[alehel]

RioRico brought my attention to talk of monochrome sensors. This really got me thinking. I think it would be awesome to have the Pentax K-7 body combined with a monochrome sensor. It would probably be economicly pointless for Pentax to produce but I would love to have one nonetheless . Would love to take it out for some street shooting combined with my 35mm and 70mm ltd lenses .

Anyone else like to se something like this?

[WerTicus]

Yeah i would think there might even be a decent market for this, and all they would need to do is remove the colour filter in front of the sensor and instantly gain 4x the resolution

perhaps they could make a camera that can internally move the colour filter out of the way to make it optional.

[alehel]

Actually sounds like something Sigma might end up doing :P

Think of the resolution you would get from the 645D

Maybe they could offer them as custom orders. They could just take the colour filter out of the K-7 on request. Would probably be quite expensive though. I've no idea how difficult that would be.

[RioRico]

As I've intimated, I would f**king LOVE to have a high-res B&W digicam. It needn't even be a dSLR. ("Burn that heretic!!") If someone's gonna do it, it would make sense to do a test run on something smaller, and would give immense bragging rights to whoever does it. So, a medium-res P&S suddenly becomes THE HIGHEST RESOLUTION COMPACT CAMERA EVER MADE!

So here's my fantasy scenario: Sony revives their DSC-V1 production line. It's a 5mpx Bayer-filter cam with NightShot (IR) and nice Zeiss optics, 4x zoom (7-28mm, f/2.8-4, equivalent to 34-136mm). Stock, its speed range is ISO 100-800; in NightShot it's effectively ISO 2500. The 1/1.8" sensor's density is only 13.8 mpx/cm2. A decent performer, admittedly noisy at high speed.

Now comes the CARTIER-BRESSON EDITION: DSC-V100!! Same platform, maybe use a (noisier and denser) 7mpx sensor, but with no Bayer filter. (And kill that sniperscope green in NightShot -- let's have REAL infrared!) Now it's THE FIRST 28 MEGAPICKLE COMPACT CAMERA, and its speed range is ISO 200-3200; and shove it into overdrive (uncrippled NightShot), it tops out at ISO 9600!! (Or something like that.) And with reduced noise! RAW output, of course. At its base ISO, the IQ would rival a FF dSLR. And all for just US$500. [drool]

No, that won't happen. But maybe Sony or Panasonic or Samsung could introduce something more modest, but still with higher-quieter resolution than anything now available. The Retro-Cam, 48 megapickles, colour not included. Would it hurt so much, to see how it sells?

perhaps they could make a camera that can internally move the colour filter out of the way to make it optional.

Sorry, that's not possible. The Bayer filter is layered right on top of each photoreceptor (pixel) in the sensor, and has a layer of microlenses right atop that. We're talking right-on-the-chip-level here. Nope, it's either B&W and sharp, or colour and fuzzy.

---

Last edited by RioRico; 03-14-2010 at 11:15 PM.

[Zav]

I want Ricoh to produce a monochrome module for the GXR! You want colors? Swap your modules!
Oh and it can also be a square sensor

[RioRico]

I want Ricoh to produce a monochrome module for the GXR! You want colors? Swap your modules!
Oh and it can also be a square sensor

First, Ricoh needs people to buy their camera. Maybe that'll happen.
Oh yes, square sensors. Just like using Verichrome Pan in a Yashicamat.

Meanwhile...
The discussion (and much info) on B&W sensors is here: Manual Focus Lenses :: View topic - Monochrome sensor; very interesting outlook

As for obtaining B&W sensors... It's a basic issue of sensor fabrication. A sensor consists of a big array of photodetectors (a megachip) topped with the filter array (often Bayer, sometimes others) topped with the microlens array. The photodetectors are transistors, either CCD or CMOS (and that doesn't matter here). These are made by VERY CAREFULLY depositing layers of semiconductor stuff (technical term) onto a silicon substrate aka wafer. The wafers are cut into chips. Wires are attached, etc.

Now we get tricky. Sometimes the fabrication plant (which is rather expensive to build and run) sprays a color filter layer onto the wafer, then glues on a separate sheet of microlenses. And sometimes another factory takes a sheet of filters, and glues them on, and then glues on the microlenses. Then the chips are packaged and shipped to the camera maker and the rest of the job continues.

The project cited in the link above doesn't have a fab plant. They're trying to take off-the-shelf sensor chips and peel away the glued-on filter layer in order to convert them to monochrome, which ain't easy. Doing so necessarily removes the microlens layer, which reduces the sensor's sensitivity a bit. Does this result in a lower base ISO level? I dunno. Panatomic-X fans, take heed.

The project shows samples of pictures with and without Bayer filters. The WITHOUT shots are strikingly sharp. But I wonder about getting image data from each and every pixel. With CMOS at least, each photodetector sits atop a signal amplifier. Its data is sent to a signal processor which adds it to a RAW file or whatever. Is each pixel addressed separately, or only as a cluster of four (1B+1R+2G)? To access each B&W pixel, must the underlying chip architecture be changed?

So, to get a commercial B&W sensor, we need a company that has its own fab plant(s), and that's willing to change their fabrication process (which ain't cheap), and maybe the chip architecture. Which is a cheaper change: spraying a clear layer instead of a color filter layer? Or gluing on a clear sheet instead of a filter sheet? I dunno. I'm a bit rusty on chip/sensor fabrication. It's not like you can just send the design off to a silicon foundry and get back a waferful of B&W sensors.

Who fabricates sensors? Canon, Kodak, Samsung, Nikon (I think), Sigma, and Sony all come to mind. Various others. Kodak made a high-res B&W sensor some years back, then stopped. Who is willing to pick up where they left off?

So maybe we'll this this happen, and maybe not. Start signing petitions.

[robgo2]

You guys are entitled to your fantasies, but wouldn't a monochrome sensor deprive one of the ability to adjust tonal separations by applying color filters digitally in post-processing? That, to me, is one of the most powerful aspects of using digital technology for B&W images. With a monochrome sensor, you would have to revert to the use of colored filters at the time of capture. How twentieth century.

Rob

[RioRico]

You guys are entitled to your fantasies, but wouldn't a monochrome sensor deprive one of the ability to adjust tonal separations by applying color filters digitally in post-processing? That, to me, is one of the most powerful aspects of using digital technology for B&W images. With a monochrome sensor, you would have to revert to the use of colored filters at the time of capture. How twentieth century.

For monstrous resolution, I'll use filters. And I'll LIKE it!!

[jbinpg]

Invoking the memory of Cartier-Bresson is misplaced in any discussion about high resolution imagery. CB would be appalled, actually. He loathed arguments about the superiority of one camera over another. Yes, the master produced his incomparable shots of Sartre, Capote, etc., with his Leica but he didn't wield it as a be-all,end-all sword.

[Gimbal]

For monstrous resolution, I'll use filters. And I'll LIKE it!!

Would the resolution actually be so much better? I doubt it. I mean the k10 (for example) pretty much renders details down to pixel-level. One might get a little better contrast and a little better details but wouldn't it be very marginal?

[WerTicus]

Yes you get 4x the resolution out of the sensor because currently it is taking 4 black and white pixels with a RGGB filter in front of them to make one colour pixel.

[er1kksen]

Real-world usage of those rare monochrome CCD cameras has shown some resolution increase but 4x the resolution it has not. The luminance resolution of a normal RGGB sensor with no AA filter isn't that much worse (the monochrome sensors lack AA filters, widening the gap over traditional filtered RGGB sensor).

Theoretically one could write a RAW-interpreting program that reads a normal RGGB sensor as a monochrome sensor, just using the luminance value of each photosite. It would still know which photosites were filtered with which color, so it would be able to add or subtract luminance inferred by the surround (differently colored) pixels. Note that this is different than interpolation, it's just using info from the surrounding pixels to "fill back in" the light that got filtered out, rather than combining data from different pixels to get a color value. This would also possibly retain the potential for choosing the sensor's color response rather than needing to use filters.

It's possible that that would be a lot more complex than your typical demosaicing algorithm, and it's also possible that it wouldn't work at all. But I think there's a possibility that it would work, and that it could yield excellent results.

[ilya80]

It's possible that that would be a lot more complex than your typical demosaicing algorithm, and it's also possible that it wouldn't work at all. But I think there's a possibility that it would work, and that it could yield excellent results.

Excellent idea, however I believe that you can only get 8 bits per color out of your RAW file now. If only they actually saved 12 bits per pixel as per ADC does it....

[Gimbal]

Yes you get 4x the resolution out of the sensor because currently it is taking 4 black and white pixels with a RGGB filter in front of them to make one colour pixel.

Both yes and no, if you take a normal 10Mpixel camera and remove the color-filter you would end up with a 10Mpixel black and white sensor. It would be a little bit sharper, but it wouldn't be a major difference.