Would you buy a monochrome DSLR?

[johnb]

<i>"Would you buy a monochrome DSLR?"</i><br>

<b>NO!</b>

Why?<br>

Because I see in color!

[david_kieltyka]

Scott brings up an interesting point about tonality. With a monochrome digital sensor you'd be limited to its particular tonal response. You could alter this to an extent with filters, but filters can't turn Plus-X into Delta 100. Each b&w camera could only use one kind of "film".

 

I wonder how much improvement we'd get in b&w output if we had de-mosaicing algorithms tuned specifically for it? For example, use only the green photosites to create output pixels...but bias the values of those pixels based on the values of the red & blue photosites. This would halve the number of pixels but might not halve the sensor's spatial resolution. It would also allow for in-camera or RAW conversion tonal adjustment. I'd love to see someone like Dave Coffin give it a shot.

 

-Dave-

[jbq]

David: personally, I can't stand the response curve from my 10D - it's never "right", typically never contrasty enough. Printing directly from the memory card always yields very disapointing results. What I like however in the digital workflow is that I can adapt the tonal response curve when I see the unprocessed shot, in the comfort of my home, and I don't need to worry about it while shooting. Once I have made the proper adjustments, I can get prints that look much better from a digital source than they would have looked on film.

[troyammons]

>I'd still like to see a true, 10-16mp version of the Sigma/Foveon >sensor though

 

The next one should be up there but not that far.

 

As i understand it the original Foveon prototype was a 16mp x 3 full frame chip. Thats the one I am waiting for. Doubt it will ever happen though, but if it did you would need a lot better glass in front of it.

[vivek iyer]

Since this is a speculative/hypothetical question, here is my take on it;

 

Mamiya ZD-

 

It has a slot for filters. Add an appropriate one to make the camera see in B/W.

 

I hope, the Mamiya ZD will do well and shake the CaNikon hold on the DSLR market a bit.

[godfrey]

No: Good digital B&W for pictorial use depends upon capturing a full spectrum of light and

then processing it to provide tonal separation and contrast in the translation to monochrome.

 

A monochrome DSLR might prove to be a little faster and a little higher rez, but would limit

you to one spectral response, unless you took three exposures through R, G and B filters for

every photograph and then integrated the results. (That's what the Mars Rovers do.)

 

Godfrey

[digitmstr]

>> typically never contrasty enough <<

 

With film you have choice of:

 

1) use a high-contrast emulsion and/or soup/agitation, etc... to increase contrast

 

2) do it during printing

 

Digital is similar. If you have a high-contrast image from the sensor you'd be limiting the "printing" choices. Instead, you get a lower contrast, full range "negative" which, with proper "darkroom" techique can deliver excellent B&W prints.

[jan_brittenson]

<i>The dynamic range would be identical to that with the color filter array since the filter array does not affect either of those parameters.</i><p>

This is true for white light. For anything else, the color filtering increases dynamic range as measured in SBR at the cost of sensitivity. If you think about it, an SBR of ten stops in red will saturate the red sites but not green and certainly not blue. As a result, red brightness levels beyond the saturation point of red can be differentiated through the relative levels of the two other channels. The more spectrally narrow the light, the higher the dynamic range.

[grepmat]

Jan, I think you are excessively complicating the simple concept of dynamic range by invoking scene-dependent, filter-efficiency and demosaicing-algorithm issues. What you are arguing is that spatially distinct pixels should somehow augment the dynamic range of each other, which would seem to be complicated to make work well in general situations. For example, consider the original poster's scenario of astronomy: Let's say that a star's light hits a single pixel - there is no way that neighboring pixels can help the dynamic range of that pixel since they see none of the star's light.

 

However, there is one case where what you are suggesting certainly could work - and that is essentially similar to the "super ccd" used in a few cameras. These contain sub-pixels with different responses. So imagine a pixel that is actually 2 pixels in one, but one of them has a neutral-density filter in front of it (or equivalent), cutting light by say 3 stops. Once the normal pixel is saturated, the second pixel would still have 3 stops of useful range remaining. Assuming that the noise floor is set by the first pixel and is uncompromised, the combination of the two would have 3 stops more dynamic range than an ordinary pixel. You can extend this further by adding more sub-pixels, each with even further-reduced response. This scheme obviously has some drawbacks, such as reduced spatial resolution since more than one physical pixel is used per output pixel. Also, it does little or nothing to improve the noise floor (which is typically the bigger problem), but only helps with brighter and brigher scenes.

 

Cheers.

[daniel_kreithen1]

OK, but here's something I don't understand about dynamic range in DSLRs (which I think most people will agree sort of stinks compared to B&W film). If you read the imageing chip spec sheets, the dynamic range specifications for these chips are in the ballpark of 60 - 70 dB (which translates to > 10 stops - which DSLRs do not attain). Is there something about the Bayer filter, demosaic-ing algorithms, subsequent signal processing that fails to preserve dynamic range, or is it simply a design choice by the various manufacturers of how much noise to truncate at a given ISO? I've always suspected that the "low-noise" sensors produced by certain manufacturers are not that at all, but rather a product of different design choices in signal processing, at the expense of dynamic range. Certainly it is the case that if you're seeing no or low noise in the shadows, that must indicate that some sort of truncation is being performed.

[jim_roof]

I have shot images with my 1Ds MkII that have captured at least 9 stops in the CR2 raw file. It takes a little work in PS to incorporate all of that range however.

 

I think the 10 stop range in a good DSLR sensor is correct. The jpg's will never show it because they are interpreted from one pass on the internal raw file (and then the raw file is lost unless you save it). The ten stops are there. For a fact.

 

As for a monochrome DSLR, I suspect the numbers simply are not there to support one at a decent price point. For every 1000 DSLR shoppers there might be .05 sho want a monochrome camera. Consider the possibility that, for every 1000 DSLR shoppers, there may only be 100 who know what monochrome means...

[peter_nelson1]

<I>I've actually written Canon about this, but alas I don't know if anyone who cares read it.<P>

 

This is one of several unfulfilled niches still left in digital photography. The company that fills it will be rewarded</I><P>

 

Sure, <B>they'll be rewarded with a trip to the unemployment line!</B>

 

Anyone who knows anything about technology manufacturing knows that the factory cost is only a small factor in producing a product. Marketing, packaging, staff-training, parts inventory systems, bug-tracking systems, shelf space, and a zillion other things all have to addressed whenever a company brings a new product to market. Adding a new product is a big deal not to be taken lightly.<P>

 

A monochrome DSLR is a niche product for a small geeky niche market that is best served by small, geeky niche aftermarket companies like the ones that make products for black-and-white printing. Anyway, aren't there already companies that make digital astrophotography cameras?

[nikos peri]

<i>Anyway, aren't there already companies that make digital astrophotography cameras?</i><p>

Yes there are. See a few hundred pages higher in this very thread. ;-)

[grepmat]

In engineering terms, dynamic range is most often quoted as the ratio of the maximum signal that can be recorded by the sensor to the root-mean-squared noise floor of the sensor (signal to noise ratio or SNR). A "stop" is just a factor of two in SNR. Thus, a sensor with a 10 bit dynamic range (1024:1 SNR) could be said to have 10 stops of dynamic range.

 

The problem is that a stop is a photographic term, and photographers tend not to like that last stop of performance, where the signal is only twice the RMS noise and images at that level are so unattractive as to be nearly useless. So they want "real" stops, where the last stop is several times the RMS noise. The problem is, that's just an arbitrary judgement call.

 

I have often had clients come to me and say "I want 10 bits of dynamic range" for the sensor they want designed. I say "o.k., 10 bits of SNR is pretty easy..." But then they interrupt me and say "Oh, but they have to be real bits!" I ask what they mean and they say something like "well, the last bit has to be maybe 4 bits above the noise. Then I have to say "O.k., so you really want something like 13 or 14 bits?" They are acting like photographers who want 10 "real" stops instead of just saying they want, say, 13 or 14 stops as would be quoted by an engineer.

 

Cheers.

[jbq]

"Is there something about the Bayer filter, demosaic-ing algorithms, subsequent signal processing that fails to preserve dynamic range?"

 

I'll assume for a moment that dynamic range is limited by the perceived noise in the shadows for a properly exposed frame (that's my personal experience with a 10D), and that at low sensitivities dynamic range isn't a problem (that's my personal experience with a 10D). I'll also assume for an instant that as a general rule exposures where fewer electrons are stored per pixel are the ones that have a lower dynamic range (I have a better dynamic range with my 10D at lower sensitivities than at higher ones). If your experience with dynamic range is different, skip the rest of this post.

 

I claim that for a given sensitivity (i.e. for a spatial density of photons hitting the front of sensor) and for a given technology, a limiting factor in terms of the number of electrons that are generated for a given number of photons is the bayer filter, whose job is to eliminate photons of the wrong color. Removing the bayer filter would allow more photons to generate electrons, i.e. to increase the dynamic range.

 

Another way of saying the same thing is that the dynamic range gets lower and lower as you increase the amplifier ratio in the sensor, and that removing the bayer filter would allow to use a lower amplifier ratio for the same sensitivity.

 

On top of that, if the brightest part of the scene is not white, you will be able to give more exposure for a given sensitivity if you don't have a bayer filter: with a bayer filter you'll have to make sure that both red, green and blue components of the light don't saturate, whereas without one you only need to worry about their average.

 

 

I also claim that if you want to shoot through a deep red or deep blue contrast filter (either a hardware filter or a digital one), the bayer matrix will seriously eat into the resolution. Check at DPreview the comparison of the SD10 and 10D, in the red and blue channels, for the gray patch. There's little doubt that the SD10 does a much better job.

 

 

Well, obviously, all those of somewhat hypothetical, the only way to really know would be to have such a camera in hand. But, once again, if Canon thinks that there's a big enough market for a 20Da (and it's not like they are alone on the market of digital astronomy cameras), maybe there could be a bug enough market for a 20Dm. Oh well, time will tell, and in the meantime I'll keep my fingers crossed.

[daniel_kreithen1]

Yes, the idea that 10 stops of sensor specification does not mean that those 10 stops are actually useful squares with my experience with 4/3 or APS-C sized sensors. My estimate is that I get about 6-7 stops of usable dynamic range, perhaps the full 7 with RAW format. As mentioned above, the full 35mm frame sized sensor in the 1DII gives better dynamic range, at least in the RAW file. This is also consistent, since I imagine the "wells" in that sensor are somewhat larger than that of the smaller sensors. If it is really true that removing the Bayer filter and dispensing with the interpolation algorithms that necessarily go with it will give better sensitivity (and therefore some increase in dynamic range) then perhaps it is something that some manufacturer will pursue in order to give it an edge in the niche B&W market. But it sounds like the only real way of buying-back (B&W) film-like dynamic range on a single exposure basis is to make the sensor and associated "wells" physically bigger. Pity. Anyway, a good reason to still use film in certain applications.

[grepmat]

Jean-Baptiste: You are discussing the dynamic range of an under-exposed *picture* as opposed to the dynamic range of a *sensor*. The dynamic range of a sensor would not change if the filter were removed. But, by definition, an under-exposed picture would not utilize the entire dynamic range of the sensor that took it. You are correct that an under-exposed picture would have a higher dynamic range without a filter since less light is thrown away. This is precisely what I meant when I said that a higher sensitivity sensor can have a similar advantage to a higher dynamic range sensor under certain conditions. However, there would be no advantage with or without a filter in the dynamic range of a picture if it is properly exposed to utilize the full dynamic range of the sensor. So, yes, in the end, "available darkness" photographers forced to underexpose would appreciate a filterless sensor, just exactly as they would *any* sensor with higher sensitivity. Cheers.

[Jochen_S]

Yes, I'd buy; but make it a Pentax please, or Konica/Minolta. Maybe even a affordable Epson DRF. I have no Canikon and wouldn't like to get into these systems.

[rayfraser]

With a small market, they would have to charge more - probably several $100 more. Now since any DSLR is out of my range (doesn't meet my dynamic range requirements), I'll never actually try scraping away a DSLR bayer. But I've gained enough confidence that I'd follow my 1 hour process before unloading whatever first DSLR cost plus $200 to get a second bayerless one. Hey if you fail you can just give up and buy another color saving $200. <p>

 

<a href="http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00CM0R">Bayer Begone</a> <p>

 

If anyone has serious suggestions of follow-up comparisons of my debayer vs bayer imagers, please post your thoughts.