Go Back   IceInSpace > Equipment > ATM and DIY Projects

Reply
 
Thread Tools Rating: Thread Rating: 157 votes, 4.99 average.
  #1  
Old 07-07-2013, 08:41 PM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Mono DSLR Mod - removing the Bayer CFA

Hi

I've been thinking of this for a while after reading similar attempts at the SGL, so I bought a 350D off ebay and a USB microscope and successfully removed the CFA.
The Camera still works and opening the CR2 file in Maxim displays a monochrome image as expected.

the whole rationale behind it was to get a decent sized mono sensor without the thousands. I know it still won't compare to a 16 bit astro ccd camera, but it should fit in between and should make narrowband more affordable.

this was more of an experiment so I can then move on to a newer model with a larger sensor and convert it to mono, replace the IR filter and cool it with a cold finger.

any suggestions on which medium level DSLR sensor has reasonably high QE?
I have a 10inch F4, so smaller pixels would suit.
the 1100D and 20D have their glass screens glued on and are almost impossible to remove to get to the sensor so they're out of the question.

I just followed the 350d filter mod guide to open up the camera and get to the sensor.
after that, I removed the IR filter.
the sensor is behind another piece of glass. It was next to impossible to lift this without breaking it, so I stuck some tape and took it apart in pieces.
that exposed the sensor surface. no damage done to the sensor itself.

For a tool, I used the top end of a wooden paint brush, sharpened. this was to prevent permanent scratches to the photosite layer although its actually quite hard.
gentle scraping removes the microlens layer which is almost like a sticky coating and easily comes off.
the CFA is a bit harder but once you get an opening, the rest just comes off. its quite brittle.
I used the microscope with 500x to make sure I didn't touch the wires at the two sides as that could render it dead.

I managed to scrape off the cfa in about half an hour. so assembled and ready to test, I find out that the battery is flat and the battery charger was'nt included.

I was getting a bit impatient, so I snuck two wires into my 1000D's battery and connected it in parallel to the 350D battery and it was charged half in about 15 minutes.
enough to test.

here are a few test shots. i used a 18-55mm lens with manual focus so the focus may be a bit off but I was pretty happy with the results.

the 350D doesn't work too well with backyardEos, so I'll have to get the timer for longer exposures.

close ups show that there are still bits of the CFA left. I need to do go over it with the microscope a give it a proper clean.

I'm aware that there will be a drop in sensitivity with the microlenses gone, but it should more than make up for it with the 4x increase as now all 4 pixels are used for luminance.

With the IR filter removed, it should also suit narrowband.

Here are some videos of the "scraping".

http://www.youtube.com/watch?feature...&v=aJ65Ve4bdOg

higher mag, you can see the bayer matrix here.
http://www.youtube.com/watch?feature...&v=E9kxsYutm5U

now to do some real tests without the IR filter and the 10inch F4.
Will post results once I get some clear skies, and also when I move on to the next camera.
I might get a 1000D off ebay for its 6micron odd pixels and size and give that a go as well.
comments welcome.



Cheers
Alistair
Attached Thumbnails
Click for full-size image (2013-07-06 20.15.47.jpg)
134.3 KB1145 views
Click for full-size image (2013-07-06 20.54.01.jpg)
127.5 KB993 views
Click for full-size image (IMG_1012 (Large).jpg)
169.6 KB939 views

Last edited by alistairsam; 07-07-2013 at 10:40 PM.
Reply With Quote
  #2  
Old 08-07-2013, 06:31 AM
Merlin66's Avatar
Merlin66 (Ken)
Registered User

Merlin66 is offline
 
Join Date: Oct 2005
Location: Junortoun Vic
Posts: 8,904
Interesting!
The guys in the UK (and JTW in Netherlands) seem to have had success...
Definately not a job for the faint hearted!
Look forward to seeing your results.
Reply With Quote
  #3  
Old 08-07-2013, 07:19 AM
bojan's Avatar
bojan
amateur

bojan is offline
 
Join Date: Jul 2006
Location: Mt Waverley, VIC
Posts: 6,932
That's awesome..
I always thought the Bayer matrix is part of the sensor, therefore not possible to remove it without damaging the chip itself.. Because each sensor (photodiode) must have it's own individual filter, which is deposited directly on the the silicon at appropriate place(s).

