Advice on building multi-channel image splitter

[Adam Glaser]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Hello everyone,

I was wondering if anyone could offer advice on building an in-house image splitter for microscopy.  

We are hoping to build one to split a rectangular ROI vertically into 2-4 colors.  The design seems straightforward, similar to the schematic of the Optosplit by Andor http://www.andor.com/scientific-cameras/multi-wavelength-imaging/optosplit-iii#prettyPhoto[pi_gal]/1/.

Are there any non-obvious factors to consider when building?  We will be using 2 identical tube lenses, the first to image an adjustable slit for creating the cropped rectangular ROI, the second to re-focus onto our sCMOS camera, and super resolution imaging grade dichroics from Semrock.

Thanks,
Adam

[Michael Giacomelli]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Hi Adam,

The main issue we have flatness, since even very flat ones will introduce
geometric distortion.   Rather than use the high grade ones, we just
calibrate it out by imaging a grid target with white light and then using
imagej to compute the unwarp matrix.  Takes less than 100 ms per megapixel
per core to unwarp two channels in matlab this way.  Make sure you have
tip/tilt control somewhere for the dichroics as well, otherwise centering
the cameras is very painful.

Christopher Rowlands, who I think you met, also had a clever design at SPIE
this year for 6 (or maybe 8?) channel detection that balanced distortion
between channels to minimize the total error.  Not sure if you need
something like that, but you could run your design by him.

Mike

On Mon, Jul 31, 2017 at 4:11 PM, Adam Glaser <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Hello everyone,
>
> I was wondering if anyone could offer advice on building an in-house image
> splitter for microscopy.
>
> We are hoping to build one to split a rectangular ROI vertically into 2-4
> colors.  The design seems straightforward, similar to the schematic of the
> Optosplit by Andor http://www.andor.com/scientific-cameras/multi-
> wavelength-imaging/optosplit-iii#prettyPhoto[pi_gal]/1/.
>
> Are there any non-obvious factors to consider when building?  We will be
> using 2 identical tube lenses, the first to image an adjustable slit for
> creating the cropped rectangular ROI, the second to re-focus onto our sCMOS
> camera, and super resolution imaging grade dichroics from Semrock.
>
> Thanks,
> Adam
>

[Sripad Ram-2]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Hi Adam,
In my experience, mechanical stability is another issue that you need to
take into account. How are you planning to mount and adjust the
dichorics/mirrors? When we built ours (which was years ago), we had to
contend with whatever  micro-optomechanical mounts we could get and they do
drift with time. Things could have improved and you want to look for
mounts/holders that have ultralow or zero drift.



On Mon, Jul 31, 2017 at 1:32 PM, Michael Giacomelli <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Hi Adam,
>
> The main issue we have flatness, since even very flat ones will introduce
> geometric distortion.   Rather than use the high grade ones, we just
> calibrate it out by imaging a grid target with white light and then using
> imagej to compute the unwarp matrix.  Takes less than 100 ms per megapixel
> per core to unwarp two channels in matlab this way.  Make sure you have
> tip/tilt control somewhere for the dichroics as well, otherwise centering
> the cameras is very painful.
>
> Christopher Rowlands, who I think you met, also had a clever design at SPIE
> this year for 6 (or maybe 8?) channel detection that balanced distortion
> between channels to minimize the total error.  Not sure if you need
> something like that, but you could run your design by him.
>
> Mike
>
> On Mon, Jul 31, 2017 at 4:11 PM, Adam Glaser <[hidden email]>
> wrote:
>
> > *****
> > To join, leave or search the confocal microscopy listserv, go to:
> > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> > Post images on http://www.imgur.com and include the link in your
> posting.
> > *****
> >
> > Hello everyone,
> >
> > I was wondering if anyone could offer advice on building an in-house
> image
> > splitter for microscopy.
> >
> > We are hoping to build one to split a rectangular ROI vertically into 2-4
> > colors.  The design seems straightforward, similar to the schematic of
> the
> > Optosplit by Andor http://www.andor.com/scientific-cameras/multi-
> > wavelength-imaging/optosplit-iii#prettyPhoto[pi_gal]/1/.
> >
> > Are there any non-obvious factors to consider when building?  We will be
> > using 2 identical tube lenses, the first to image an adjustable slit for
> > creating the cropped rectangular ROI, the second to re-focus onto our
> sCMOS
> > camera, and super resolution imaging grade dichroics from Semrock.
> >
> > Thanks,
> > Adam
> >
>

[Peter Brunt AVR]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

*Commercial Response* From AVR Optics (Semrock's research and academic distributor for the US and Canada)

Hi Adam,

As previously mentioned in this thread, the key factor in this type of setup is the flatness of the splitting dichroic.  An insufficient flatness will introduce wavefront errors that will distort your image.  With a lot of standard dichroics which are used in a microscopes epi-fluoresence path, the flatness requirements are not as strict as is typically required for splitting images.  To address this, Semrock, have developed a range of image splitting dichroics.  These use a thicker 3mm glass substrate that will support a higher lambda/10 flatness, equivalent to ~lamda/5 reflected wavefront error.  This level of flatness is generally regarded as a strong standard of optical flatness for even the most demanding optical setups and should be ideal for image splitting.  

