Arri Alexa and Mysterium-X...

[Phil Jordan]

It seems to me that the math is fairly simple. Arri has a 3.3K sensor that can be recorded as uncompressed raw, good luck with that, or compressed RGB at 1080P. Compared to Redcode Raw it seems unspectacular at most. Alexa seems to be marginally comparable to the Red One MX sensor but certainly not EPIC. The pics on Bravins myspace shows that Alexa is BIGGER than the Red One. It seems to me that Alexa is a day late and a dollar short just about anyway you look at it.

 

[Stephen Matthews]

from the Arri website...

"The ALEV III sensor design provides 32 pairs of outputs. Each channel is divided into a high 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 lowlights are reconstructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful information in all 16 bits."

... there is more there if you want to look into it further, but I'm sure Mr. Bravin or someone else will clear things up before that.

 

[Phil Jordan]

from the Arri website...

"The ALEV III sensor design provides 32 pairs of outputs. Each channel is divided into a high 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 lowlights are reconstructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful information in all 16 bits."

... there is more there if you want to look into it further, but I'm sure Mr. Bravin or someone else will clear things up before that.

I'm not sure how this gets them more than the stated 13.5 stops of DR or more resolution. This fact doesn't make their camera smaller or lighter or give them more frame rates. What did I miss from your post?

 

[Mike Gross]

It seems to me that the math is fairly simple. Arri has a 3.3K sensor that can be recorded as uncompressed raw, good luck with that, or compressed RGB at 1080P. Compared to Redcode Raw it seems unspectacular at most. Alexa seems to be marginally comparable to the Red One MX sensor but certainly not EPIC. The pics on Bravins myspace shows that Alexa is BIGGER than the Red One. It seems to me that Alexa is a day late and a dollar short just about anyway you look at it.

Yes it does appear bigger than an MX red one but it has a nice cut out for the shoulder. Alexa looks as if to ride higher though.

I'm still bummed that they did not go for high resolution to compete with red one MX or Epic. Whyyyyyy?! I can't buy a camera that gives me fewer options for my clients when my competitors will have a camera with twice as much resolution for a 1/3-1/4 the price.

And if red one MX is the competition for Alexa, it makes it harder for me to choose Alexa knowing Epic will trump Alexa by a wide margin.

Sigh. Very frustrating that Arri isn't competing on all fronts.

 

[Sanjin Jukic]

Someboby would might think that I'm against Alexa but I'm not.

I like more a real things and facts only.

So here we are: For all of you new Arri Alexa fans you can get more picture at

ARRI's Photos Facebook>>>

 

[Stephen Matthews]

I'm not sure how this gets them more than the stated 13.5 stops of DR or more resolution. This fact doesn't make their camera smaller or lighter or give them more frame rates. What did I miss from your post?

Was trying to help out Roberto, nothing else was meant by it, the Alexa is far out of my price range so I have very little interest in it at this point.

Edit: though I realize that I just restated what Roberto has already herd most likely... ah well I try.

 

[Phil Jordan]

One area where Alexa is going to have problems is with 3D rigs. People are complaining that a Red One is big for this purpose and Alexa is bigger. Epic is half the size and will make a great camera for 3D rigs. I just don't understand why Arri would build this camera. It is too big, not enough resolution and not that great of recording options. I admit becoming a Red fan but it is because they seem to understand where the market is going and not where it has been.

 

[Jannard]

I think you meant Facebook...

Jim

 

[Sanjin Jukic]

One area where Alexa is going to have problems is with 3D rigs. People are complaining that a Red One is big for this purpose and Alexa is bigger. Epic is half the size and will make a great camera for 3D rigs. I just don't understand why Arri would build this camera. It is too big, not enough resolution and not that great of recording options. I admit becoming a Red fan but it is because they seem to understand where the market is going and not where it has been.

Neither big nor small, something in between:

From ARRI's Photos - Stijn Van der Veken, SBC - ALEXA prototype shoot
"Preparing a low shot of the Brandenburg Gate before magic hour. Look at how small that camera is!"

