ARRI ALEXA LF - 4k, FullFrame announced

[Steven Abrams]

So, with that in mind, I have a humble request of RED: stop fucking around with this lowlight bullshit…um, pretty please. I want RED to design me a sensor that sees 9 stops into the highlights.

The "lowlight bullshit" is actually what you kind of need for more stops in the highlights (if you already have the dynamic range).

Let's make a camera/sensor and say the native iso we've decided is 800, you have 14 stops, and the DR we decided is +6/-8. If you want +9/-5 that will push middle grey on the log curve down 3 stops. So our 'native' ISO800 just became 'native' ISO100. You might be fine with that, many will not. If you want a more 'reasonable native ISO' lets say ISO400 - that would be equivalent of taking the noise etc you get from our sensor at ISO3200 and making that ISO400. How would you feel if your ISO400 camera had the same noise as it would have had at ISO3200? That's where lowlight performance helps. To be competitive and have a native ISO800 we would now have the noise performance our sensor gave us at ISO6400. Damn I hope our lowlight is good.

So at ISO800 Arri has +7.4/-6.6, about a stop and half more than our hypothetical camera. That means they decided their noise performance at around ISO2500 with +6/-8 was good enough to call it ISO800 with +7.4/-6.6 and push more range into the highlights.

That's what highlight range is. Full well capacity AND sensitivity/noise. Less noise and better lowlight with high full well capacity means you can choose to rate the camera lower and put more stops into highlights. Arri has large photosites which collect a lot of light (i.e they are very sensitive) and a large full well capacity. If they had decided to put less dynamic range into highlights for their log curves, they would have a good 1-2 stops higher ISO ratings than currently. Arri doesn't want to shrink their pixels until they can achieve the same performance or better. Seems like they haven't yet. Neither has anyone else.

If you want your +9 stops now, rate the camera around 3 stops slower than you currently do (assuming +6 in highlights at native), output linear from the RAW files and put your own log curve on that puts middle grey 3 stops lower than Red does. You'll probably wish it had better lowlight performance though. But if not, job done. You can do this already.

Add to this it was only recently that Red acknowledged that their iso ratings were actually a stop darker than everyone else and changed it to match what others do. Maybe we should do that with our hypothetical camera. Just say our ISO400 is actually ISO800 without changing anything. Still means our noise and low light performance at ISO3200 needs to be good since it will now be called ISO800, but seems more doable than saying ISO6400 is ISO800. Are you sure we can't do +6/-8 ?

 

[Jacek Zakowicz]

But the highlights will blow out early if you rate it 3 stops slower....

 

[Sabovic Adis]

Blah...

If You really wanted filmlike response You'd shoot film.

This is what You think You say You want

and this is what You actually are saying. :lurk5:

Okay, fight me.

Pussy! :rofl:

 

[Brian Merlen]

Where can i get this FULXA® sensor? count me in!

 

[Tom Gleeson]

It's pretty clear what the dream camera is for most professionals: The resolution and versatility of the R3D format in 8K, the small camera-body of the Weapon, the built in wireless video and focus with standard protocols like in the Arri, the full frame like in both, the film-like colors and DR handling of the Alexa, the stability and reliability of the Alexa.................It's hard in practice, but it begins by admitting what is actually good about the competitors. Red could easily create a sensor and color science that closer resembles Alexa colors and DR handling.

Christopher,

No argument from me and I agree with what you say but I do wonder if Arri are still the DR kings? I would love to see and maybe somebody has done it but I suspect the Monstro chip could challenge Alexa on dynamic range? When I tested the Weapon Dragon I felt Alexa just won by a nose. I'm tempted to setup another test but its so much work and the hate mail is such a disincentive.

*Ideally you would need to test an Alexa LF against a Monstro

 

[Brian Boyer]

If You really wanted filmlike response You'd shoot film.

That statement runs counter to everything digital cinema camera manaufacturers have been trying to do for the last 10 years. There are basically only two moving picture capture technologies of note - video and celluloid film.

Despite what you might think, they are contemporaries. Film is older, but video cameras date back to the early 1900s. The film look is the preferred look for narrative storytelling. The video look lost in that area.

Digital capture eventually supplanted celluloid for numerous reasons but that’s no reason to throw the baby out with the bath water. The film look is still what we’re after even though actual film is out of reach or impractical for what most of us are shooting.

In addition, your simplistic statement contradicts pretty much every DoP coming from a film background who I’ve read or heard share their thoughts on why they chose to shoot a project on Alexa. Not every time, but more times than I can remember.

And, for the record, it’s more like blah blah blah.

Pussy! :rofl:

I don’t quite know what to make of this but, if it’s meant as an insult, it’s not doing much for me. I’d like something more clever, if you think you’re up to it.

 

[Brian Boyer]

Where can i get this FULXA® sensor? count me in!

It certainly has an appeal. RED may deem it a bit too confrontational a moniker for an official release in the long run but it could be a good code name.

I’m drifting back to working my name in. Brilux? Boyerlux? Still a WIP.

 

[Sergio Perez]

One thing is for sure. Top 2 are Arri and Red in the digital cinema World. This is undisputed. What we are discussing here is how to take the battle for the number 1 spot. That battle was neck and neck on the MX days but now seems just that little bit leaning towards Arri. But obviously only on the Digital Cinema usage front.

We are obviously all speculating about the performance of the LF.

I would really like to see a fair comparison between Monstro and the LF.

Monstro will be irrespective of the results my next sensor. Like I said, RED is not simply a Digital Cinema Camera. Its an Artist's camera, a tool that transcends merely the "movie" world.

 

[Sabovic Adis]

That statement runs counter to everything digital cinema camera manaufacturers have been trying to do for the last 10 years.

No, it counters Yours statements which counter common sense.

There are basically only two moving picture capture technologies of note - video and celluloid film.

Which one do You prefer?

Despite what you might think, they are contemporaries. Film is older, but video cameras date back to the early 1900s. The film look is the preferred look for narrative storytelling. The video look lost in that area.

I know that, but what's a film look? What's a video look? The look of what film? The look of what video? Spit it out!

Digital capture eventually supplanted celluloid for numerous reasons but that’s no reason to throw the baby out with the bath water. The film look is still what we’re after even though actual film is out of reach or impractical for what most of us are shooting.

But why is that?

In addition, your simplistic statement contradicts pretty much every DoP coming from a film background who I’ve read or heard share their thoughts on why they chose to shoot a project on Alexa. Not every time, but more times than I can remember.

Verecundiam much?

And, for the record, it’s more like blah blah blah.

I know, I was just too lazy.

I don’t quite know what to make of this but, if it’s meant as an insult, it’s not doing much for me. I’d like something more clever, if you think you’re up to it.

:cryin:

 

[Brian Boyer]

The "lowlight bullshit" is actually what you kind of need for more stops in the highlights (if you already have the dynamic range).

Let's make a camera/sensor and say the native iso we've decided is 800, you have 14 stops, and the DR we decided is +6/-8. If you want +9/-5 that will push middle grey on the log curve down 3 stops. So our 'native' ISO800 just became 'native' ISO100. You might be fine with that, many will not. If you want a more 'reasonable native ISO' lets say ISO400 - that would be equivalent of taking the noise etc you get from our sensor at ISO3200 and making that ISO400. How would you feel if your ISO400 camera had the same noise as it would have had at ISO3200? That's where lowlight performance helps. To be competitive and have a native ISO800 we would now have the noise performance our sensor gave us at ISO6400. Damn I hope our lowlight is good.

So at ISO800 Arri has +7.4/-6.6, about a stop and half more than our hypothetical camera. That means they decided their noise performance at around ISO2500 with +6/-8 was good enough to call it ISO800 with +7.4/-6.6 and push more range into the highlights.

That's what highlight range is. Full well capacity AND sensitivity/noise. Less noise and better lowlight with high full well capacity means you can choose to rate the camera lower and put more stops into highlights. Arri has large photosites which collect a lot of light (i.e they are very sensitive) and a large full well capacity. If they had decided to put less dynamic range into highlights for their log curves, they would have a good 1-2 stops higher ISO ratings than currently. Arri doesn't want to shrink their pixels until they can achieve the same performance or better. Seems like they haven't yet. Neither has anyone else.

If you want your +9 stops now, rate the camera around 3 stops slower than you currently do (assuming +6 in highlights at native), output linear from the RAW files and put your own log curve on that puts middle grey 3 stops lower than Red does. You'll probably wish it had better lowlight performance though. But if not, job done. You can do this already.

Add to this it was only recently that Red acknowledged that their iso ratings were actually a stop darker than everyone else and changed it to match what others do. Maybe we should do that with our hypothetical camera. Just say our ISO400 is actually ISO800 without changing anything. Still means our noise and low light performance at ISO3200 needs to be good since it will now be called ISO800, but seems more doable than saying ISO6400 is ISO800. Are you sure we can't do +6/-8 ?

I know what you’re saying and if you read back through my TL;DR inducing posts, you’ll see I directly mention or at least allude to the very things you’re talking about. See the one with Phil’s charts.

I said in one of those posts that maybe we get the 9 stops through a new calibration of an existing sensor. I also mention exposing for highlights and bringing up mids and shadows in post. In the profanity laced post you quoted from and another one, I leave room for a different sensor technology (not CMOS) driving the expanded highlight latitude.

I called it lowlight bullshit mainly because a desire for more latitude up top is not what was driving the lowlight requests from RED owners. Their concern was strictly what they considered to be noisy images when they’re trying to shoot in virtual darkness - and under exposing to begin with in a lot of instances.

