All Arri Alexa News!

[georg lamshöft]

Just take a look at www.arridigital.com - interesting new infos.

 

I'm just not sure about their new style of marketing with simplifiaction of technical facts (like giving DR one number) & large, stylish images - is it really a consumer product at 50k€?

 

Why Fuji-lenses for budget work with ARRI-label?

[georg lamshöft]

I had some trouble with playing the first Alexa-clip, so here's the direct link:

http://www.arridigital.com/sites/default/f...World%20Cup.mp4

 

What about a 1080p-clip with 1:1 comparison S16 + 3k->1080p DI and S35 + 6k->1080p DI with Arri Relativity before somebody else does questionable comparisons again, I assume Arri knows how to use the full potencial of it's equipment best... - would at least give an idea how HD-broadcast/Blu-Ray compares ? To me it looks nice but still just like video, although hard to tell from such a small video...

[John Sprung]

The link didn't work for me, either. But I saw the Alexa presentation at the DGA last night. Even with the not so great black level on the projector, it's impressive. They've clearly leapfrogged the Red MX.

 

 

 

 

-- J.S.

[Scott Bullock]

They've clearly leapfrogged the Red MX.

 

Hi John,

 

Can you expand on your comment? I'm not a RED apologist or anything, and I'm a long-time lover of Arri, but I was just looking at the specs for the Alexa and it seems that the Epic specs are just as impressive or more so. The Epic has a 5K sensor compared to Alexa's 3.5K sensor, both are rated at EI 800, and the Epic will shoot up to 100fps compared to Alexa's 60fps (not sure if the Epic will do 100fps at 5K though).

 

Anyway, just looking for some additional details, especially from someone who's seen the Alexa demonstrated.

 

Thanks.

[Brian Drysdale]

I assume he's referring to the RED One upgrade, since no one outside RED has really yet seen an Epic.

Edited April 8, 2010 by Brian Drysdale

[Adrian Sierkowski]

Also don't forget the major selling point of an Alexa to me, uncompressed output...

[Brian Drysdale]

The ProRes will also be a big selling point for those people needing quick turnarounds.

[Scott Bullock]

The ProRes will also be a big selling point for those people needing quick turnarounds.

 

Yeah, I can definitely see the benefits of that.

[Adrian Sierkowski]

Most certainly... especially for me since i already own SxS cards! It's nice to feel a head of the curve once in awhile.

[John Sprung]

We haven't used the Epic yet, but we've done a couple pilots on MX-upgraded Red One's. Based on Sam Nicholson's available light night exteriors, Alexa looks to me to have a good 2-3 stops more range. Size and weight are similar, the user interface is simpler, and it's a more robust physical object. The electronics are in a sealed module with heat pipes to a big sink and an extremely quiet fan. Arri gets it about how we really use cameras.

 

There's more to life -- and sharpness -- than just counting pixels (photosites, really). To put more photosites in the same area, you have to make them smaller. That costs you dynamic range and decreases the actual light sensitive area, because you can't scale down the stuff between the photosites. Arri made a courageous decision to go for the best pictures, not the biggest numbers.

 

Alexa and Red are two different price/performance points. I'm sure we'll be using both, depending on the specific needs of each show.

 

 

 

 

-- J.S.

[Scott Bullock]

Thanks for the responses, folks. Those are a lot of good points.

 

John wrote: "Arri gets it about how we really use cameras."

 

Yeah, decided edge to Arri on that one. Now that the biggest camera rental house in Denver has closed down, I'm not sure if I'll get a chance to use the Alexa anytime soon. There are lots of REDs around though; too many, in fact, for this geographical area.

[georg lamshöft]

Hi John,

 

Can you expand on your comment? I'm not a RED apologist or anything, and I'm a long-time lover of Arri, but I was just looking at the specs for the Alexa and it seems that the Epic specs are just as impressive or more so. The Epic has a 5K sensor compared to Alexa's 3.5K sensor, both are rated at EI 800, and the Epic will shoot up to 100fps compared to Alexa's 60fps (not sure if the Epic will do 100fps at 5K though).

 

Anyway, just looking for some additional details, especially from someone who's seen the Alexa demonstrated.

 

Thanks.

 

The "I have more k than you" argument is a bit tricky:

The ALEV-III (sensor of the Alexa) has a pixel-pitch (distance from one center of each photosite to the next) of 8,25µm (=68,06µm²) which results in 3072 (3,5k are overscan for the viewfinder) "pixels" (pixels are a mathematical/IT term within the data, the photosites are the actual, physical entities on the sensor) at about 25mm wideness. The RED uses a 5,4µm (= 29,16µm² - less than half the area of Alexa given similar fill-rate) pixel-pitch which results in 4520 pixels (the active area at 4k is therefore a bit narrower than S35). Even when stronger downsampling (from 4k instead of 3k) to 2k might enhance quality, the smaller photosites limit the "per-pixel"-IQ (noise and DR).