 

[Sanjin Jukic]

I think you meant Facebook...

Jim

Thanks Jim.

 

[Mike Gross]

One area where Alexa is going to have problems is with 3D rigs. People are complaining that a Red One is big for this purpose and Alexa is bigger. Epic is half the size and will make a great camera for 3D rigs. I just don't understand why Arri would build this camera. It is too big, not enough resolution and not that great of recording options. I admit becoming a Red fan but it is because they seem to understand where the market is going and not where it has been.

It's designed to be a TV show cam.

Epic will be very nice for 3D for both beamplitter and side by side as it looks like it can get very close. I heard it's a 2.5" to 3" interocular.

Alexa in a side by side rig would get you to a 6" interocular.
Mr Bravin, is that about right?

 

[Roberto Lequeux]

from the Arri website...

"The ALEV III sensor design provides 32 pairs of outputs. Each channel is divided into a high 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 lowlights are reconstructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful information in all 16 bits."

... there is more there if you want to look into it further, but I'm sure Mr. Bravin or someone else will clear things up before that.

I read that, but all that means -to me- is that they spread information through the bit depth so that the size of the file is perhaps used a bit more effectively, to carry the most information it can hold for a given size, be it uncompressed or compressed.

But this doesn't sound like, or explain how, it could increase DR...

Edit: though I realize that I just restated what Roberto has already herd most likely... ah well I try.

It is very appreciated. I still don't get it.

 

[Michael Bravin]

[/URL]

Back on topic, I just don't understand how manipulation of a signal increases DR, other than through downconversion where you get to average out some of the noise.

Does anyone understand how this separate gain step increases DR? I am at a loss...

It really sounds like an analog curve if anything, which could make the response more the way ARRI may like it to be...

...but how does this increase DR??

Looks like Michael is busy with the Alexa program, so could someone else that might understand this technique jump in and explain?

Roberto

Here's a picture http://arridigital.com/teaser/sensor


CMOS imager design provides for amplifiers to amplify the very tiny analog voltages that are read off the CMOS.

DGA splits them into a high gain amplifier path for lowlight data and a low gain amplifier path for highlight data.

These 64 (2 X 32) signals are sent from the amplifiers to a 14 bit analog voltage to digital data converter (A>D). Combining the two 14 bit signals yields a 16 bit signal with meaningful data from lowlights (with modulation above the noise floor) to highlights (with unclipped modulation) yielding 13.5 usable stops of dynamic range.

Not sure where you got the idea that we manipulate or increase dynamic range. We achieve the 13.5 stops by capturing from the CMOS using DGA. We START OUT with 13.5 stops of dynamic range and we tell you how we do it.

Hope that make sense.

 

[Michael Bravin]

It seems to me that the math is fairly simple. Arri has a 3.3K sensor that can be recorded as uncompressed raw, good luck with that, or compressed RGB at 1080P. Compared to Redcode Raw it seems unspectacular at most. Alexa seems to be marginally comparable to the Red One MX sensor but certainly not EPIC. The pics on Bravins myspace shows that Alexa is BIGGER than the Red One. It seems to me that Alexa is a day late and a dollar short just about anyway you look at it.

Phil

Alexa has a 3.5K sensor and outputs 2K ARRIRAW data which is uncompressed off the CMOS sensor (no luck needed). 1920 X1080 uncompressed 4:4:4 video and additionally an internal recording method which will be publicly revealed on April 6th at the DGA

My understanding is that Redcode is wavelet compression of uncompressed bayer raw off the sensor that is compressed right before going to media. That is not like ARRIRAW which is uncompressed data off the sensor.

 

[Roberto Lequeux]

Gotcha. Thank you Michael. I wonder if R1-X achieves its 13.5 with more than one "channel" of voltage amplification or with just one.

While I have you Michael, I'd love to ask you one more. What are the plans for the recording side? All I remember is that there are plans for an "on-board module".