Actually, all my CCD chips I accumulated over time are like that.

I shall certainly consider your approach... and I am very tempted to pay you a visit when you have some free time (we are living close) to directly see and discuss the process.

Last edited by bojan; 08-07-2013 at 07:33 AM.
Reply With Quote
  #4  
Old 08-07-2013, 08:13 AM
Merlin66's Avatar
Merlin66 (Ken)
Registered User

Merlin66 is offline
 
Join Date: Oct 2005
Location: Junortoun Vic
Posts: 8,904
Bojan,
Think of the CFA as a "coloured screen print" onto the surface of a mono chip.
Obviously both the colour correction and IR filters will be removed from the camera body - ending up with a FULL full spectral modded mono CCD.
Reply With Quote
  #5  
Old 08-07-2013, 08:43 AM
bojan's Avatar
bojan
amateur

bojan is offline
 
Join Date: Jul 2006
Location: Mt Waverley, VIC
Posts: 6,932
Quote:
Originally Posted by Merlin66 View Post
Bojan,
Think of the CFA as a "coloured screen print" onto the surface of a mono chip.
Obviously both the colour correction and IR filters will be removed from the camera body - ending up with a FULL full spectral modded mono CCD.
Yes, I know bayer matrix filter is "printed" on the chip..
In the past I was engraving various markings on the transistor chips using microscope, sharp needle and free hand.. I was always intrigued by how actually soft the metal deposits are and how easy it is to damage them.. and also how easy it is to control the hand movements if they are controlled by visual feedback (by watching what's going on under the microscope).

That is why I appreciate Alistair's effort and success so much.
Very brave! And successful.

Now, maybe there is a way to remove the matrix by some chemical process? Not likely..
Reply With Quote
  #6  
Old 08-07-2013, 10:39 AM
rcheshire's Avatar
rcheshire (Rowland)
Registered User

rcheshire is offline
 
Join Date: Apr 2010
Location: Geelong
Posts: 2,617
Good one Alistair. I look forward to seeing the results.
Reply With Quote
  #7  
Old 08-07-2013, 10:58 AM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Thanks Ken, Bojan, Rowland,

the guys in SGL have gone through several chemicals and some of the partially successful methods were
using a buffing tool with a dremel - worked for a 1000D
some sort of a paint remover
toothpaste!!

but after thinking about it and the fact that the photosites are composed of multiple layers sandwiched and are quite hard, i thought the wooden tool was the safest because it doesn't cause any scratches on the layer beneath the CFA and also because the CFA layer is brittle.

one of the members there, Gina, did every imaginable thing possible to remove the glass off a 1100D sensor and finally gave up. these sensors are actually a lot tougher than people think.

and as Ken mentioned, removing the IR filter and removing the CFA gives you a full spectrum camera, only limited by the slight loss in sensitivity due to the microlenses being removed which is unavoidable, and the fact that the QE of these sensors are lower than astro CCD's.

but still, making use of all 4 pixels opens opportunities for so many applications, narrowband, Ha, luminance, all at a fraction of the cost of a similar sized CCD.

I can now do narrowband imaging from my light polluted backyard. once I cool it of course which is another challenge on its own.

I figured, if I'm going to spend 2.5K toward a CCD, I might as well spend a few hundred trying it with DSLR's and worst case, each 350D sensor is $60 on ebay. so its very cost effective and is not really that hard if you're good with small things.

it certainly requires a steady hand and I'm used to soldering SMD's by hand, so I had no issues controlling or restricting my movement.

Bojan, it takes a bit of practise to control the hand movement based on the microscope visual feed, but its quite easy.
as I mentioned, the two edges with the contact wires are the most risky cause one contact on those wires and its very likely the sensor would be dead.

I'm thinking of moving on to a 400D or a 1000D which I know is noisier but is quite cheap and the glass does come off.

Bojan, you're welcome to come over. but be prepared to spend atleast a few hours. I have heaps of questions for you as well. pls drop me a PM.

Cheers
Alistair
Reply With Quote
  #8  
Old 08-07-2013, 12:08 PM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
I've come up with this SS to determine which one is going to be the next scapegoat so I can invest some time and money and remove the cfa, replace the IR filter, and cool it.
I just need one which I can control via BEOS and has 14 bits and between 5 and 7 microns for imaging galaxies.

this SS might be useful to someone.
The text file is tab delimited so you can import it in to excel.
Wish I could get the QE and dynamic range values as well.

I also found this site very informative in comparing various sensors.
Interesting to see that the read noise of most dslr's are lower than the kaf-8300 if i'm interpreting this correctly.

I've been very tempted to do the same with a QHY8 which uses a nikon D40 sensor.
only issue is that the sensor is pretty much glued on to the PCB so if I do damage it, i won't be able to replace it.
if it were replaceable, it would be really cool.

http://www.clarkvision.com/articles/...mance.summary/


Cheers
Alistair
Attached Files
File Type: txt dslr-canon-sensors.txt (6.6 KB, 142 views)
File Type: pdf dslr-canon-sensors.pdf (291.1 KB, 306 views)

Last edited by alistairsam; 08-07-2013 at 12:19 PM.
Reply With Quote
  #9  
Old 09-07-2013, 08:48 AM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
hi

if I replaced the factory IR filter, will I need a UV IR filter to prevent bloated stars or will the bandpass characteristic of a Lum filter or a LP filter suffice to cut UV/IR?

any suggestions on where I can get a piece of glass cut to size to replace the IR filter?

As this is just a test, I cant really get the baader acf filter

thanks
Alistair
Reply With Quote
  #10  
Old 09-07-2013, 09:11 AM
Merlin66's Avatar
Merlin66 (Ken)
Registered User

Merlin66 is offline
 
Join Date: Oct 2005
Location: Junortoun Vic
Posts: 8,904
I have and use a FULL full spectrum modded 1000D (both filters removed - not replaced)
To maintain focus with std lenses I've added a clip-in Astronomik CL filter.
Different filters have different UV cut-off.
For astro work, you can add the UV-IR cut anywhere in the optical path.
Reply With Quote
  #11  
Old 09-07-2013, 10:02 AM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Thanks Ken,

I can't get away with not having a filter as I've removed the glass covering the sensor.
if I leave it open, moisture could cause a short, so I'll need a layer of glass.
I'll see if I can get a piece of clear glass cut and use a UV/IR filter at the end of the nosepiece.

Can I use an LP filter as the L filter?

Transmission curve for the LPS-P2 filter
http://www.sciencecenter.net/hutech/idas/filtdata.htm

It looks like it covers the lower and upper ends of a normal L filter, but the added bonus is that it cuts off or reduces the Light pollution wavelengths as well.

Curve for astronomik LRGB filters
http://www.astro.uni-bonn.de/~mischa...nomik_lrgb.png


Will there be a discernible difference between a 12bit and 14bit dslr for DSO's?

I'm hoping to try it tonight if I can get the glass cut.


Cheers
Alistair

Last edited by alistairsam; 09-07-2013 at 10:13 AM.
Reply With Quote
  #12  
Old 09-07-2013, 10:34 AM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
I was thinking of doing something like this at some point. I did read some posts where one guy was offering it as a service and I think it was Nikon D90's he was doing.

As you point out some may be harder than others.

I have modified a few 20Ds. The UV/IR cut filter is glued in at its edged with black silicone. I did not take it any further but I take it you are referring to the glass cover slip over the sensor?

Good work.

As far as which sensor I would think a Sony Nex 6 or a Nikon D7100 or Fuji XE1.

