DSMC == FF35

[Chad Clemans]

When I hear the phrase "Digital Stills and Motion Camera" (DSMC), I usually think of stills as being Full Frame 35mm Film, and I usually think of motion-pictures as being S35. Since FF35 can envelope S35 and 16mm, B4, etc, I think of DSMC as FF35.

The announcement by Red Digital Cinema to release Scarlet DSMC's in 2/3" and S35 is great, as is the announcement to release Epic S35's -and I'm not criticizing; however, I'm wondering if what really makes "obsolescence obsolete," with respect to a DSMC, is only the FF35.

Although there seems to be pools of manifested interest isolated around each of these smaller sensor sizes, I wonder if what filmmakers want most is an FF35 sized sensor, which could use any of their lenses from FF35 on down to B4, as the base DSMC, with the option to upgrade their base DSMC with a module (ASIC) that would grant them the processing power for higher framerates, datarates, and higher resolution (Epic size it).

From my own studies in ASIC design, I've come to respect anyone up to the task, so if RED has deemed such an idea impractical, then I'm sure they know best. I do however think that if they were to only focus on designing one sensor size, it could save them a lot of effort and money. If it wasn't practical to modularly separate the extra processing power (ASIC), RED could even just sell the Epic FF35 (at the 2/3" price) and allow users to purchase, or even rent (temporarily unlock) extra processing power (via expiring firmware updates), and thereby RED could dominate not only sales of professional cinematography tools (because everyone would buy one [or more]), but control rental in that respect also (because everyone would have bought one and rentals would go through RED without hardware transactions). While some may wince at the idea of giving Epic FF35's away at the price of a 2/3" Scarlet (with the Epic sized framerates, datarates, and resolution locked out), when it comes to research and development, it's often said that the first ASIC cost one-hundred million to make, and the second one cost less than a buck. The latter idea would clearly have its own set of issues (firmware hacks, etc), which is why the former would appear better (Epic Size ASIC Module).

I don't expect this to make any waves over at RED as I'm sure they'll do what they think is best; however, I am interested to hear from Redusers on whether this is what they really want, or if the currently proposed smaller sensor sizes happen to be what they really want, why? The floor is yours-

 

[Jannard]

When I hear the phrase "Digital Stills and Motion Camera" (DSMC), I usually think of stills as being Full Frame 35mm Film, and I usually think of motion-pictures as being S35. Since FF35 can envelope S35 and 16mm, B4, etc, I think of DSMC as FF35.

The announcement by Red Digital Cinema to release Scarlet DSMC's in 2/3" and S35 is great, as is the announcement to release Epic S35's -and I'm not criticizing; however, I'm wondering if what really makes "obsolescence obsolete," with respect to a DSMC, is only the FF35.

Although there seems to be pools of manifested interest isolated around each of these smaller sensor sizes, I wonder if what filmmakers want most is an FF35 sized sensor, which could use any of their lenses from FF35 on down to B4, as the base DSMC, with the option to upgrade their base DSMC with a module (ASIC) that would grant them the processing power for higher framerates, datarates, and higher resolution (Epic size it).

From my own studies in ASIC design, I've come to respect anyone up to the task, so if RED has deemed such an idea impractical, then I'm sure they know best. I do however think that if they were to only focus on designing one sensor size, it could save them a lot of effort and money. If it wasn't practical to modularly separate the extra processing power (ASIC), RED could even just sell the Epic FF35 (at the 2/3" price) and allow users to purchase, or even rent (temporarily unlock) extra processing power (via expiring firmware updates), and thereby RED could dominate not only sales of professional cinematography tools (because everyone would buy one [or more]), but control rental in that respect also (because everyone would have bought one and rentals would go through RED without hardware transactions). While some may wince at the idea of giving Epic FF35's away at the price of a 2/3" Scarlet (with the Epic sized framerates, datarates, and resolution locked out), when it comes to research and development, it's often said that the first ASIC cost one-hundred million to make, and the second one cost less than a buck. The latter idea would clearly have its own set of issues (firmware hacks, etc), which is why the former would appear better (Epic Size ASIC Module).

