Why not release mini SI (camera head only) based on good 1/2,7"-1/3" cmos image sensor ?
We can see now how good image from Canon HV20 1/2,7" sensor.
In most cases HV20 with 35mm adapter and HDMI output to CineformHD is very comparable to SI2K samples :)

P.S.What business is better : sell 10pcs of SI2K at $12K or 1000pcs (1/3") at $2K ? ;)

Personally I'd think a version with a larger (35mm) sensor would be more interesting. Sensor size or more to the point photosite area, is the most significant parameter affecting image quality. Also using a smaller imager is not going to significantly impact manufacturing costs. The costs of machining lens mounts isn't going to change and the price point that you're talking about is what a lens mount alone costs.

Bob, i agree with you about 1" sensor.
But in 35mm digital pro area more cams and strong competition:
Panavision, Arri, RED, NoX...

SI need to make full range of camera head: 1/3", 1/2", 2/3", 1"
Time to have products similar to oldstyle 8mm, 16mm, 35mm, 70mm :)
Instead of film celluloid we have now the digital negative -CineformRAW.

About price and quality...Bob, look at samples from Elphel camera.
Only drawback of this fpga camera is compression.
With CineformRAW this modular tapeless camera can kill those hdv/avchd camcorders ;)

I'd love to see a 1/2" (or maybe APS-C sized -- 24mm x 15mm) SI mini with an HDMI port that could be used with the proposed Cineform CF flash recorder. Bring it in around $5K?

I'd love to see a 1/2" (or maybe APS-C sized -- 24mm x 15mm) SI mini with an HDMI port that could be used with the proposed Cineform CF flash recorder. Bring it in around $5K?

It sounds attractive however the MINI head doesn't do the de-bayering or much in the way of image processing. It feeds out RAW data to the recording device. About as close to a digital negative as it gets.
Converting that digital negative into something that could be encoded into mpeg-2 would destroy all the advantages of recording RAW and require some serious processing. That processing needs a lot of power to be done in real time. That power means heat and big batteries. This seems to be the stumbling block at the moment for the industry as a whole. Advances in silicon might change this in the medium term.

Converting that digital negative into something that could be encoded into mpeg-2 would destroy all the advantages of recording RAW

I thought the SI recorded to Cineform codec, not RAW. Am I confused?

you are right and wrong. SI records to Cineform RAW.

Ahhhh... now I get it. Thanks for the clarification.

Hi Everyone,

Good thread . . . yes, the "big deal" is that the Mini doesn't have on-board recording and it doesn't have on-board control surfaces either. There's a couple TTL inputs, etc., but that's not enought to give you comprehensive control over the camera like we give you inside of SiliconDVR.

While theoretically a pre-debayered HDMI port version could be constructed with a FPGA implementation, but it would 1) not be RAW, 2) probably not support the "fancy" stuff like 3D LUT's, etc. without a bigger FPGA, and 3) the CineForm device would not give you any control over the camera as it's only a recorder, so there would have to be a control surface of some type.

Adding "intelligence" to the camera head in the form of another FPGA, with then the associated graphic overlays or feedback screen, or something of that ilk starts getting things a bit more complex. I know technically it would be doable, but not for $5K, at least not the way things stand right now, especially with the physical construction of the camera head (it's machined, not cast, so per-device prices are not cheap).

Thanks,

Jason

Jason, do you know about Sumix camera head based on same altasens sensor as SI2K and coming soon Cineform SOLID DVR?
Both devices are anounced at price less than $2K each.
IMHO, same silicon of SOLID can be used to capture CineformRAW also. Shoot to digital negative (Cineform RAW) same way as film but with mp4 or mpeg2 low res proxys for preview. All LUT's in post...

I'm very aware of the Sumix camera. They do some very nice work and sell at a great price. I've also been participating in the CineForm thread on this site, so I'm aware of their developments as well.

Some notes I see from the current developments:

1) Sumix's camera is not using the same sensor as us at this point in time.

2) I don't see a HDMI output on the Sumix or a Gigabit port on the CineForm recorder, so there will have to be some device to mesh this disparate connection/protocol/signal mis-match.

3) The CineForm recorder does not actually control the Sumix camera head, so even if there was a way to record the RAW stream output to the CineForm SOLID, there is no way to actually send commands to the camera head to change settings such as shutter, frame-rate, resolution, etc. The Altasens sensors are quite configurable, and there would need to be some separate serial interface, or specific development on CineForm's part, to control the camera head. Since each camera head typically "talks" a different serial control language based on the way the serial interface is implemented, this would make the support for the Sumix head extremely targeted, and therefore volumes will be low compared to the generic SOLID that CineForm is working on right now. Either that, or someone will have to work on some type of break-out box or CCU-style unit that will take the gigabit input, output RAW over HDMI, and then offer a control interface for the Sumix itself. This new device won't be "cheap", especially if volumes are low. I would estimate another couple thousand. But now you have three pieces to the puzzle and a lot of cables, etc. In fact, I bet it would probably be cheaper for whoever would make this CCU to simply license the SOLID electronics from CineForm and then build a single box just for the Sumix.

The point I'm making is that for the quality people are wanting, at this point in time, a total cost of less than $5K for the camera and recorder is not feasible without compromises being made somewhere along the line either in the recording codec, the mechanical quality, or the comprehensive feature-set.

I think someday this sub $5K price point will be realistic, but not right now. Not when you have multiple devices involved. It will probably take an all-in-one device with higher volumes to get the price-point down that far.

I like to dream of super prosumer priced cameras, but lets get realistic.

A complete large sensor HD camera below about $10,000 is impossible for the units we are discussing here. And if you try to get this to $5,000 you can't afford to use CCD and the software and development will have to be free. It will not happen. I have spend quite a lot of time trying to find a way to make it possible at that cost and there is really no way today.

Even if it becomes possible in the future when every part (except ccd) will cost 1/2 as much, this will not happen for marketing reasons. In 5 years you will still have to spend $8,000-$10,000 to get highly compressed CMOS based camcorders with some manual control and the sensor sizes will be 1/4 or 1/3". Even now, large sensor CCD and 3x1/3" sensors is considered very professional. Who would put a superior system against those? A dreamer and a bad businessman I guess.

A $3,000 portable uncompressed HDMI (also capable of good lossy codecs at extra cost) HDD based deck will probably appear soon. But what are we going to use with that? 1/3" tiny pixel cmos or 3x960x540 1/6" or 1/5" ccd? These will gain practically nothing from the better recording quality. We are talking about $4,000 to $8,000 cameras for using this recorder and it will still be tiny 1/3" CCDs and small noisy 1/4" CMOS. The total cost will be $10,000.

What about workflow cost? Storage, capable HDD for multitrack, capable cpus, software, etc?

Anyone realistic with $30,000-35,000 in his pocket can go for an acquisition/editing package and actually get quality video.

I also don't understand the point of looking for something that does not exist and probably never will. There are many solutions on the market and some of the affordable ones produced ok results even for feature filmmaking. The camera is never the limit in creative uses of video.

2) I don't see a HDMI output on the Sumix or a Gigabit port on the CineForm recorder, so there will have to be some device to mesh this disparate connection/protocol/signal mis-match.

Yes, but you can have the sumix mounted to a custom made box with a computer inside and record the same quality of footage as the SI2K. This was actually how the SI2K started and what it actually is in it's core. A camera strapped to a computer. Will it be the same in terms of usability as the SI2K+DVR? No. But it will cost less than a 1/5 and get you the "same" quality of image, resolution and post flexibility.

Getting from the point of writing images, to a dependable video camera with controls that make sense and behave properly, is a very long journey. You can get software that writes video to a capable storage subsystem at very low cost and camera heads are from $1000 to $10000 depending on sensor size and technology and most come with recording software and SDK. The debayer will usually be a joke, it might have serious implementaation issues (bugs) or be unusable in real time preview. Sometimes the debayer will loose half of the sensor resolution vertically and horizontally, the frame rate will be innacurate and practically anything except what you had in mind, you will have no audio and it will not be portable or battery powered or anything resembling a camera in actual use. There are lots of engineering issues and these have to be solved by someone. The same applies to GUI, software, codecs etc. You can easily spend 1/3 and get nowhere and 1/2 and get close with numerous usability issues after hundreds of hours of testing, ordering components and development. Time is not free. Development cost of a working camera with off the shelf components can be 100,000s of USD in engineering time. The markup in thse devices represents exactly that, engineering time and the assurance that it will do the job. Most people will not pay 1/3 for something that will not do the job, because that 1/3 is money down the drain. That's what we call turnkey. Engineers can always build something to do the job, but no sane man can normally justify the cost for building only a couple of cameras, even if he/she has the engineering capacity.

Personally I'd think a version with a larger (35mm) sensor would be more interesting. Sensor size or more to the point photosite area, is the most significant parameter affecting image quality. Also using a smaller imager is not going to significantly impact manufacturing costs. The costs of machining lens mounts isn't going to change and the price point that you're talking about is what a lens mount alone costs.

Hi Bob,

Size is generally not connected directly to performance. The majority of development money from sensor manufacturers goes to small and medium sized sensors. This has a nice result, 1/3" interlaced sensors are leaping in performance once in a couple of years, but larger sensors show little progress. The 2/3" sensor could have 60 ot 70% of quantum efficiency and the larger sensor will have 20% on 25% in some series. This could easily be 2 or 3 stops of difference in the blue color of a bayer sensor. Larger pixels are usually more noisy. The only advantage you will positively have is shallow DOF capability with a disadvantage in low light when you need the DOF to get the shot. And that's if you can afford lenses with the same relative apperture with the ones you are using in the small sensor.

Even if the large sensor comes with identical resolution and you increase size in both directions by 2x, same as going from 1/3" to 2/3", the theoretical advantage is only 6dbs of signal/noise performance and it is usually reduced by the smaller quantum efficiency of the large sensor and the availability of lenses with large apertures. You can get f1.0 lenses for up to 1", but above than you have to use very expensive lenses with much smaller relative apertures. The large sensors have advantages in light gathering but they also have a lot more dark noise so they need the light to saturate and get a good signal per noise performance. And if you are stuck with f2.0 lenses on the large format, a pixel that is 1/4 the area on a sensor with identical resolution using f1.0 has identical light gathering and shallow DOF capability.

Once you go over 1/2", the difference in the performance of larger sensors is not that impressive.

Yes, but you can have the sumix mounted to a custom made box with a computer inside and record the same quality of footage as the SI2K

While this hasn't been discussed much, the SI-2K is actually quite a bit more than simply wiring up a CMOS sensor to run at a certain clock and then recording whatever comes out the other side direct-to-disk. There is quite a large imaging pipeline to perfect the image quality from these sensors that we have developed over the course of the past two years, optimzed to achieve very high-end image quaility that has been honed with lots of user feedback and a lot of in-house proprietary development. This is not to say that others cannot accomplish the same task, or as a put-down to other manufacturers who I'm sure are working very hard to make their products a success, but suffice to say there has been a lot of development effort poured into this product at the hardware and imaging pipeline level.

Also keep in mind that just because two cameras might have the same sensor does not mean they will have the same image quality across the entire spectrum of scenarios that typify the cinematography experience. To get that last 10-15% of performance can require months, and months, and months of work, and re-working of back-end algorithms, methodologies, and workflows. So what you're paying for in the SI-2K is us having gone over the image with a fine-tooth comb and eeked out every little drop of performance we can get from our sensor. I have found the Altasens senors to be a superb platform to start from for achieving the goal of a top-quality digital cinema camera, but the sensors are only the start. Lots of man-years have gone into the camera product that you see, both on the imaging aspect side, and other portions of design that all come together to make a single "camera".

Now I don't want this to sound like a "we're better than everyone else" thread, because that's not what I'm saying, and I think it would be a dis-service to the other members of this community who are working very hard on their own designs. The point I'm trying to make is that we feel the pricing for the SI-2K is valid based on the investment that we have made in the product, both at the imaging end, recording technology, mechanicals, optical precision and accuracy, and other aspects of a camera that need to be in-place in order to have what we deem to be a professional product that meets our users goals for a digital cinema camera.

Lots of food for thought there. I can see why most of the R&D dollars are spent on the smaller sensors. The cost of fabricating chips seem closely related to die size and the smaller chips in smaller cameras are where the high volume / high profits lie. What I'm uncertain about is why quantum efficiency is dropping with sensor size. With photosite size maybe but then again going from a 1/3" SD imager to a HD imager to retain the same photosite size you need a 2/3" sensor size. You've only got to look at the difference between say the PD170 and the Z1 to see this effect.
The other issue that you've rightly touched on is the availability of suitable lenses. I don't see any 1/3" cameras with T1.3 lenses however that might be a catch 22 issue. No one makes fast HD lenses for 1/3" cameras because none have interchangeable lenses and those that currently buy such cameras would baulk at the price of such lenses.

Sony just released some new 1/3" sensors that are have about 4 times higher sensitivity than the sensors on a PD170. Those are actually the faster sensors on the entire sony line, including 1/2" and 2/3" interlaced used in eng cameras! We will never see those in video cameras though, because affordable cameras (<$15,000) will soon be cmos only and HD has become the norm.

Non 3ccd primes can be built quite cheaply. 1/3" has quality issues but 2/3" and 1" are great. About 300 dollars will give you a very nice metal construction and modern glass, low distortion, multicoating and good aberation correction with fully manual focus and iris without steps. The cost issues start when it has to be 3ccd, it needs to have motors or a zoom ratio. That takes a $300 prime to $5,000 easily while reducing its optical performance.

There are good sensors with large pixels, but the cost gets out of hand. In these designs you can't afford to spend more than $5,000 to get the sensor to a working camera block including development, and that does not leave many options. The sensor itself has to be cheaper than about $1,000 or $1,500, capable of video rates and smaller than super35 size. With that kind of money you can get 1 or 1.2" Kodak CCDs or CMOS of larger size which is cheaper to manufacture. But nothing exotic.