The current lowest noise APS camera is a Fuji XE1 (about $715 on ebay body only, Nex 6 is similar). Sony Nex 6 is not too bad either. Nikon D7100 is 24mp, no AA filter and is also low noise. Sony Nex 5r is cheaper and same sensor. I think the Nex 3N may also be the same 16.3mp Sony Exmor sensor (not 100% sure) so that may be the cheapest.

Canon EOS APSc not sure Canon advanced past the 350D/20D/40D/1100D models. The others seem a noisy. 6D 1Dx and 5D3 are clean at high ISO. 5D2 is in between. A 5D2 may be fairly cheap now or a 5D with its largish pixels.

I would not worry about DSLR pixel size - they are all smallish compared to CCDs.

Greg.
Reply With Quote
  #13  
Old 09-07-2013, 10:45 AM
Merlin66's Avatar
Merlin66 (Ken)
Registered User

Merlin66 is offline
 
Join Date: Oct 2005
Location: Junortoun Vic
Posts: 8,904
As Greg has already asked...is the glass you removed the internal (bluish) filter? or another layer of glass.
If it's just the original filter then I wouldn't worry about moisture etc.
This filter element is the one usually removed in the DSLR -increased Ha response mod.
I've been living without any filters for the past four years with no issue.
Re filters
The Hutec LPS would work as the Uv-IR cut, as well as light pollution rejection. You colour balance will obviously shift due to the filter.
Can't really comment re DSO - I only do spectroscopy.
Reply With Quote
  #14  
Old 09-07-2013, 02:53 PM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Hi,
I was referring to the glass directly over the sensor as in the pictures attached.
the first shows the glass lifted at the top right corner, it turns grey as you lift it, if you lift too much, then it just cracks.
the second pic shows the glass removed. you need to remove this to get to the sensor surface.
so basically the wires and the sensor are exposed.
I was thinking of just sticking a clear glass over the sensor and removing the IR filter.
in the 1000D's and similar, there are actually two glass layers, the top most one is the one that's normally removed or replaced for the mod.
I believe the bottom one is the AA glass and has the piezo for dust cleaning.

I'll most probably mod a 1000D or a 400D next. the 450D seems attractive as its 14bit. but a bit more pricey for second hand.

As for the filter, I was hoping to use the IDAS-LPS-P2 filter as the "L" filter for LRGB as this from what I can see, covers the normal "L" wavelengths and cuts off LP wavelengths.

Greg, in terms of QE, I know the dslr's are only between 25 to 30%. would you know of any mid-range dslr's that have relatively high QE?
I don't think I'll try the Fuji or the Sony just yet.

yes, Guy, author of BEOS has also done mono conversions for Nikons. I prefer Canon's just for the software support although if I can debayer a D40, I might take on a QHY8!!
two second hand QHY8's (in case I break one), are still cheaper than a comparable mono.

Cheers
Alistair
Attached Thumbnails
Click for full-size image (1000d.jpg)
140.1 KB380 views
Click for full-size image (1000d-without.jpg)
142.9 KB374 views
Reply With Quote
  #15  
Old 09-07-2013, 03:07 PM
Merlin66's Avatar
Merlin66 (Ken)
Registered User

Merlin66 is offline
 
Join Date: Oct 2005
Location: Junortoun Vic
Posts: 8,904
Alistair,
On the 350D there's only one filter element, about 2.8mm thick, which I assume just sits in front of the "protection" glass you show.
The filters on the 1000D etc are much thinner, each is 1.0mm thick.
Edmund Optics sell thin plates of coated glass.....
Reply With Quote
  #16  
Old 09-07-2013, 03:49 PM
pmrid's Avatar
pmrid (Peter)
Ageing badly.

pmrid is offline
 
Join Date: May 2008
Location: Cloudy, light-polluted Bribie Is.
Posts: 3,665
At the risk of asking a silly question, can you achieve a similar result by setting the camera to shoot in B&W mode?
Peter
Reply With Quote
  #17  
Old 09-07-2013, 03:52 PM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Hi Ken,

That's right. the 350D has the glass on the sensor or just above it and the IR glass on the plastic frame above the sensor glass which is normally what is removed for a mod.
for the 1000d, it'll be the topmost glass that's replaced for an IR mod, but I'll be removing all three for the B&W mod.

