Hi all,

I've been away for some time, but I'm here again.

I don't see what's the real problem with this camera. I mean, if you can set a constant framerate and shoot lossless video to the computer without frame drops, then all you need is the best possible debayer algorithm you can find. I mean, you need a computer with the SI2K Mini as well and all the new setups with the Canon HV20 need computers to capture too.

So far the only difference I can find between the Sumix and the SI2kMini is resolution (1080p vs 2K) and specific film software and codecs. It may be a bit harder to use, but you're saving more than $10,000 to get a very close image quality and motion feeling.

My question is: Is there a real problem to use this camera for film purposes?

I've got another question. I'm planning on filming many vfx shots so rolling shutter is my main concern mainly because of the camera matching software. Is it very noticeable? Could you provide us with a quick 1-2 seconds panning shot so we can see it ourselves?

Thanks.

Is anyone here using one? Or are there some communities elsewhere with people already experimenting?

The head sounds great but without images or raw data it's difficult to justify buying one 'blind' just to have a look.

Is the API available online anywhere? Or any whitepapers or technical documents?

many thanks!
paul

Is anyone here using one? Or are there some communities elsewhere with people already experimenting?

The head sounds great but without images or raw data it's difficult to justify buying one 'blind' just to have a look.

Is the API available online anywhere? Or any whitepapers or technical documents?

many thanks!
paul

Im eager to get my hands on this camera, but before I commit to buy I want to see more examples. Right now ($2,000?) feels like a shot in the dark. Im concerned about the ERS, noise, and sensitivity. Im very impressed with the HV20. From my experience with it, its sensitivity rivals CCD and the ERS... Well it hides it so well I honestly cant even tell its there. However, I have learned that not all CMOS chips are alike. Its difficult to really get a good idea of how the cameras perform in practice based only on technical data.

If Sumix or a customer could provide any more videos, or images demonstrating sensitivity or pan movements it would certainly give me more confidence to buy. Would be pleased to see any more at all.

Anybody is tested combination of
1) SMX-12A2C ($2500) http://sumix.com/products/cameras/smx-12a2m/index.html
2) StreamPix 4 ($ price unknown) http://norpix.com/products/multicamera.php
3) NEO4K ($999) http://cineform.com/products/NeoHD.htm

It is possible to perfectly sync 4 SMX-12A2C camera heads by using StreamPix4 ?

I assume you're looking at streampix for cineform RAW encoding? I would think you'd be hard pushed to encode 4 simultaneous streams of cineform RAW unless you had a very high spec PC. (I think in the cineform forum David mentioned a core duo for one). You'd also be looking at 4 x 12MB/s recording rate, that would be 4 separate streams though not just a single 48MB/s.

Also im not sure norpix has updated or had a chance to update streampix to work with that sumix head.

But in theory it should work if you have the hardware to do it.

cheers
paul

Paul:

Why Am I saying the actual version of the camera is not quite usable?

Because:

API is not available online. Nor any whitepapers or technical documents.
The internal compresion is "something" only they know how it works and how much processing intensive it is.

Quoting Farhad:

"The catch is that SMX-12A2C does not have standard manual control. Everything must be controlled from PC side using a custom software (yet.) This camera is designed to give the total control of the sensor to the user."

How can a user have full sensor control when it is really hard to get to know how to decompress the compressed stream (when aplicable), there is no ducumentation, and the "user" cannot even write its own drivers, and need to wait an undefined time, to get a different version of a closed one?

"The only limitation is that at 1080p 60 fps GigaE bandwidth does not allow all 12 bits. You must choose a table look up to transfer 8 bits to PC at 60fps."

So again, the "total control" doesn't mean you can load "custon LUTs" or even get 10 bits from the camera, instead of the fixed 8 (linear/log?) or 12 bit linear.

So total control again means, "you can have your car in any color, as long as it is black"


"One of our customers needed LDVS output of the sensor so they can use their own interface they decided to buy our camera and use our API to control the sensor."

Yes, this is quite positive.Are they shooting high quality stuff for entertainment imaging purposes?

"This camera is as flexible as it comes, except for little GigaE bandwidth limitation. The price is low only because we decided to set it so. Our business model is to let the resale value to be above our price so our risk of inventory and production will be low."

Of coursethat it is as flexible as it comes, but it seems to me that it comes short.

Maybe after the years I really became mad, but still think SI has a clearer vision of what needs to be done.A vision Andrey and Sumix are still lacking.

PS: If someone here really believes this camera in its actual version (I repeat this because it could be easilly changed in future revisions) is as flexible as Farhad says, please let me know it.

Juan, just a couple of points:

- Farhad says it's not possible to get 12 bit from the camera at 60fps which is something the SI2K can't do either. You can have full 12bit color in 24, 25 and 30fps. I don't see the problem with that.

- Again, the SI2K mini doesn't have full manual controls either. It does have a more cinema targeted interface though. The only problem I see is that with the Sumix you have to learn how to use its interface for cinema purposes which means some features will probably be totally useless (I mean, they have nothing to do with cinema shooting), but also I'm sure it has everything we need.

So yes, I do believe the Sumix can be used for cinema shooting.

Are there any issues with GigE and bandwidth limitations? 30 fps @ 1920x1080 @ 12 bit is 88.98MB/sec which seems pretty close to the limit for GigE (which is 1000mbit/s?)

I do remember Sumix saying something about compression at the head though, unless packing 12 bits is the compression they mean?

cheers
paul

They say lossless realtime compression, so I guess real numbers will be lower than those 88,98Mb/sec and easier to deliver for the GigE interface.

there's something about the clips i just didnt like, did anyone else notice RGB noise

Jose, are you able to tell us how much CPU power do you need for decompressing the compressed stream from the head in Realtime and recompressing it again with "something"?

Mabe just recoding it uncompressed would be easier.

Does anybody know what kind of algorithm are they using in their hardware compression?

Wouldn't it be much better to have a 10 bit log LUT inside the camera head instead of the 12 bit linear?

Lots of questions I think...

Roshdi, what clip are you speaking of?

Juan, I don't own the camera yet. You mean doing it with any of the clips they've got online?

Blue color
 

Roshdi,

Our IR filter is severely cutting the blue component. Altasen sensor is sensitive to shorter blue wavelength comparing to other sensors like Micron. Recently we found out that the IR filters (Hot mirror) that we ordered to Schott for this camera are made not exactly according to our specs. We are ordering new filters and send replacement to our customers.

Compression
 

No compression is needed for 30fps 1080p 12bit.
Compression can be turned on or off with our application program or by API. presently it works only at 8 bit (that is after selecting a table look up) and it is needed if you wish to shoot at 60 fps 1080p (8bit)
Compression now is only a linear prediction coding scheme with a limited 1.7 times compression ratio. We provide decompressing formula and example code to our customers. Later we will put more details on our website.
We are working on some fancy new compression that will allow 12 bit (firmware not ready yet.)

Juan,

We are dedicated to supporting our customers with them using our cameras. Anything not posted on our website can be provided if it is needed and is not against our obligations to sensor manufacturer. We are still supporting our original camera SMX-110 with Zoran sensor. For our Altasens camera we need to get individual permission from Altasen to provide information in data-sheet of sensor to specific customer. I hope they will allow us to post a gleanings of the data-sheet on our website.

10 bit table look up will be added later.

This camera is designed to be flexible and simple. everything that can be done on the PC side must be done on the PC side. Other interesting features like continuous changing of frame rate, frame by frame, is done on camera side.

are there heat problems with the casing? (there doesn't seem to be ventilation and i thought sensors get quite hot)

Paul, So far we have not seen any problem with heath. Camera is very stable indoor or in below freeze outside. Howerver, the body gets warm.
There is a heath sinking design that transfers the heath from sensor board to the Aluminum alloy solid case.

Would a laptop work as a capture station, what are the real minimum specs? (there's a wealth of small form factors PCs and ultra portables out there).

Some laptops are excellent and some do not allow full bandwidth of GigaE. We had good results with IBM, and Gateway. I think any newer laptop will be ok.

What is the data rate down that GigE?

We are using 92% GigaE


What are the realistic time scales for these being readily available? (especially with the new altasens sensor - the one in the SI 2K)?

We are making 80 more with the same Altasens 3562. There is no advantage of going to 4562. Later we consider Aatachrome 2/3" and perhaps new Sony sensors.

Is the RAW data from the head standardised, so if a new head came along the software would be the same?

Almost. Some features will be different from sensor to sensor, all else will be identical.

What are the real issues in pulling something together that works in the field?

we are too busy with making camera heads. we give you tools and support to do the integration for something that works in the field. This is not our expertise yet.

cheers
paul[/QUOTE]

>We are making 80 more with the same Altasens 3562. There is no
>advantage of going to 4562. Later we consider Aatachrome 2/3"
>and perhaps new Sony sensors.

Farhad,

Im following your progress with increasing interest.

