From: David Dyer-Bennet on
Roger N. Clark (change username to rnclark) wrote:

> There are companies that remove the IR filter (Heutech (sp?) for one),
> but removing the Bayer filters would require removal from the
> surface of the sensor, and I don't know anyone who does that.

Two of the places that do IR removal have confirmed the general
impression that removing the Bayer filter isn't practical on current
cameras. I'm not terribly surprised, but I figured getting more
information than just my own impressions was worth the trouble of asking.

> I would buy a 20D class black and white camera if such were
> to come on the market.

I'd be extremely tempted, at least if it really had a stop or two more
sensitivity (or more; more is always okay). Although for me it'd have
to be a Nikon mount, or else I'd have to find enough money to replace
bunches of lenses (some of which can't really be replaced in the Canon
world). I was figuring that a D40 plus $300 conversion plus $150
KatzEye screen (including installation) was something I'd have to find
money for if the conversion were possible.
From: David J. Littleboy on

"Roger N. Clark (change username to rnclark)" <username(a)qwest.net> wrote:
>
> Removal of the Bayer filter would give a lot more light throughput
> than 1.5 stops. You have not only the transmission of the
> filters but their bandwidths too as factors. If you removed the IR filter
> too, I think I computed once that you would gain about 50x in speed!

Hmm. At first glance, that 50x seems way over the top, but if the filters
actually attenuate in the pass band as well, a factor of 4x (assuming that
the pass band is 1/4 of the band from IR to near UV) plus another 2 for the
in-band attenuation, and that's a factor of 8x. So three stops is possible.
It sounds like you are thinking of shooting under dim incandescent lights
where there's a lot of IR and not much blue and green.

I had originally thought that you'd only get a stop or so going to B&W, but
if you think about the far higher QE of digital than film, it makes sense
that seriously high sensitivities should be possible.

One problem may be that lenses that can focus both IR and visible at the
same time are few and far between...

> Removal of the blur filter would also make very sharp B&W images.

I'd leave the low-pass filter on. At least on the 5D, it's way too weak
already, and probably just about right for 12.5MP B&W.

> There are companies that remove the IR filter (Heutech (sp?) for one),
> but removing the Bayer filters would require removal from the
> surface of the sensor, and I don't know anyone who does that.

At least on the P&S sensors, the Bayer CFA filters are below the
microlenses, and both are fabricated on the chip.

> I would buy a 20D class black and white camera if such were
> to come on the market.

I'd rather a (16+MP) 5DII with interchangeable sensors and special-order
options with and without the IR cut filters and with and without the
low-pass filter so everyone can get the combination they want. I bet Canon
would make a fortune on the color + no LPF combination. (The availability of
this latter combination would bankrupt Foveon<g>.)

David J. Littleboy
davidjl(a)greedy.com
Tokyo, Japan


From: ASAAR on
On Mon, 16 Apr 2007 10:48:09 +0900, David J. Littleboy wrote:

>> Removal of the Bayer filter would give a lot more light throughput
>> than 1.5 stops. You have not only the transmission of the
>> filters but their bandwidths too as factors. If you removed the IR filter
>> too, I think I computed once that you would gain about 50x in speed!
>
> Hmm. At first glance, that 50x seems way over the top, but if the filters
> actually attenuate in the pass band as well, a factor of 4x (assuming that
> the pass band is 1/4 of the band from IR to near UV) plus another 2 for the
> in-band attenuation, and that's a factor of 8x. So three stops is possible.
> It sounds like you are thinking of shooting under dim incandescent lights
> where there's a lot of IR and not much blue and green.

At first glance I agree with your first glance opinion. :)
Charles Schuler posted an informative link in this thread a couple
of days ago that may clarify the amount of attenuation that takes
place if the Bayer Filter Transmission vs. Quantum Efficiency chart
(Figure 4) is a fair representation. It shows fairly broad filters
with a good deal of overlap, and quantum efficiencies at different
frequencies ranging from about 8% to 38%, maybe averaging a little
bit more than 20%.

> Interesting idea. Would one necessarily lose the micro-lenses as well and
> would that mitigate against removal? I don't really know, by the way. Here
> is an interesting link that might be helpful:
> http://micro.magnet.fsu.edu/primer/digitalimaging/cmosimagesensors.html

From: Martin Brown on
On Apr 13, 6:11 pm, David Dyer-Bennet <d...(a)dd-b.net> wrote:
> Is any company offering removal of the Bayer filter from a Nikon-mount
> DSLR? Particularly the D40? (I suspect it may not be feasible due to
> the stacking order and how the microlenses, AA, IR cut, and Bayer are
> combined, and other issues, and I haven't been able to Google up much,
> but I thought asking might still turn something up.) I'm interested in
> a high-sensitivity B&W camera for low-light situations.

Would you not be better off with a faster IS lens?

It might be possible to bleach the filters by exposing the chip to
strong UV light with some risk of sensor damage increasing the noise.
Or for certain camera combinations there are pin compatible monochrome
only sensors. Consumer grade astronomical cameras tend to use them for
semi serious science work (as opposed to single shot colour). However,
even though they are excellent in low light they lag behind state of
the art MP rankings by using larger area pixels.

I doubt you will find a consumer SLR that permits swapping the sensort
chip, but I could be wrong.

Regards,
Martin Brown

From: Ilya Zakharevich on
[A complimentary Cc of this posting was sent to
ASAAR
<caught(a)22.com>], who wrote in article <pir523dn2lkd71hn0i9u5n00edp94lr70u(a)4ax.com>:
> Charles Schuler posted an informative link in this thread a couple
> of days ago that may clarify the amount of attenuation that takes
> place if the Bayer Filter Transmission vs. Quantum Efficiency chart
> (Figure 4) is a fair representation. It shows fairly broad filters
> with a good deal of overlap, and quantum efficiencies at different
> frequencies ranging from about 8% to 38%, maybe averaging a little
> bit more than 20%.

If you mean

http://micro.magnet.fsu.edu/primer/digitalimaging/cmosimagesensors.html

I do not see any QE spectral data on Figure 4. So what do you mean?

To find the effect of removing of Bayer, one needs to average these
curves with the coefficient given by the QE curve (well, it must take
into account the IR filter too). In principle, one could try to guess
the QE curve by dividing the "throughput" measured QE of
IRF+Bayer+Sensor by the data for the Bayer.

[Do not have the link for the measured QE at hand, but it was you who
gave it to me. ;-]

Hope this helps,
Ilya

P.S. Let me reiterate that my estimates of "throughput" QE of a 2005's
Canon's sensor (with sunlit scenes) were about 14%.