Details on Semrock's range can be found online here: https://www.semrock.com/image-splitting-dichroics.aspx

As an additional note, I think this has already been mentioned by mounting will also come into play.  I would recommend using higher quality optical mounts for these elements to avoid drift over time.  Everything from mechanical shock to thermal changes in the environment can cause the mechanics to move so something with a locking mechanism or a well thermally balanced mount will save a lot of time in the long run.  

I hope this helps.


Best Regards

Peter Brunt

AVR Optics

[Zdenek Svindrych-2]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Hi Adam,
generally, you can split and combine either in the "aperture space"
(infinity space) or "image space". Both have some interesting properties,
pros and cons.

The (very common) Andor approach (splitting and combining in the infinity
space) means that the different "channels" arrive at the sensor from
different angles. Possibly important implications are:

Vignetting: with zoom 1 optics the beam size at your combining mirrors (that
are close to the aperture plane) is the same as the back focal plane (BFP)
aperture of the objective lens. Away from the aperture plane the beam will
likely be even bigger. You want to make sure you are not blocking part of
that beam by other mirrors. This would reduce both intensity AND
resolution...

Aberrations: apart from the possible aberrations introduced by the
dichroics, the last lens will be critical, too. The aperture of this lens
will be much higher than that of the rest of the optics, by a factor of 2-4
(the number of channels) or even more, depending how close you can get the
individual beams when combining (D-mirrors will be helpful here). It may
happen that a doublet (even at f = 180 mm) may not perform diffraction-
limited-ly. Also, you may find that 1-inch optics is just too small,
especially with big CMOS chips.

Note: the flat dichroics are really flat when Semrock mounts them into their
TIRF cubes. You need to be careful (or ask Semrock) if you want to mount
them yourself...

As mentioned before, you will need fine adjustment mounts for both the
dichroics and the combining mirrors to get both the aperture and the field
exactly where you want them.

Good luck!

Best, zdenek
--
Zdenek Svindrych, Ph.D.
W.M. Keck Center for Cellular Imaging (PLSB 003)
Department of Biology,University of Virginia
409 McCormick Rd, Charlottesville, VA-22904
http://www.kcci.virginia.edu/
tel: [phone]

---------- Původní e-mail ----------
Od: Adam Glaser <[hidden email]>
Komu: [hidden email]
Datum: 31. 7. 2017 16:11:43
Předmět: Advice on building multi-channel image splitter
"*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Hello everyone,

I was wondering if anyone could offer advice on building an in-house image
splitter for microscopy.

We are hoping to build one to split a rectangular ROI vertically into 2-4
colors. The design seems straightforward, similar to the schematic of the
Optosplit by Andor http://www.andor.com/scientific-cameras/multi-wavelength-
imaging/optosplit-iii#prettyPhoto[pi_gal]/1/.

Are there any non-obvious factors to consider when building? We will be
using 2 identical tube lenses, the first to image an adjustable slit for
creating the cropped rectangular ROI, the second to re-focus onto our sCMOS
camera, and super resolution imaging grade dichroics from Semrock.

Thanks,
Adam
"

[James Kerin]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

*Commercial response from Cairn Research Ltd - The home of emission image splitting*While it is of course possible to correct for distortion afterwards, in our opinion it's better to minimise it in the first place, as we aim to do in our products.&nbsp; It has to be a good scientific principle to "mess around" with your data as little as possible!We've been familiar with these issues for quite a while, and we were instrumental in Chroma Technology's introduction of "ultraflat" dichroics, in order to minimise distortion from that source.The other important consideration is to reduce differences in distortion between channels by making the optical pathways as similar as possible.&nbsp; We consider this to be a key advantage of our Optosplit, Triplesplit and Multisplit image splitters in particular, but it also applies to our multicamera products.&nbsp; Seeing is believing, so we're not afraid of any comparative tests here!&nbsp;&nbsp;
https://www.cairn-research.co.uk/products/fluorescence-detection/image-splitters/
&nbsp;
Dr Martin Thomas, President, Cairn Research Ltd
&nbsp;
&nbsp;

&nbsp;
 
&nbsp;
 
&nbsp;

&nbsp;
&nbsp;
&nbsp;
&nbsp;

&nbsp;




On  07/31/2017, 09:11pm, Adam Glaser ([hidden email]) wrote:
*****
 
To join, leave or search the confocal microscopy listserv, go to:
 
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
 
Post images on http://www.imgur.com and include the link in your posting.
 