Alistair
Reply With Quote
  #18  
Old 09-07-2013, 04:00 PM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Quote:
Originally Posted by pmrid View Post
At the risk of asking a silly question, can you achieve a similar result by setting the camera to shoot in B&W mode?
Peter
Hi Peter,

there are lengthy explanations for this, but the simplistic answer from my understanding is that it won't provide the same sensitivity.

below is just my understanding, I'm no expert though.
B&W or mono images from a colour camera use interpolated values for luminosity. that is bright and dark areas.
this doesn't change the fact that red light will get through only one of the 4 pixels due to the colour filter array above the sensor, similarly blue will enter only the pixel with the blue filter.
by removing the CFA, all wavelengths reach the four pixels and hence provides higher sensitivity and resolution.
That's the reason mono CCD's with LRGB images have more detail in them.
as you may know, with mono imaging, its the "L" channel that carries all the details.

This CFA mod appealed to me because I could use a relatively large sized chip for mono imaging at a fraction of the cost, compromise is the lower QE, noise, and lower dynamic range compared to astro CCD's.
it also allows you to do narrowband imaging using the full resolution of the chip. I've done Ha imaging with an OSC, and I can only use the red channel. so a 11M image becomes a 3M image as its only 1/4th the data.
with the 350d now being mono, I can use all pixels for Ha or other NB.


Edit: the company maxmax which also does BW conversions have a very good write-up here with example images clearly showing the difference.
http://www.maxmax.com/b&w_conversion.htm
They do sell B&W dslr's but for that price, you could get a mono ccd.

i'll post my own results a little later on.


Cheers
Alistair

Last edited by alistairsam; 09-07-2013 at 05:40 PM.
Reply With Quote
  #19  
Old 09-07-2013, 06:41 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Greg, in terms of QE, I know the dslr's are only between 25 to 30%. would you know of any mid-range dslr's that have relatively high QE?
I don't think I'll try the Fuji or the Sony just yet.

yes, Guy, author of BEOS has also done mono conversions for Nikons. I prefer Canon's just for the software support although if I can debayer a D40, I might take on a QHY8!!
two second hand QHY8's (in case I break one), are still cheaper than a comparable mono.

Cheers
Alistair[/QUOTE]

I did have a link at one stage to a listing of various DSLRs and their QE and read noise.

Nikon D5200 was highest with a Toshiba APSc sensor and around 60% QE. Nikon D800E was next at 59%.

Pentax K5 might be one to look at. Its the same Sony 16.3mp Exmor APSc sensor in the Fuji XE1 and Nex series. Its also in the Nikon D7000 which may be cheaper now there is a D7100.

Of the Canons a 5D may be good. It has over 30% QE as I recall and largish pixels (still small by CCD stds). They are also sometimes quite cheap.

A 5D2 is around 26% or so and 5D3 was 39%. I will see if I can find that listing and post the link.

I like your work. I think this could be a productive use of a DSLR and turn them into a nice imaging machine potentially.

59% QE with a Bayer CFA and UV/IR block in D800E. Wow, imagine a naked mono sensor what QE it must run at - it must be up over 80%.

Greg.
Reply With Quote
  #20  
Old 11-07-2013, 11:54 AM
rcheshire's Avatar
rcheshire (Rowland)
Registered User

rcheshire is offline
 
Join Date: Apr 2010
Location: Geelong
Posts: 2,617
Alistair. If you still need a filter I have one in a spare 1000D sensor, if it will fit. I can send by post when I get back from Brisbane next week?

In the 1000D, the top glass is the AA filter/Piezo, while the glass in the plastic frame closer to the sensor is the IR-Cut filter - replace with Baader or Astrodon and leave the AA out altogether as it softens focus.

Something I've been meaning to develop for a while is a sensor face heater for a DSLR - it needs to be very small, but having cooled two DSLR's by various means, sensor face heating is essential. Small sticky strip with Nichrome filament.
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 09:00 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement