Why couldn't somebody take a 7D/5D/GH1/D3s/etc., remove the sensor and build a camera

[Andrew Thomas]

from it? I see all these DIY camera projects using (seemingly) much worse performing sensors. Is the way the sensor connects to the "computer" proprietary? Would it just take a lot of reverse engineering? It seems to me the sensor in the 7D for example is capable of a heck of a lot more than the rest of the hardware in the camera will allow. What might stop somebody from ripping that sensor out, providing it proper cooling, and recording the raw output to a computer?

I really have no knowledge here so be nice

 

[morgan_moore]

I have been trying to get my head round this

if a D3 can do 13FPS in crop mode stills its close to 24

surely a dslr maker could somehow port that data out to a second box which would be a raw data recorder and could also handle sound properly and maybe even provide the monitor signal

Trouble is that box might cost red money

Surely the limit of DSLR frame rate is largely affected by the mechanics of the mirror/shutter rather than the data rate -Locking those mechanics open would solve that surely ??

But it seems not

Seems that these sensors for some reason I dont understand need a shutter

(I think it is to do with the actual read speed being slower - just like a fast still frame is created by moving a slit across the film plane)


Such a device woulde therefore AFAIK need a cine style shutter and a huge recorder, else great compression or all three

All this tech is begining to sound familiar .. some one is indeed doing it .. Red ?

 

[Jannard]

The native sensor frame rate tells you the story. To double the frame rate, you need to cut the top to bottom in half, which gives you about 1/4 of the overall resolution. What we are building is COMPLETELY different than what the DSLRs are currently doing... and in more ways than just the frame rate.

Jim

 

[Wesley Scoggins]

Pixie dust I presume.

 

[morgan_moore]

Im still confused as to how a DSLR can capture 1/250 th with out have a read time of one 1/250th and there a potential framarate of maybe 120 assuming write=read

I pick the 1/250th number because that is the flash synch time where the shutter is fully open

I understand that shorter still exposures are created by sweeping a shutter slit across the lens (which even creates 'jello' in stills theoretically)

Shutter on leaf lens stills cams (H1, Rollei) of course can take a whole photo at their top speeds, 800 and 1000 but those chips are famous for being about 1FPS !

Again I know my old digi back neede to creat black reference files which took time too

S

 

[Mitch Gross]

There is so much more to it than just the sensor.

 

[Erik Widding]

It seems to me the sensor in the 7D for example is capable of a heck of a lot more than the rest of the hardware in the camera will allow.

There is a lot of noise reduction going on in those ASICS - particularly at higher ISOs... it is more than just a sensor, it is a system.

Im still confused as to how a DSLR can capture 1/250 th with out have a read time of one 1/250th and there a potential framarate of maybe 120 assuming write=read

I pick the 1/250th number because that is the flash synch time where the shutter is fully open

I understand that shorter still exposures are created by sweeping a shutter slit across the lens (which even creates 'jello' in stills theoretically)

1/250th is just the fastest shutter speed that has the mechanical shutter completely open. The readout is finished with the mechanical shutter closed.

 

[David Rasberry]

My GL2 is a good example of slow readout for stills. It has no mechanical shutter being a video cam and the processors are optimized for that function. But it shoots a modest JPEG still mode to at shutter speeds up to 1/250sec. It takes a full 5 seconds to write that out to an SD card however.
The still camera's raw frame rate probably represents close to the maximum the sensor is capable of. If it was faster they could probably do better 1080p H.264 scaling too.

 

[J. Bernard Vallon]

the fast shutter speeds of DSLRs is completely due to the mechanical shutter, same as film. It has nothing to do with the read-reset speed of the chip itself. this is why lots of these cameras have bad rolling shutters in video-mode, because the mechanical shutter takes care of that when shooting stills - there primary function.

 

[morgan_moore]

indeed all true but how slow is the D3 considering it can shoot great stills at 13FPS (in crop mode which is red size)

it cant be slower than 1/13th

shortly technology developments should double that ?

Of course this all need a lot of other bits put on the back, and dealt with which is exactly what Red are doing

if nikon gave a raw recorder it would surely create scarlet prices anyway

it cant be that easy or it would have been done is probably the best answer !

S

 

[Andrew Thomas]

I appreciate the comments, some very interesting points brought up. Though I do think my question itself wasn't really addressed (and admittedly it was probably poorly written). Based on what Canon themselves have said, the issue is not with the sensor itself but with the engine that runs it (the digic 4). In the future they will have a new version of the digic that will have dedicated video...stuff, for lack of a better word.

So if that is the case, why wouldn't somebody build a camera around the sensor out of a 7D that uses better scaling methods or even just better compression (maybe to a 10 bit codec for example). Is there information about the sensor that would only be known by Canon to do something like this?

 

[Jannard]

Canon uses line-skipping to get frame rates. You can't get 24fps out of a 6fps sensor unless you do something "creative".

Jim

 

[Tom Lowe]

Canon uses line-skipping to get frame rates. You can't get 24fps out of a 6fps sensor unless you do something "creative".

Jim

Which is why their "HD" video sucks and often leads to disastrous problems with artifacts and aliasing. Canon took the "quick and easy path...." and we all know that leads to the dark side.

 

[Andrew Thomas]

Canon uses line-skipping to get frame rates. You can't get 24fps out of a 6fps sensor unless you do something "creative".

Jim

I'll try and find the article that had the interview with a Canon tech guy, but he said the reason they can't run the sensor full blast (if you will) is heat and computing power.

Are you saying that is likely inaccurate?

(And thank you Jim for responding, nice to hear something from a more definitive source )

 

[Jannard]

I'll try and find the article that had the interview with a Canon tech guy, but he said the reason they can't run the sensor full blast (if you will) is heat and computing power.

Are you saying that is likely inaccurate?

(And thank you Jim for responding, nice to hear something from a more definitive source )

There are a ton of things that have to be solved to get real high frame rate large sensor motion. Heat and processing are two of them. There are a bunch more. But it all starts with a true high speed sensor.

Jim

 

[Andrew Thomas]

Very interesting. So I take it this is a major reason why you decided it would be best to build your own sensor designs?

And maybe this is a question best suited for Gramme, I dunno, but is there any method of line skipping or binning that would not result in the massive reduction of resolution and all the aliasing/moire that we see on the 7D if still using the same sensor?

And one more line of questioning. You guys are shooting still and video at the same resolution, while everybody else is shooting at different resolutions. Am I correct in assuming that a big hit to the image quality of these hybrid DSLR cameras is that the OLPFs are designed for the full resolution sensor and due to the skipping/binning instead of capturing full sensor and scaling, they lose image quality as the OLPF is not tuned for the resolution they are shooting in video mode? Could you theoretically replace the OLPF with one matched for the video mode and improve image quality that way?

 

[Jannard]

Very interesting. So I take it this is a major reason why you decided it would be best to build your own sensor designs?

And maybe this is a question best suited for Gramme, I dunno, but is there any method of line skipping or binning that would not result in the massive reduction of resolution and all the aliasing/moire that we see on the 7D if still using the same sensor?

And one more line of questioning. You guys are shooting still and video at the same resolution, while everybody else is shooting at different resolutions. Am I correct in assuming that a big hit to the image quality of these hybrid DSLR cameras is that the OLPFs are designed for the full resolution sensor and due to the skipping/binning instead of capturing full sensor and scaling, they lose image quality as the OLPF is not tuned for the resolution they are shooting in video mode? Could you theoretically replace the OLPF with one matched for the video mode and improve image quality that way?

Why not just buy a camera that does all you want? Yes... our stills are the same resolution (and RAW) as motion.

Jim

 

[Tim Whitcomb]

There are a ton of things that have to be solved to get real high frame rate large sensor motion. Heat and processing are two of them. There are a bunch more. But it all starts with a true high speed sensor.

Jim

hmm, that makes me think Red Gunner is definitely a COOLING system... that is my revised guess pre -10/30 announcement.

Those translucent tube renders and connectors sure look like liquid lines to me.

Red Gunner is an advanced cooling system which is used primarilty for high frame rate shoot and capture... modularity being the key, can also be removed for standard frame rate shooting.

Am I close?

 

[Andrew Thomas]

Why not just buy a camera that does all you want? Yes... our stills are the same resolution (and RAW) as motion.

Jim

It doesn't exist...yet:001_smile:

 

[Eddy Robinson]

from it? I see all these DIY camera projects using (seemingly) much worse performing sensors.

The main problem (aside from $ and time and the need for pretty advanced engineering skills) is that there's a limit to how fast you can read out the sensor before you have to start worrying about overheating, which creates noise on the image. This is one reason for the rise of CMOS vs CCD, the former uses much less power and so makes engineering easier.