*****
 

 
Hello everyone,
 

 
I was wondering if anyone could offer advice on building an in-house image splitter for microscopy.  
 

 
We are hoping to build one to split a rectangular ROI vertically into 2-4 colors.  The design seems straightforward, similar to the schematic of the Optosplit by Andor http://www.andor.com/scientific-cameras/multi-wavelength-imaging/optosplit-iii#prettyPhoto[pi_gal]/1/.
 

 
Are there any non-obvious factors to consider when building?  We will be using 2 identical tube lenses, the first to image an adjustable slit for creating the cropped rectangular ROI, the second to re-focus onto our sCMOS camera, and super resolution imaging grade dichroics from Semrock.
 

 
Thanks,
 
Adam

[Adam Glaser]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Thanks everyone for the input.  Optics wise, is it important to ensure that the separated images are incident on the final tube lens parallel to the optical axis?  And simply shifted laterally to achieve the spatial offsets?  As in this cartoon illustration: http://www.andor.com/scientific-cameras/multi-wavelength-imaging/optosplit-iii#prettyPhoto[pi_gal]/1/

Or is generating the spatial offsets using a combination of tilt/tip of the dichroics and/or mirrors also adequate?  In this case the different beam paths might not all be perfectly parallel to the optical axis.

I am just thinking through my head the alignment/optimization of our home built image splitter.

Thanks again,
Adam

[Craig Brideau]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

The parallel image paths must both pass through the same tube lens to be
focused on the CMOS/CCD array. You will want to make sure the two paths
pass on either side of the center of the tube lens to get adequate focusing
onto the array. If the tube lens diameter is too small compared to the
combined width of the beam paths you will see distortion or lateral defocus
from operating too close to the edge of the lens. Alternatively you can use
two separate tube lenses with long working distances (200mm is typical for
a tube lens anyway) and combine them with a dichroic as close to the lenses
as possible. This could lead to slight distortion if the dichroic is of
mediocre quality as you will not be in infinity space when you combine the
two light paths. That said, you can then easily offset one path relative to
the other by translating the position of the dichroic, and the image cast
onto the CMOS/CCD should be flat for both paths with minimal distortion.

Craig

On Wed, Aug 2, 2017 at 2:02 PM, Adam Glaser <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Thanks everyone for the input.  Optics wise, is it important to ensure
> that the separated images are incident on the final tube lens parallel to
> the optical axis?  And simply shifted laterally to achieve the spatial
> offsets?  As in this cartoon illustration: http://www.andor.com/
> scientific-cameras/multi-wavelength-imaging/optosplit-
> iii#prettyPhoto[pi_gal]/1/
>
> Or is generating the spatial offsets using a combination of tilt/tip of
> the dichroics and/or mirrors also adequate?  In this case the different
> beam paths might not all be perfectly parallel to the optical axis.
>
> I am just thinking through my head the alignment/optimization of our home
> built image splitter.
>
> Thanks again,
> Adam
>

[Zdenek Svindrych-2]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****


Hi Adam,
it's just a cartoon...
Ray-optical properties of a thin lens are simple. The rules are (1) a bundle
of parallel rays will be focused by the lens to a spot somewhere on the
focal plane (and vice versa), and (2) rays passing through the center of a
thin lens continue un-deviated.
So clearly, to achieve spatial offset in the focal plane the rays of the
individual channels have to impinge on the lens from different angles. The
angles will be fairly close to parallel with the optical axis for f=200mm
lens.

It really may be more difficult than it seems, I recommend using some
raytracing software so you get an idea. Or google for "telecentric system"
or "4f system".

Note also, that real lenses are not thin lenses and the spectral properties
of the dichroics depend on the incidence angle...
You may also find that a more complicated design with individual focusing
lenses gives you more flexibility in alignment...

See e.g. this cartoon:

https://www.cairn-research.co.uk/wp-content/uploads/2015/12/Optosplit-III-LS
-Image-Splitter1.jpg

Best, zdenek


--
Zdenek Svindrych, Ph.D.
W.M. Keck Center for Cellular Imaging (PLSB 003)
Department of Biology,University of Virginia
409 McCormick Rd, Charlottesville, VA-22904
http://www.kcci.virginia.edu/
tel: [phone]

---------- Původní e-mail ----------
Od: Adam Glaser <[hidden email]>
Komu: [hidden email]
Datum: 2. 8. 2017 16:03:11
Předmět: Re: Advice on building multi-channel image splitter
"*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Thanks everyone for the input. