand...

What is there to explain? It's pretty clear what's going on, without getting into the technical specifics of sensor operation -- they use a high and low read-out on their sensor to deliver the total acquired range. This is not a unique process.. Last time I checked, Sony and Canon were essentially doing the same thing on their DSLR sensors and employing both electronic and digital gain control as well. Not sure where you're coming up with 16 stops or internal HDRx (or whatever you want to call it). The Alexa uses a 14bit Analog to Digital Conversion (ADC) process and stores the data in a 16bit container. EPIC stores its data in a 16bit container as well, RED has not stated the resolution or bit precision of their ADC process.
 
Looking at the R3D files, for me it seems: The Epic footage was not expose properly - to close to the clipping point in the highlights. You need to handle each sensor in a different way to get the best out of it.
Second, the Alexa has some kind of diffuser built on the sensor if I am correct - add a simple diffusion filter to your Epic setup and you will get about the same milky 13.5 stops with the Epic.
And yes, Alexa uses a kind of HDRX (a better method for that I think) to get more latitude.
Both are great cameras!
 
Almost every test video against red epic or red in general is done shitty almost purposely to make it seem like a bad camera it seems. I remember before I shot with an epic I'd hear people talking shit about it etc without me personally using it and I'd be like yea fuck red blah blah <_< shot with an epic once edited some red files on my computer straight off the camera in premiere....instant love...these fools can EAD(I'm sure some will figure out what that means) I'm not a fan boy but it's a dope camera and too many tests videos done poorly to make another good camera shine
 
Even though I have not met Ryan Walters, I like his style and professionalism, and
have always been impressed with his work.

:)
 
I've worked with Ryan on many commercials and I can't say enough about his professionalism. I've also worked with him on tests (although quite awhile ago) and he is an very detailed tester. I'm a little surprised with the results of this test in regards to the Epic's handling of highlights but I think denying that the Alexa has an edge in highlight roll-off is silly. I think the Dragon and new color science will change this. I would still rather shoot with an Epic for other reasons. I like that Ryan keeps emphasizing that the cameras are just tools. This is what should be taken from camera tests by other people.

I'm surprised how sharp the BM camera is... I agree filtration and/or old glass seems like a must.
 
Jeff Kilgroe said:
What is there to explain? It's pretty clear what's going on, without getting into the technical specifics of sensor operation -- they use a high and low read-out on their sensor to deliver the total acquired range. This is not a unique process.. Last time I checked, Sony and Canon were essentially doing the same thing on their DSLR sensors and employing both electronic and digital gain control as well. Not sure where you're coming up with 16 stops or internal HDRx (or whatever you want to call it). The Alexa uses a 14bit Analog to Digital Conversion (ADC) process and stores the data in a 16bit container. EPIC stores its data in a 16bit container as well, RED has not stated the resolution or bit precision of their ADC process.
Click to expand...


What are you talking about?, that doesn't have anything to do with analog to digital conversion, even when is mentioned there.

It clearly states that, shadow areas are reconstructed from the high gain path (OVEREXPOSED), and the highlights are reconstructed from the low gain path (UNDEREXPOSED).

That is clearly HDR blend to me.. And I never said that Arri is the only one doing this, as a matter of fact, Sony and Aptina are licensing each other technologies, in order to take advantage of this new Dual readout sensor technology, that not only fixes rolling shutter problems, but also is able to do HDR in real time too.


But, forget it, lets jump back to the subject..


What Ryan Walters is trying to demonstrate, is how is the response in the roll off to the highlights that every camera have in particular WHEN OVEREXPOSED, and, IN CERTAIN SITUATIONS, and this is not the first time I see a test like this.

What he is doing, is very professional in my opinion.
 
Last edited:
Patrick Tresch said:
16 stops for Alexa. Some one smoked something!

Pat
Click to expand...

I own and shoot EPIC and SCARLET all the time. I also shoot Alexa all the time.

Alexa has much more in the way of dynamic range. I'd say 16 stops...and the base ISO of the chip is more sensitive.

Epic on the other hand makes up for it in all kinds of other ways.
 
remember this was an overexposure test not a beauty test.

Thing is Bjorn brought up a good point. HDRx is for overexposure that's reds tool for handling it, like it or not that is what it is. Now test the camera normally with overexposure but also use the tool built into the camera to do so, HDRx. you may not like how it looks but why not let the observers decide, and use it as a base for comparison.

Would have also been fascinating to compare it to the Alexa and his claims of 16 stops as we have seen Alexa and HDRx compared before and HDRx has more range... plus more issues. I find the 16 stop conclusion fascinating as no one else has come to that conclusion after a proper test heck even ARRI don't rate it at that which I'm fairly certain if it was 16 stops they would tell us all over their website.

but hey maybe it is, in fact that would be amazing, it's just way out of left field hence why so many people are confused by it and doubt it.
 
I work with Both cameras EVERY weeks for years, own two Epics and have to hire Alexas.

The latitude of the two sensors (without HRD of course) is practically the same, I think the key is the process of this test. I bet Ryan is grading the Epic from RG3 instead of any Logarithmic space, If not I really can´t understand this result. I film and grade footage from Epic and Alexa for comercials and Docs and this difference is incomprehensible, the highlights from Misterium-X are always much more smooth until overexposure, It´s like the stop was not the same or I´m loosing something. Anybody knows if Ryan use a 18% gray to set exposure on both cameras?, or he set both on 800ISO and use the same scene and same stop on both cameras?, because 800ISO in Irvine not means 800ISO in Munich ;-) (I think this is one of the errors here)

And "usable DR" is very subjective, in my experience I can´t see more than 11 stops in any of this sensors. This for sure.
 
Pablo Moreno said:
What are you talking about?, that doesn't have anything to do with analog to digital conversion, even when is mentioned there.
Click to expand...

Yes it does. Image sensors, including CMOS sensors like the one ARRI uses in the Alexa, are all analog. They are using a high gain and low gain read-out on the sensor and using a 14bit ADC for each readout. They're not doing anything unique here. The two 14bit ADC pathways are combined to form one single 16bit composite. This is not the same thing as HDRx that RED does. In fact, dual-gain read-outs on CMOS sensors are quite common. For all we know, RED is doing it as well, they have never said. But Sony and Canon do the exact same thing on their sensors. Translation is that Sony, Canon and most (if not all) Nikon DSLRs are doing this. I suspect the Canon C300, C500, etc.. and Sony F-series cinema cameras as well...

It clearly states that, shadow areas are reconstructed from the high gain path (OVEREXPOSED), and the highlights are reconstructed from the low gain path (UNDEREXPOSED).
Click to expand...

Yes, it also clearly states the digitization precision and container size. Your point?

That is clearly HDR blend to me.. And I never said that Arri is the only one doing this, as a matter of fact, Sony and Aptina are licensing each other technologies, in order to take advantage of this new Dual readout sensor technology, that not only fixes rolling shutter problems, but also is able to do HDR in real time too.
Click to expand...

I guess you could look at it as an "HDR" blend if you want. But what makes something "HDR"??? At what point do we transition from "Dynamic Range" to "High Dynamic Range"? The technique is indeed like blending a low and high exposure image pair to make one wider-range composite. Actually not even like doing that, it IS doing that. I guess my point is that doesn't make it "HDR", and both exposures put together in this manner is what allows the Alexa to reach it's 13.5 stops of usable DR out of what seems to be just about a total of 14 stops. A lot of people I bump into are quick to assume that the 16bit data structure implies that there are 16 stops of total DR achieved, but for the 10-billionth time on these forums, bits and stops don't correlate. EPIC/ Scarlet RAW is 16bits per channel as well, but RED doesn't like to say that because, like ARRI RAW, a lot more goes into the interpretation of that data through the de-bayer/ demosaic process and it is typically presented as 32bit float to apps that can use it.

But, forget it, lets jump back to the subject..
Click to expand...

Agreed.

What Ryan Walters is trying to demonstrate, is how is the response in the roll off to the highlights that every camera have in particular WHEN OVEREXPOSED, and, IN CERTAIN SITUATIONS, and this is not the first time I see a test like this.

What he is doing, is very professional in my opinion.
Click to expand...


The test is a good idea, and yes there is an air of professionalism here. I just feel that the execution and presentation are flawed. And that none of the cameras were used in a way that shows their true potential. ...even for this test.


Paul Schneider said:
I own and shoot EPIC and SCARLET all the time. I also shoot Alexa all the time.

Alexa has much more in the way of dynamic range. I'd say 16 stops...and the base ISO of the chip is more sensitive.
Click to expand...

I shoot Alexa a lot too. No way, no how does it have 16 stops. ARRI claims 13.5+, they actually like to say it's between 13.5 and 14. Never been able to reach a full 14 on any EV chart or metered environmental exposure test. 13, yes. 13.5, yes, but any more beyond that, even just a fuzz, we're clipping on both ends.

With EPIC, it seems to be half a stop less sensitive and have about half a stop less DR vs. the Alexa. But we get cleaner images due to the higher resolution and oversampling. I think the Alexa produces fabulous color right out of the camera, especially skin tones. Alexa feels a lot more contrasty and even gives a higher DR appearance right out of the camera, even with the Log color space LUT vs. REDLOG. EPIC, I can match it to Alexa just fine... after I spend time in Resolve to do it. :( I can't quite seem to get "the look" in REDCINE-X... I've said it before on these forums a bunch of times, the Alexa is a fine camera. Indeed easier in some ways, especially if you want great looking ProRes right out of the camera. EPIC is still my camera of choice for most everything I shoot, though. And my official stance on Alexa and 16 stops is that people are smokin' crack. Show me a properly exposed EV chart showing 16 stops and then we'll talk.
 
This is turning into a dr debate between Alexa and EPIC but he point still remains that the test is flawed. I am not challenging Ryan's professionalism here.

All this test tells me is that if i you use a light meter Alexa will perform more like traditional film stocks. There is no detail about how the files are processed which is the key element in RED and determines how the image looks.

Jeff you are right about out of the box Alexa images. There is no doubt that if Ryan had moved the mid grey the results can vary vastly.

This test I'd basically butting the mid grey value at Alexa's preset and then comparing the three cameras. So much more can be obtained from EPIC footage that what this test demonstrates and in that sense it is simplistic.

Id rather hear more about what techniques can be used to extract those stops from epic rather than see an out of the box image from epic. Like film, epic images must be exposed and processed properly.

It's been REDs way from day one.
no sharpening.
no excessive manipulation.
You work for the image, like film, by knowing what you are doing and understanding the tool.

this test compares tools without exploring them much.

We know that the highlights need care.

david
 
ok people,
but then why would Ryan even did a test so absurd?
It could have been set to 18% gray and then from them and show real differences.
Instead, once again epic look shit ....
It would be helpful if someone could make a good test and explain to ordinary people how to extract the most out of our cameras, in different lighting conditions.
So any of us, beginners and less able to consult and put it to good use.
And similar tests where only creates confusion and discontent.
In the next test Ryan,demonstrate the skin tone, I already feel that in this case the Epic will be bad .....
So 4 tests and 4 times where the Epic comes out as the worst for money.
 
Status
Not open for further replies.
Back
Top