I don't expect this to make any waves over at RED as I'm sure they'll do what they think is best; however, I am interested to hear from Redusers on whether this is what they really want, or if the currently proposed smaller sensor sizes happen to be what they really want, why? The floor is yours-

Chad... you are missing a huge point. Sensors go up in price exponentially as the size increases. I don't think anyone wants to pay for anymore than they need.

Jim

 

[Stephen Gentle]

Although there seems to be pools of manifested interest isolated around each of these smaller sensor sizes, I wonder if what filmmakers want most is an FF35 sized sensor, which could use any of their lenses from FF35 on down to B4, as the base DSMC...

I think that the smaller sensors will be the most popular, since they have such high frame rates and resolution (120fps at 3K 2/3") for their relatively tiny prices.

 

[Gavin Greenwalt]

When I hear the phrase "Digital Stills and Motion Camera" (DSMC), I usually think of stills as being Full Frame 35mm Film, and I usually think of motion-pictures as being S35. Since FF35 can envelope S35 and 16mm, B4, etc, I think of DSMC as FF35.

I don't know the smaller sensor sizes have become pretty popular in the stills world as well. (APS-C, 4/3rds etc)

 

[Chad Clemans]

Chad... you are missing a huge point. Sensors go up in price exponentially as the size increases. Jim

I still may be missing it, because I would have thought:

when it comes to research and development, it's often said that the first ASIC cost one-hundred million to make, and the second one cost less than a buck.

but it sounds like the manufacturing cost of the larger sensor sizes is hefty as well (beyond just the design and lab setup), despite the savings in time, effort, and money from only producing one sensor size (aside from 645 and 617). I suppose economies of scale would become a bigger factor in reducing those costs, however,

because everyone would buy one [or more]

but maybe not enough to make it profitable on RED's end. As to this:

I don't think anyone wants to pay for anymore than they need. Jim

I wouldn't know where to draw the line in the sand on what anyone needs, as we are all hopefully progressing in our projects, and whatever profits were reduced for RED upfront [by only making FF35 sensors and selling them cheap as base DSMC's] would seem to be recouped ten-fold as users purchased and/or rented higher fps, resolution, and datarates. If the higher framerates, resolution, and datarates were enabled through a hardware module, it would prevent owners from having to sell their "brain" to upgrade and make "obsolescence obsolete." Or... as to renting, I can just imagine being on set and "everything changes" and you immediately wished you could shoot 350 fps for an unplanned slow-motion shot, so you purchase a day's worth of juice through Red.com, download the expiring firmware upgrade that unlocks it, and within ten minutes RED just enabled you to make "obsolescence obsolete."

Its your camera though and I doubt anyone could do it better, so I'm not complaining nor criticizing; thanks for the clarification.

 

[Daniel Browning]

I still may be missing it

The difference in cost between 2/3" and FF35 is easily three full orders of magnitude.

That's like the difference in cost between a bicycle and a luxury sedan. It's insurmountable.

(Personally, I detest the "lock out features with firmware" idea.)

 

[Shane Nassiri]

I think your ideas are interesting in terms of an alternative business model, but what happens if someone hacks the firmware and can enable additional features and distributes it on the Internet? I'm sure there would be plenty of peeps not prone to paying $1,000s for enabling new features when they could download a hack for free.

 

[Erik Bien]

I'm no expert, but as I understand it CMOS chips begin life as round blank silicon 'wafers' just like CPUs. Only a few of the largest sensors can fit on a single wafer, so it's likely RED take up as much of the remaining space as possible with smaller chips to minimize waste.

 

[Andrew Walker]

I think your ideas are interesting in terms of an alternative business model, but what happens if someone hacks the firmware and can enable additional features and distributes it on the Internet? I'm sure there would be plenty of peeps not prone to paying $1,000s for enabling new features when they could download a hack for free.

RED is the one hacking their own software to push the cameras as far as they can go and giving us those new features for free. See this is the difference between RED and everybody else. I'm sure Canon is trying to figure out a way to charge people for the new firmware on the 5D2.

 

[David Rasberry]

I'm no expert, but as I understand it CMOS chips begin life as round blank silicon 'wafers' just like CPUs. Only a few of the largest sensors can fit on a single wafer, so it's likely RED take up as much of the remaining space as possible with smaller chips to minimize waste.

Yield is a problem with large sensors. If a wafer makes only 1 ff35 sensor but can make 12 2/3" sensors and there is a flaw in the wafer, you could still get 11 2/3" sensors out of it, for ff35 it is junk.

 

[J. Eric Camp]

You can get everything you want, except the ability to snap your fingers and fix and oversight in camera needs, with the purposed set up. I personally think it is ridiculous to purpose that an item that is projected at a $33k price point be sold for $3k with the idea of firmware rental making up the difference. Notwithstanding that the addition of the firmware lock out protocol would add to the cost of the camera and potentially even the size.

Right now you can buy a Scarlet brain and if you need to jump up to something beyond your brains capabilities you simply rent the brain you need.

Not to mention the fact that firmware unlocks can and would be hacked and David's point about manufacturing.

It's an interesting idea but I feel that it is inherently flawed.

Just me view.

 

[Chris Pickle]

The difference in cost between 2/3" and FF35 is easily three full orders of magnitude.

That's like the difference in cost between a bicycle and a luxury sedan. It's insurmountable.

(Personally, I detest the "lock out features with firmware" idea.)

Please don't misinterpret my response, but this just doesn't... sound reasonable. I know JJ said the same thing but... again. I don't believe it or maybe it is being over-simplistic or we're talking about different things. If you said it costs way more to get the larger chip to read feaster, then that sounds like a legit dilemma.

Here's my thinking... all point and shoots have very small chips. All DSLRs have very large chips in comparison. Last year's Rebel sells for as little as some point and shoots. How can they sell a camera with what essentially has an S35 sensor for $500.00

Again, this thread is on chip size not chip speed. I know the technology to get the sensor to read high frame rates puts it in another league, but chip size alone seems to be something that has been mastered.

If a Nikkon DSLR shot 24p using a robust codec, the game would be over.

Chris

 

[David Rasberry]

Please don't misinterpret my response, but this just doesn't... sound reasonable. I know JJ said the same thing but... again. I don't believe it or maybe it is being over-simplistic or we're talking about different things. If you said it costs way more to get the larger chip to read feaster, then that sounds like a legit dilemma.

Here's my thinking... all point and shoots have very small chips. All DSLRs have very large chips in comparison. Last year's Rebel sells for as little as some point and shoots. How can they sell a camera with what essentially has an S35 sensor for $500.00

Again, this thread is on chip size not chip speed. I know the technology to get the sensor to read high frame rates puts it in another league, but chip size alone seems to be something that has been mastered.

If a Nikkon DSLR shot 24p using a robust codec, the game would be over.

Chris

As Jim has said repeatedly, real frame rates cost real money. That $500 Rebel is what a 3fps raw camera? Take that to just 30 fps and you are looking at 10 times the sensor speed and 10 times the processing power. Scale costs to speed and that Rebel would be in the Scarlet S35 price range.
The consumer video features on these cameras skip enough pixels and lines on the chip to lower the resolution to 1080 or 720, so even though you get the DOF of a larger sensor, you are getting much lower resolution, and the moire, aliasing, and pixel blocking artifacts that go along with it, to get processing speed and data rates to a low enough level to work with consumer recording media and a low cost processor of limited power.

To understand just how much processing power is needed, consider that a typical moderately high end current gen graphics workstation in Scarlet S35 price range, something like an 8 core Mac pro fully loaded, can't even play full resolution 4k Red One files at real time speeds.

 

[Daniel Browning]

Things get even worse when you get above S35.

Sensor cost rises exponentially with area, but that's only up to S35 size (1.6X). That is the largest you can make a sensor with a single reticle. [The reticle is the quartz "master" image of the sensor which the manufacturer uses to focus a lazer onto the much smaller silicon wafer.] The only way to increase size larger than S35 is to stitch multiple reticles together. With a single stitch, normal manufacturing yields drop to less than 25% (!) of normal yields, so imagine how low the yield drops with two stiches.

It took a major breakthrough for Canon to finally get yields high enough that they could do two stitches of three reticles (36x24mm) for only $1,500 more than S35 (5D vs 20D). Three years later, and it's *still* $1,500 more (5D2 vs 50D), and most other manufacturers are around the same price level.