 

It's not difficult to achieve large amounts of megapixels even on a given sensor-size (the sensor-architecture for a few cents used in mobile-phone cameras would result in 12-16k in a S35-sized sensor!), it's a conscious decision of engineers and has very little to do with know-how or costs.

ARRI decided for 3k to enable large photosites (low noise, high DR -> "per-pixel"-IQ) while still oversampling (3k-2k) to minimize negative effects/artifacts of the bayer-interpolation (2/3 of the color information is interpolated) and enhancing MTF/avoiding alaising (at 3k the contrast has to be zero, achieved by the AA-filter, which also reduces contrast below 3k). The goal is always a high MTF (contrast over frequency/resolution).

 

The term "3,5k" is just a stupid marketing gag, started by RED and has nothing to do with the actual quality of the 2k-output (4k is not very efficient because neither 4k of the MX or 5k of the EPIC will offer decent MTF and expose artifacts at this size).

 

Sensitivity is tricky as well. The actual sensitivity doesn't depend on photosite-size (while it has a positive effect on noise and DR) - the ALEVIII has a base-sensitivity of 800ASA which means that it reaches it's IQ-limit at this point (I have no idea how they want to obtain IQ below 800ASA) without any tricks/amplification while the RED-architecture still seems to offer 200/320ASA-base-sensitivity - maybe they can offer better S/N-ratio/DR pushed up to 800ASA than ALEXA - although that's unlikely (amplification always decreases IQ so their IQ has to be much better at base sensitivity already to offer enough reserves at 800ASA).

 

To make a long story short: all these numbers tell us very little about actual IQ.

 

RED was kind enough to give a concrete number of R&D-effort after my last "attack" but it's still a pretty closed business. We can assume that even while a lot of electronics/sensor-development is actually done by RED/ARRI/Panavision now, most of the know-how lies within the actual fabrication of the sensor and the limits in sensor-design are similar. It's propably not an accident that RED hits the very same fps-limit (60fps@3k, but windowed) as the ARRI and the quality depends more on different compromises (engineering vs. marketing, "k" vs. MTF, transparency vs color separation of filters.

The main differences lie within the camera, the handling, build-quality, workflow... ARRI and Panavision develop from the professional cinematographers perspective (although Alexa marketing is a little bit too bright for an engineers-company like ARRI) - ARRI offers 2k uncompressed, they waited to offer an electronic viewfinder till the technology was advanced enough, boot-time, thermal-management, bugs and they have their own production in a high-wage-country - it's a different business for a different market - I somehow doubt that ARRI will become a prosumer-player with a 50k€-cam...

Edited April 9, 2010 by georg lamshöft

[Peter Moretti]

Georg,

 

Thanks for the explanation and I think you bring up some very good points about craming pixels into a given sensor size. However some of Red's offerings will have full frame (24mm X 36mm) size sensors and larger, not S35. In those instances, it seems more pixels could be rather beneficial.

[Chris D Walker]

That larger sensor may not be as beneficial than people are led to believe.

 

This is a hypothetical: If one were shooting on a Super 35-sized sensor with a 50mm lens, to get the same angle of view on a Full-frame sensor they would need an 80mm (give or take). Focusing both cameras at the same distance from the subject (let's say 6') and at the same aperture (T2.8), the Full-frame sensor would capture an image with much less depth of field. The same applies for anamorphic.

 

What I mean to say is that operators and focus pullers have enough trouble shooting with wide-open Super 35-area spherical lenses before moving up to Full-frame 35. The number of people capable of shooting on formats larger than Super 35 is very small in cinematography, I don't have an exact number. Plus, how many cinematography lenses cover Full-frame? How many still lenses have an aperture larger than an f/2.8?

 

Some might say the solution is to shoot at a slower stop and push up the speed of the sensor. Doesn't that counteract how many people want to shoot; with available light and night exteriors with decent exposure? Super 35-sized sensors are the optimum for 75%+ of features; 20% goes to 2/3-inch chips.

 

The applications for shooting with Full-frame and medium format sensors has a very small niche, just like with Imax 15/65. Small productions have no need for them.

 

Personally, I like what Arri have done; oversampling for a 2K image, dual recording uncompressed and ProRes, and a Super 35 sensor among other things. It's what professionals have been wanting from a digital camera.

[Mei Lewis]

The Canon 5D2 is doing well with a full frame sensor, albeit in a different market segment.

[Peter Moretti]

Chris,

 

I agree that full frame and beyond are pretty limited in their application, not to mention lens offerings, but we'll see if it finds a good use.

 

Thinking more about what Georg wrote, I have to disagree w/ what he said in terms of adding pixels being easy. It IS easy to add more pixels to a cellphone or other still camera sensor, but it's very difficult to read them fast enough to have them be useful for a motion camera. This is why vDSLR's have lower video than stills resolution; they essentially have to ignore some pixels to be able to read the sensor fast enough.

 

Now unlike in the stills world, where who knows what the delivery resolution will be, we know motion is delivered in essentially SD, HD and 2K, Arri's sensor has enough resolution to go up to 2K, so adding more pixels will hit diminishing returns in terms of resolution and could become detrimental in other areas.

 

But if delivery goes above 2K, then Arri's sensor will jeopardize resolution.

[Adrian Sierkowski]

For some reason I am fairly certain Arri et'al realize the need to upgrade in the future. Is it my imagination going again or isn't the Alexa designed to allow future sensor upgrades? In truth, it doesn't sound like too hard of a proposition as we're just dealing with 1 chip at a certain distance behind an OLPF -v- the 3 chips of other Digital cameras... what I mean by that is, of course, pop out one chip and board and pop in the next one, check FFD and go? Seems possible to me.

[Peter Moretti]

Alexa is designed to be upgradeable to large degree, by the user, via replacing certain modules. Arri doesn't mention about the sensor, but that doesn't mean it couldn't be done if the camera were sent back to the factory.

[Brian Drysdale]

But if delivery goes above 2K, then Arri's sensor will jeopardize resolution.

 

Could a a trade off, in that the largest market for quite a few years will be on productions for HDTV which is unlikely to have a larger resolution in the near or perhaps even meduim term.

[Saul Rodgar]

RED has a 617 camera coming out soon? Pfff! That is nothing. I hear there is a 100K-pixel 8x10 stereo 3D hand held anamorphic super stealth motion picture camera under $500 being developed by a conglomerate of camera manufacturers determined to prevent letting RED have the biggest K's on the market. :blink:

[John Sprung]

Alexa is designed to be upgradeable to large degree, by the user, via replacing certain modules. Arri doesn't mention about the sensor, but that doesn't mean it couldn't be done if the camera were sent back to the factory.

 

Given the difficulty of getting the flange-focal distance right without damaging the OLPF, a sensor change would be a factory job, or done at an Arri regional facility, or at some of the major rental houses. But not doable by the end users on location.

 

 

 

 

-- J.S.

[Brian Drysdale]

Here's an interview at NAB, which gives delivery info etc

 

http://provideocoalition.com/index.php/NAB...b10_arri_alexa/

[georg lamshöft]

I can imagine a sensor upgrade is possible, but the whole sensor board (including the electronics that control the sensor) has to be changed. Desolder the sensor and re-use the filters ? No chance. Given the high cost for the necessary replacement parts, the replacement itself (calibration, dust-free-environment - all in the factory, definitely not in the basement of the retailer ;-) and remaining components of a camera without optical viewfinder (a well-made housing with power supply and cooling is all what remains) it won't be easy and cheap and therefore not a solution for everyone.

 

Higher pixel counts without smaller pixel-pitches can only be achieved with a beam-splitter design and three sensors which seems unlikely given the incompatibility with S35-accessoires (even Sony doesn't like this technology with larger sensors). The other possibility, larger sensors beyond S35+overscan (about ~30mm wide) cannot be manufactured at once, they are "stitched", which is a tricky process for professional purposes (of course the stitched "sensor parts" are not exactly the same). And you need new lenses (I don't think rehoused Full-Frame lenses from the amateur market are a good idea) - you lose MTF within the lens, DoF is reduced, you'll have to stop down further - you're turning in circles - just like with smaller pixel-pitches.

So we're talking about compromises in the end.

And of course the high pixel-count itself, as correctly mentioned, the read-rate has to be increased as well (or the number of outputs on the sensor) which generates a bunch of new problems (noise, heat, balancing channels)...

 

Right now, S35-sized sensors for oversampled 2k-output seem to have reached a point of acceptable quality for many purposes (and even an advantage in sensitivity in comparison to film?) but using four times (!) smaller photosites with about 2 stops less sensitivity and DR (not exactly, because of the lower relative enlargement) for 4k-output without even having a chance to record uncompressed 4k (although Aaton claims to record 800MB/s with a simple 2xSSD-Raid0 somehow!? Aren't 4k@16bit and 24fps already >>1GB/s and what happens at higher frame rates?) seems a bit far fetched in my opinion.

[Michael Lippi II]

I've been trying to find out but I haven't had any luck yet; does anyone know if the Alexa uses a global or rolling shutter? I've searched their website and literature I got from the DGA event but I haven't found any answers on it.

[Brian Drysdale]

I've been trying to find out but I haven't had any luck yet; does anyone know if the Alexa uses a global or rolling shutter? I've searched their website and literature I got from the DGA event but I haven't found any answers on it.

 

It has a rolling shutter, with a very quick refresh time. The skew doesn't look that noticeable in the World Cup video, although how well flash guns etc work is another question.