Can i ask which sony sensors do you mean?

You mention a LUT being available later on, does this mean you can update the cameras firmware once in use? I assume there's an FGPA in there and perhaps you have a method for updating the code?

Still waiting for images and movies.... :)

cheers
paul

Paul,

The Sony sensor is IMX021.

A LUT that can be set by user for 12 bits to 8 bits is already available. I mentioned that later we will add LUT for 12 to 10 bits. This we will do by interpolating, so 10 bits LUT will take the same memory as 8 bit version. User can only specify the LUT by 256 points, either 10 bits or 8 bits.

Yes, FPGA and programmer controllers are used inside the cameras and firmware is updated automatically with new camera drivers.

We just got some new IR filters. It will take some time before we post new sample videos. Altasens has extra ordinary color distinction. Using a correct filter makes all the difference. both blue and red are much enhanced.

Farhad

I'm so glad to read that, Farhad. We've got everything ready. You'll receive my order within the next few days. Just can't wait to test the camera!

>The Sony sensor is IMX021.

You mean the APS-C sensor? Is this for machine vision (because it looks to be limited to 12fps)?

Although it would be very interesting if, by binning, this rate could be increased to have lower, more sensitive HD resolution. An APS-C size sensor in one of these things would open up the lens choices enormously.

Of course i'll take an HD 2/3rds for the time being once i know it all works :)

cheers
paul

Paul,

Yes, better focus on what we can sell now. Other sensors will be available. We make sure upgrades to new sensors will be a painless as possible.

And don't hesitate to post sample images and videos here :)

Jose,

Did you ever acquire the single board computer that was being discussed in your older thread and will you be testing the Sumix on it? Or will you be testing on a standard desktop or laptop?

I look forward to seeing some sample videos from your testing!

For now I'll test it using my laptop. In a couple of months I'll buy a Mini-ITX based computer and a couple of HDDs so I can have an independent camera.

I've got news... Not about the camera, sorry, but I just received my new Computar 2/3" C-Mount zoom. I have to say I'm quite impressed with it. I haven't been able to test it yet, since I don't have the camera, but it looks fantastic.

It's a 12,5-75mm F1.2 zoom and those are 35mm equivalents, so they're real measures for 2/3" sensors. It looks even more solid than many of my canon EOS lenses, lol. Kinda like the feel of an old FD lens but brand new.

To those of you who question the option of using C-Mount machine vision lenses for film shooting, I have to say this lens has removed all possible doubts from my mind. The usual complaints about these kind of C-Mount lenses are small diameter and fixed positions when adjusting the iris (in most lenses you may notice a little "click" when going from one position to another). In this lens, the three rings are very smooth. And I mean VERY. On the other hand people complain about c-mount diameters because they are too small to control the focus. This lens is closer to 35mm in terms of size, so focusing is not a hard task. In fact I was really impressed with the size. Having used the lens from my first Micron demo board, I expected this one to be much smaller and definitely not this heavy. This' going to look fantastic with a followfocus and a mattebox.

As soon as I get the camera, I'll post more things about the lens.

I've been testing a number of lenses on a camera here to try and understand real world results. I've got 16mm lenses, EF-cmount adaptor and a 25mm f1.4 fujinon machine vision lens.

Bottom line is im impressed with the fujinon, good feeling, 180 degrees of focus, nice image - bit soft on the edges but on a 2/3rds you probably wouldn't see that (im using a 1" sensor). This is compared to an old T2 cooke 25mm, switar f.16 10mm and various canon lenses (10-22mm works nicely for example)

I would love to see some images from that computar, it's on my hit list too. Would like to see how soft at f1.2 vs f4 for example.

My feeling is that the glass in these lines probably isn't *that* different to fujinons digital cinema offerings. I mean there's probably 20x the price difference and certainly not 20 times a better image

cheers
paul

I will be getting the SMX-12A2C camera this week. Our production team is meeting up this Saturday which will be a great opportunity to test the camera.

We have a short film project coming up in the next 3-4 weeks and if ready, I plan to use the SMX-12A2C. I will start putting together an interface and writing software for it on Sunday.

Paul, as soon as I get the camera, I'll start testing it and posting as many clips as I can. Let me know if you want me to do any other specific test.

I also think the quality of these lenses is quite close to digital cinema standards. If you think about it, tack sharp images are a priority in machine vision systems. In fact I think these lenses lack the organic (and a little more soft) feeling you get with old 16mm lenses, which for me is also very interesting. I've read things about old Angenieux super 16mm zooms which can capture very beautiful images with a great balance between sharpness and softness, impossible to get with new lenses without post work. But anyway, I prefer to have my clips as clean and sharp as possible and then get them dirty in post.