The difference in COGS between S35 and FF35 is still far, far too great relative to the sale price per unit, even for Canon, who sells millions of units per year.

 

[David Rasberry]

FF35 is a pretty small part of Canon's sales too. Measured in tens of thousands, not millions. Way less than 1%

 

[Chris Pickle]

As Jim has said repeatedly, real frame rates cost real money. That $500 Rebel is what a 3fps raw camera? Take that to just 30 fps and you are looking at 10 times the sensor speed and 10 times the processing power. Scale costs to speed and that Rebel would be in the Scarlet S35 price range.
The consumer video features on these cameras skip enough pixels and lines on the chip to lower the resolution to 1080 or 720, so even though you get the DOF of a larger sensor, you are getting much lower resolution, and the moire, aliasing, and pixel blocking artifacts that go along with it, to get processing speed and data rates to a low enough level to work with consumer recording media and a low cost processor of limited power.

To understand just how much processing power is needed, consider that a typical moderately high end current gen graphics workstation in Scarlet S35 price range, something like an 8 core Mac pro fully loaded, can't even play full resolution 4k Red One files at real time speeds.

That's why I qualified that I understood that chip speed was another matter. We were talking about chip size, and although I understand that size and speed is inexorably linked, as you wouldn't want a 3fps motion camera, I thought the Chad's questions & thoughts were valid. There is a certain base-line investment amount in R&D regardless of size. It was a good question--as the technology, chip sizes and chip speeds evolve.

I personally am not interested in FF but I often wonder how it will impact single sensor motion cameras in the future.

CHris

 

[Chad Clemans]

The sensor in the Epic/Scarlet S35 and the R1 upgrade are the same. The RED ONE can only record a certain amount of bits through it's internal pipes however, so it cannot record the full 5k that Epic/Scarlet can.

This is exactly my point!

As it says under the faq on RED.com, "Keep your RED ONE and upgrade your sensor to Mysterium-X for $4500;" therefore, one can only assume that RED can manufacture a Mysterium-X S35 sensor with sufficient read-reset capabilities for EPIC or SCARLET for less than $4500.

This then begs one of the questions I initially posed: why not make a hardware ASIC module for higher fps that could be added onto an already purchased sensor (lets say the SCARLET S35 or Scarlet FF35) so the user can upgrade without the hassle and loss on resale value of selling their brain?

(If RED just doesn't want to because it wasn't in the initial program then that's one thing, but it just seems so compelling, but what do I know...)

 

[Steven Caesare]

Sensor size (in resolution), speed, and the ASIC processing power(and power draw/heat disappation) to handle the data generated are all related. An increase in effective sensor resolution requires changes in the other areas.

Each item also costs significantly more as you move up the food chain. Including all of the necessary items in the "base model" adds to the size/expense of ALL units sold, when only a small percentage may opt to 'unlock' them, or even WANT the extra weight and bulk in their base model.

Instead, RED has isolated all those interdependant parts into a single module, and called it the Brain. You get to pick and choose your brains accordingly, and keep all your other modules (audio ins, video outs, etc...) that aren't sensor-size dependant.

It's the model that best meets the requirements of both modularity and size/weight/expense, IMO.

-sc

 

[joakim_hansson]

I don't know the smaller sensor sizes have become pretty popular in the stills world as well. (APS-C, 4/3rds etc)

True true but not in professional circles, even the prosumer market are turning away from smaller than "full frame".

In the RED sensor dilemma I would above all preferably see a 2K S35 sensor with monstrous dynamic range. That would be the perfect tool for me and a lot of others I believe, even 4k is overkill for me. Being able to shoot in low light and natural light conditions with the lens stopped down a bit is much more important than lots of pixels to me.
But maybe the 4K+ race will continue, RED is creating a market for it and pushing the other developers - Arri, Sony, p+s etc. to develop their own 2k+ cameras. Rumor has it that Arri has something up their sleeve for ibc 2009, something not in the 150k € pricerange, something with a much more attractive price than D21 as an alternative to red.

And the winner in this technology race?
-We, the present and future customers

 

[Gavin Greenwalt]

I've been informed that sensor size not pixel size affects Dynamic Range. In which case 2k would just be lower resolution, not better.