sumix SMX-12A2C fullHD 60p

Also, talking about sensitivity is not moot, because there is a 100% limit in light gathering for each pixel. If you believe the 2/3" sensor will exceed physical reality by a factor of 4 or 6, that's your problem and being an engineer, I don't want want to discuss it further.
 
A simple explanation on light gathering:

Camera A with proven high quantum efficiency sensor technology: 2/3" f1.4 lens, 1920x1080 pixel=2073600pixels

Camera B with unknown sensor technology: 2/3" f2.8 lens, about 3072x1728=5308516pixels

Both have the same 2/3" area so light gathering per pixel in camera B with its fixed f2.8 is 5308516/2073600*(f2.8/f1.4)^2=10.2 inferior if we ignore quantum efficiency.

So the pixels on the scarlet get less than 1/10 the light per pixel. That's more than 3 stops. The Sumix is already operating on a very good quantum efficiency, close to the limit for CMOS techology really, so I predict with this solid and simple calculation the Scarlet will be 3 stops slower than the Sumix with f1.4 lenses. Scarlet could cover the distance by having zero read noise, unlimited dynamic range, or quantum efficiency of 800% or 1000%. This is technically impossible since 100% is an actual physical limit for these technologies.

There is no reason to ignore physics just because Sumix do not have a good marketing department.
 
Last edited:
Dpreview DOF calculator results, it doesn't have the precise apertures and focal lenghts I needed and relates to slighly narrower 1/3" sensors (photographic aspect ones) but I got close:

1/3"35mm f1.4 from 2m, near sharp 1.980 far sharp 2.021
2/3" 70mm f2.8 from 2m, near sharp 1.982 far sharp 2.018
crop factor slr 200mm f8 from 2m, near sharp 1.984 far sharp 2.016
full frame slr 300mm f11 from 2m, near sharp 1.986 far sharp 2.014

Ignoring calculation errors of the extreme scenario, all have similar dof despites the small differences in focal lenght and aperture compared to my results on the previous post and provide identical angle of view and coverage. Those are practically identical imaging systems from a physics point of view, but the smaller sensors get a light gathering advantage when achieving identical results. More is not always better for various practical reasons including sensitivity. You need to take into account pixel quantum effiency and number of pixels (pixel area). Then there is the out of focus character and quality issue, but that's not down to absolute physics and simple calculations just yet:)

This method can be used to compare any two cameras with different aperture and sensor size. Just multiply relative aperture by sensor size ratio to get the identical dof info.
 
Last edited:
A comparison of two sensors, one 1.6x larger than the other. Comparing 1/3 to 2/3" which is 2.0x larger does not change anything.

full frame 50mm 2.2
vs 1.6x crop factor 35mm f1.4

http://bp3.blogger.com/_LMV40M_lPGA/R3MEFbN1BGI/AAAAAAAAACk/kLUJ3iUqYe4/s1600-h/5035-full.jpg

another example:

http://bp2.blogger.com/_LMV40M_lPGA/R3MP2LN1BJI/AAAAAAAAAC8/mTrqAOCZO2o/s1600-h/13585blur.jpg

So, by using 1.6x smaller focal lenght and 1.6x smaller relative aperture, you get identical results. This proves the Scarlet 2/3" f2.8 will get 1/3" f1.4 results in shallow DOF.
 
Last edited:
Alright, I learn something new everyday. :)

But what's the workflow for this? Are you tethered to a macbook/pc? How practical is this? Can you shoot in uncontrolled settings? Or is this strictly useful for a studio situation and not clients? And where can one get these "megpixels primes."

I'm very curious about what kind of shooting people are accomplishing with this.
 
I haven't got a Sumix but I have used boards that use the same sensor and other cameras of the type and I'm building a camcorder that can be made compatible with this head although it is more oriented on CCD solutions.

Sumix takes a regular pc to work, just like the Silicon Imaging Mini because you need storage monitoring and control which a camera like this lacks, but this doesn't mean it has to be big or use too much energy. It's just not practical at this time for Sumix. but we can actually get smaller form factors compared to the professional camcorders of the same sensor size. The Silicon Imaging original camera was based on a head similar to the Sumix. Sumix is not ready for production because it lacks many features at this time such as a color matrix for accuracy and an organised workflow, but we are not far from puting these cameras in a user friendly production ready camcorder. Although it's not something you would get to shoot weddings with, it is similar in practically or better compared to recording systems of the same quality and far more affordable, so it will do the job for feature lenght filmmaking.

The megapixel primes come at 2/3" and 1" formats and are fully manual. In machine vision and control, very high optical quality corner to corner is required because algorithms are applied on the video to detect precise mechanical features. Being primes, the lenses are very economical when they do not come with motors. So these lenses come with modern design, but you have to choose a lens class that fits the pixel size. For a 2megapixel camera like the Sumix you need at least 1.5megapixel 2/3" lenses, preferably 5 megapixel lenses such as the Fujinon 2/3" 5 Megapixel Series at this page:

http://www.1stvision.com/lens/megapixellenses.htm

Since these cameras are single sensor, they do not require 3ccd design (unlike B4 camcorders), they can use any 16mm and super 16mm lenses with simple non optical adapters.
 
Interesting lenses, any idea on price? I checked the site and can't find any key to prices.

Is your camcorder build for the Sumix in prototype mode already? Are you shooting with it and getting good results?
 
http://www.rmassa.com/lenses.htm

A disadvantage of these lenses over cine 16mm lenses is that the barrel is quite small, some are as small as the Bolex prime lenses. The Fujinon ones are larger and have barrels of 50mm so they are on the limit of being usable with follow focus units. The advantage is that they have excellent corner to corner resolution even wide open at very high MTF.

I did the first working prototype in december and posted images and some videos. It is a 2/3" 768p ccd model. I have also tried an 1" 1920x1080 ccd with it. I have tried the sensor that is used on the Sumix on an implementation from another manufacturer. The resolution and sensitivity is ok but it wasn't up to ccd saturation without an agressive color matrix and it comes with rolling shutter and inferior low light compared to ccd. I am considering Sumix 2/3" as a more affordable alternative for those who do not need CCD performance, so it might get to be in our entry level model. Generally I'm not a big fan of CMOS although it is becoming more acceptable today that manufacturers are abandoning ccd in anything but their most professional camcorders in order to reduce cost.
 
Hey,

So I'm still not sure what you're recording to? Is your prototype a self enclosed computer for recording?

Can I see footage links for your prototype?
 
Played with those lenses too. They are nice and sharp (very close focusing distance) but the sensor size( 2/3") comes with the known look (I am reffering to the DOF). The focus is precise but those lenses are made with a certain purpose in mind (and they deliver!) but for constant back and forth..... as required in cinematography....I didn't like them all that much (kind'a stiff)
 
The whole system is based on a battery powered computer and records in removable 2.5" media, based on solid state or hard disk, but it's really open to practically anything that can connect to SATA or GigE. Computers these days are not what they used to be. My present prototype is smaller than an F900, consumes less power and has similar weight:)

Clips from the 2/3" model on the January prototype. It was a simple straight gamma and fully uncorrected at that time, now it's an advanced film response model with a proper color matrix. About 2 stops extra dynamic range.

http://www.archive.org/download/Extinq/extinq.wmv
Lit by two tiny candles, one flickering in the process.

http://www.archive.org/download/3CandlesToDarkness/CandlesToDarkness.wmv
Demonstrating black purity.

http://www.archive.org/download/TwoBottlesInAndOutOfFocus/bottlesinoutoffocus.wmv
Pulling focus on a cheap cctv lens, lit on available light (a single 60watt incadescent). I think I got video levels wrong on the export, so it clipped shadows and highlights.
 
The Sumix camera does require a computer for operation, similar to the SI-2K Mini. I built a prototype system with a small computer on rails with the camera & a 7" touch screen. It was all pretty compact, about 12 inches long and weighed ~7lb without batteries.

It was fully contained, and more or less point and shoot. I was able to walk around with it, take it outside, and the only setup was putting it on a tripod.

Sumix does have plans for a standalone camera, but I have no idea of what kind of time frame. I'm waiting on updated software myself, as well as a new color filter for my camera. Should not be too long now.

Right now, with the incorrect filter, I'm losing a lot of color information. My images are desaturated, and far from what they should look like. Nevertheless, here are some more recent stills I took a few weeks back.

http://www.dreamstonestudios.com/delete/boat1.jpg
http://www.dreamstonestudios.com/delete/boat2.jpg

I believe the lens was a 50mm Minolta, set at f/4. It was hot glued onto a C-Mount -> Canon FD converter. (I had no canon lenses to use at the time)

I don't have any videos to share at the moment.
 
I have been looking, but can not find a .raw file from the sumix camera. I would like to download a sample to see the process of using a raw file. can anyone hook me up?
 
Would this camera work with a fast laptop? Maybe even a super fast tablet that could act as a monitor as well as capture device?
 
> I believe the lens was a 50mm Minolta, set at f/4.
> It was hot glued onto a C-Mount -> Canon FD converter.

Awesome. Do you have pics of that?
 
zcream said:
Interested how this camera faded out...
Click to expand...

The problem with small companies that already have other products on the market is that they never focus and invest enough time. Integration of a full system is a full time job for a small team of engineers for a minimum of 2 years. It's quite an investment in time and money and it's hard to even consider it when you are already busy with other products. In any case, you can't expect the user to handle integration, the user expects a turnkey system. Anything less is simply not worth the investment in time.
 
SUMIX camera is no good

SUMIX camera is no good

They have a plenty of camera troubles probably due to basic technology lacks.
The rumor said their trouble rate is very closed to 30%, some tech said.
Do not use for your :zombie_smiley:business.
copy.png
trans.png
 
You must log in or register to reply here.
Back
Top