And Daniel, it's great to know there will be more independent software projects for this camera. Can we expect a public release for people like me who can't write a single line of code?

I also hope I can get my cam soon. I've mailed Farhad asking about it and got no reply yet. I paid for mine about 10 days ago and I just can't wait to get my hands on it. How long did it take for you Daniel?

I plan to release the software as open source, but it may be a while till then. Right now im going to concentrate on getting the camera ready for production.

It took a little while. I first contacted Farhad with my intent to buy about 25 days ago. Everything was settled about 14 days ago.

The camera was shipped this Tuesday and I got the invoice a little later. It just came in earlier today. Well not quite... FedEx could not find my home so im going to pick it up at 4:00 from the location in town.

If things go well I will have some images/videos to post here soon.

It's already there!! I can't wait to see clips!

Ok, so if I paid 10 days ago... in about 5 to 10 days more I may have it here (if everythings goes ok).

Please post something as soon as you get home with the cam.

Ok, I got a mail from Farhad saying they'll ship the cam next monday. Can't wait!!

Hey Jose,

Is this the lens you got?
http://www.avsupply.com/details/m6z1212.shtml


I'm seeing prices all over the place, so wanted to know what you paid for it too.

I might just be with you guys if the footage looks good.

Hi Samuel,

Yes, that's the lens I bought. I paid $297 plus shipment and I find it quite cheap since I've been told this' one of the best possible choices for 2/3" HD sensors in terms of quality/price. I know Fujinon is even better, but they didn't have anything like this, so the options were this zoom or three Fujinon primes. Personally I prefer a zoom, so I can have every single focal lenght from 12,5 to 75mm.

Before anyone points it out, I know there're very high quality super16mm lenses from old cameras out there giving better (or more cinematic) results at similar prices or even cheaper. Let alone the ones which are even more expensive. But I also knew there're many people trying to sell crap and since I don't know much about old lenses, I wanted to start with this one. My next lens will probably be an old Angenieux zoom and I'll ask a lot before buying.

I'm replying here because I feel its a more appropriate place.

Note that my camera was shipped with the old, incorrect filter. It cuts too much blue and red, I should be getting the new filter from Sumix sometime soon...

Meanwhile, any results I post should not be used to judge the overall capabilities of the camera. In addition to that, i'm still just learning how to use it so my white balance and other settings are likely to be incorrect.

Here are some test grabs from earlier today. Each is accompanied by a quick auto levels, contrast, and color (cc in file name).

http://www.dreamstonestudios.com/per...s/FRAME_24.jpg
http://www.dreamstonestudios.com/per...RAME_24-cc.jpg

http://www.dreamstonestudios.com/per...s/FRAME_30.jpg
http://www.dreamstonestudios.com/per...RAME_30-cc.jpg

http://www.dreamstonestudios.com/per...s/FRAME_34.jpg
http://www.dreamstonestudios.com/per...RAME_34-cc.jpg

Not aiming for anything artistic, did not have time to do a big setup. Only lit with 2 150w Fresnel lights. Had to turn the gain up considerably.
Used a TV zoom lens.

You're right... Too much green. But it looks promising... Way better than any HDV cam, as expected, and closer to a DSLR image. What kind of tv lens are you using? It looks extremely sharp! (Notice the upper leafs in the second and third image)

I'd like to see a DOF test with a focal lenght of 50mm so we can see what are those 2/3" capable of.

Other than that it looks like we've got a really good cam here. As with all digital cams, it has little problems with highlights, but that's something we're used to and can be retouched in post.

The absorptive filter on this camera takes out more than 60% of blue light that the Altasens sensor is sensitive to. We have been experimenting with two other filters. Although they are much better, they are still not satisfactory. One is Reflective IRC30 which has narrow acceptance angle not suitable for small F# when used inside the camera. The other one is absorptive IRC20 which is still quite thick and not optimal. We either solve this problem or you must use IR cut filters in front of your lenses.
We are now attempting to design a reflective filter with several coating layers that will accept sharp angles with a final hard coat that would allow easy cleaning. This may take a while before we can produce such filter at low cost.
Would you help us to decide on parameters of a basic lens that you would find useful--such parameters as FOV, F#, focusing range, depth of focus at different focuses? Putting the filtering inside the lens will make it easier as it can be placed where light rays are most parallel.

So which one is the filter you're including with my cam?