Will that module record RAW 2k, or only RGB 1080p? And if RAW, are there plans to compress that file at all, even if only in a mathematically lossless option?

and additionally an internal recording method which will be publicly revealed on April 6th at the DGA

Oh... I guess I may be asking about that.

 

[Michael Bravin]

Gotcha. Thank you Michael.

I wonder if R1-X achieves its 13.5 with more than one "channel" of voltage amplification or with just one.

While I have you Michael, I'd love to ask you one more. What are the plans for the recording side? All I remember is that there are plans for an "on-board module".

Will that module record RAW 2k, or only RGB 1080p? And if RAW, are there plans to compress that file at all, even if only in a mathematically lossless option?

as I mentioned to Phil "...and additionally an internal recording method which will be publicly revealed on April 6th at the DGA"

ARRIRAW is uncompressed, isnt the point of capturing raw to not compress anything? The 4:4:4 HD Video is uncompressed also.

Also Alexa is thinner and not as tall (including the built in shoulder mount) when compared with Red One.

 

[Dan Graetz]

Neither big nor small, something in between:

From ARRI's Photos - Stijn Van der Veken, SBC - ALEXA prototype shoot
"Preparing a low shot of the Brandenburg Gate before magic hour. Look at how small that camera is!"

Seeing the Alexa sitting on that enormous Anton Bauer brick made me wonder - has anyone heard what the battery life of the Alexa is?

 

[Roberto Lequeux]

as I mentioned to Phil "...and additionally an internal recording method which will be publicly revealed on April 6th at the DGA"

ARRIRAW is uncompressed, isnt the point of capturing raw to not compress anything? The 4:4:4 HD Video is uncompressed also.

Also Alexa is thinner and not as tall (including the built in shoulder mount) when compared with Red One.

Yes. I noticed your last post and edited my last one when I saw that you have to wait till the 6th to announce.

ARRIRAW is uncompressed, isnt the point of capturing raw to not compress anything? The 4:4:4 HD Video is uncompressed also.

What about compression that only allows for visually irrelevant advantages? Or is it that because Alexa will record only 2k you want to make sure that -for lack of a better term- nothing goes wrong when pushing in post? But even then, what about at least offering a mathematically lossless compression? Which seems like a fantastic idea.

I mean, I know that ARRI sort of naturally markets towards those who would otherwise shoot 35mm, so I suppose this generally means customers wouldn't worry much about the cost of RAID storage and LTO backup. But...

...does Arri really see no benefit in super small, super light, affordable magazines that can hold multiple 1,000 ft. mags in one serving? Current SSDs are already quite big and they aren't getting any smaller.

On the other hand, I would imagine that the smaller a file is, the easier and cheaper it is to store it in a reliable way till it gets backed up.

So I totally understand the idea of retaining all information, but I don't see a justification for keeping an uncompressed file from becoming smaller through mathematically lossless compression. Right?

 

[Deanan]

My understanding is that Redcode is wavelet compressed data off the sensor. That is not like ARRIRAW which is uncompressed data off the sensor.

Uncompressed bayer raw off the sensor and then compressed right before going to media. (if that's what you meant to say

 

[Michael Bravin]

Yes. I noticed your last post and edited my last one when I saw that you have to wait till the 6th to announce.



What about compression that only allows for visually irrelevant advantages? Or is it that because Alexa will record 2k so you want to make sure nothing goes wrong? Or what about mathematically lossless compression?

I mean, I know that ARRI sort of naturally markets towards those who would otherwise shoot 35mm, so I suppose this generally means customers wouldn't worry much about the cost of RAID storage and LTO backup. But...

Does Arri really see no benefit in super small, super light, affordable magazines that can hold multiple 1,000 ft. mags in one serving? Current SSDs are already quite big and they aren't getting any smaller.

Of course you're right Deanan... my bad, I edited the original.


Roberto

Why would you compress if there are available uncompressed recording devices like S.Two, Keisoku Giken, and Codex that record ARRIRAW?

With dual link and 3G HDSDI outputs Alexa will support "several" workflows. April 6th is just 4 days from now. We'll talk all about this... and more.