I wasn’t really suggesting lowlight performance isn’t important, only I’d sacrifice some of it in practical use if it meant getting 9 stops up top.

I know how to get 9 stops over by rating the sensor differently. I guess what I’m talking about is, however it’s done, implementing it in such a way that is transparent to the end user. They turn on the camera and set it to ISO 50, 100, 400, or whatever setting gets them +9/-6 with the understanding that this implementation, however it was achieved, was designed for highlight latitude and color fidelity above all other concerns.

Take those noise complaints elsewhere.

 

[Antony Newman]

I wonder which camera Chinese tent-pole productions would rather use : Monstro of LF?

With NASA, (possibly Musk), and Panavision choosing RED Sensors - they are more than endorsing RED image acquisition; they are to some extent also staking their own reputation on them.

It may take a few years to sway DP's who dislike Red Colour from a few years ago, to ones that actually want Monstro because they have seen what it can do. They need comparative tests, and people to drag it over-the-coals to demonstrate its image resilience. And - they probably need the tests to come from places that they trust outside the direct marketing of RED.

With Arri forced to release a larger sensor of their old technology (albeit with improved noise reduction - probably in exchange for resolution), I think it demonstrates just how difficult it must have been for RED to build their mobile 'supercomputers'. I am also guessing that Arri customers might wonder how many years it will be until Arri successfully get their next gen prototype out of the doors. If Arri release their next camera before 2020 - twould be a bit of a bitter pill for their existing owners (especially if there is no upgrade or trade in).

AJ

 

[Brian Boyer]

No, it counters Yours statements which counter common sense.

Which one do You prefer?

I know that, but what's a film look? What's a video look? The look of what film? The look of what video? Spit it out!

But why is that?

Verecundiam much?

I know, I was just too lazy.

:cry in:

It’s ironic that you would accuse me of Ad Verecundiam when your posts follow the “I am right because I say so” model. You haven’t bothered to back up one thing. Instead, you glibly answer a question with a meaningless question as if you’ve proven something. Pot, meet kettle.

Also, it doesn’t apply to my argument because the premise you put forth was if people wanted a film-like response, they would shoot film. Yes, I countered with the argument “I am right because X says so” but X represents at least some of the people in question, the people who are shooting digitally on Alexa and not film.

When asked why, many said explicitly it was because Alexa behaved more like they were used to with film. It’s direct evidence from X about what X chose to do and why. That’s enough data to invalidate your definitive and declarative assertion, which was actually mere opinion being passed off as some sort of truth. Where’s your evidence for that, son?

I’m not going to provide links because I don’t think you’re worth the time and effort and such quotes are easy to find on the internet but, for the sake of argument, I’ll pretend I did.

 

[Björn Benckert]

The Width of the DR counts and the sensitivity. iso is just numbers. Put a ND 6 in camera and it got more highlight protection than the arris.

 

[Sabovic Adis]

It’s ironic that you would accuse me of Ad Verecundiam when your posts follow the “I am right because I say so” model.

Really? And not because You try to back up Your argument by hiding behind pretty much every DoP's opinion?
Now let me ask You for Yours - what's a film look?
This:

or this:

this:

this:

this:

this:

this maybe:

A? They can't all be film looks, so which one is it?

You haven’t bothered to back up one thing. Instead, you glibly answer a question with a meaningless question as if you’ve proven something. Pot, meet kettle.

Also, it doesn’t apply to my argument because the premise you put forth was if people wanted a film-like response, they would shoot film. Yes, I countered with the argument “I am right because X says so” but X represents at least some of the people in question, the people who are shooting digitally on Alexa and not film.

When asked why, many said explicitly it was because Alexa behaved more like they were used to with film. It’s direct evidence from X about what X chose to do and why. That’s enough data to invalidate your definitive and declarative assertion, which was actually mere opinion being passed off as some sort of truth. Where’s your evidence for that, son?

I’m not going to provide links because I don’t think you’re worth the time and effort and such quotes are easy to find on the internet but, for the sake of argument, I’ll pretend I did.

Crap...

 

[Brian Boyer]

The Width of the DR counts and the sensitivity. iso is just numbers. Put a ND 6 in camera and it got more highlight protection than the arris.

True, but in that sense, whatever method you employ on a RED (or other) camera to achieve more stops on top you can do to the Alexa as well, so you’re still left with less highlight latitude than Alexa, all things being equal.

That’s not giving some of these higher end production that are favoring Alexa a compelling enough reason to look elsewhere.

And with LF being even better in low light, it doesn’t bode well for anyone wanting to claw back some of that business. Aside from the 9 stops, that’s what I’ve been getting at this whole time.

If comapanies want to release an Alexa-killer, they’re going to have to beat it on easy-to-achieve, no-brainer highlight latitude. That’s my theory.

 

[Björn Benckert]

Film handles out of range values very well. The celuoid creates a very nice and dynaminc roundning for over exposure and out of gamut colors. Digital cameras are linera in that sense and clip sharp and abruptly, so to make em behave more celuoid like colorsience and luts are applied.

Most dps seam to have great difficulties judge whats actually captured and what the color sience / post processing is adding and tend to think they work as one and can not be seperated.

 

[Björn Benckert]

True, but in that sense, whatever method you emply on a RED (or other) camera to achieve more stops on top you can do to the Alexa as well, so you’re still left with less highlight latitude than Alexa, all things being equal.

That’s not giving some of these higher end production that are favoring Alexa a compelling enough reason to look elsewhere.

And with LF being even better in low light, it doesn’t bode well for anyone wanting to claw back some of that business. Aside from the 9 stops, that’s what I’ve been getting at this whole time.

If comapanies want to release an Alexa-killer, they’re going to have to beat it on easy-to-achieve, no-brainer highlight latitude. That’s my theory.

Wrong as all else is far from equal as Monstro has a greater DR. So if you apply the same nd on the alexa it will be short in the botom end.

so its all about how you distribute the DR.

just a simple example:

one camera has twice the dr and twice the sensitivity, it does not mean it has half the highlight capability it only means you need to adjust your exposure accordingly and when thats done you still have two times the DR.

Sensitivity is a good thing.

 

[Thai Christen]

Giving the camera out to DOPs for a trail run didn't work for AJA Cion and I doubt it would work for RED (since said DOPs likely wouldn't care to learn its ins and outs; they'd just want it to perform optimally out of the box without futzing around).

You cannot compare AJA Cion with RED Monstro. RED as a brand and Monstro as a camera, completely different level.

As I said, no one has the magic formula right now. But I think RED has to tackle from a technological standpoint (Monstro in a different form factor? Fully working kit more towards production?), AND from a business perspective by trying something different and more agressive. You also mentioned about REDucation.

Time to think outside of the box and stop talking about numbers, pixels and such.

 

[Brian Boyer]

Wrong as all else is far from equal as Monstro has a greater DR. So if you apply the same nd on the alexa it will be short in the botom end.

so its all about how you distribute the DR.

just a simple example:

one camera has twice the dr and twice the sensitivity, it does not mean it has half the highlight capability it only means you need to adjust your exposure accordingly and when thats done you still have two times the DR.

Sensitivity is a good thing.

When I said all things being equal I didn't mean in terms of individual camera specs. I mean anything you can do to coax more latitude out up top (NDs, higher ISO rating, etc) on a RED you can do to Alexa, particularly to LF, as well.

I probably should've worded it better but I was typing on my phone at the time and revisions can get to be a pain at that size.

Yes, the non-LF Alexas will suffer more on the bottom end than LF but I've tried to make it abundantly clear I'd be happy with a noisier or more textured bottom end in order to get the stops on top.

I don't think I ever said sensitivity was bad. Like I told Steven, I mentioned the lowlight fixation mainly in terms of people complaining about noise. My profane post was also meant to be humorous.

Let me try to be clearer. I know that sensitivity is a good thing. What I'm saying is, in practice, I will gladly give up being able to see or use as much as the sensor can capture at the bottom if it means I can see way more at the top. Does that make more sense?

Also, I said previously that I've talked about DR many times before in connection with the way brake and traffic lights blow out to white very easily when shooting digitally. I've said maybe the only way to fix that is through sheer brute force DR with like 18 or 20 total stops.

My thing is, more sensitive sensors provide more total DR. So, if Alexa's highlight latitude is merely a function of how Arri chose to calibrate or define middle gray, and there have been sensors from every other manufacturer that are more sensitive, why is no one using that extra sensitivity in a way that provides more highlight latitude?

They always taut more sensitivity as a feature solely to achieve cleaner shadows. That's what I mean by a fixation on lowlight performance. Why isn't anyone marketing a sensor as a beast when it comes to highlights?

Are they reluctant to do so because so many customers want to go out and shoot with practically no lights and will complain if they see one speck of noise?

If that's true, then it's another reason why the sensor swap program is good because it will allow RED to release niche sensors/sensor-calibration combos for people who aren't obsessed with blacks that look like they were generated in Photoshop.

 

[Kemalettin Sert]

Film handles out of range values very well. The celuoid creates a very nice and dynaminc roundning for over exposure and out of gamut colors. Digital cameras are linera in that sense and clip sharp and abruptly, so to make em behave more celuoid like colorsience and luts are applied.

Most dps seam to have great difficulties judge whats actually captured and what the color sience / post processing is adding and tend to think they work as one and can not be seperated.

Really? Most DP's i known have better color science knowledge than most of the colorists.You should really check out Steve Yedlin.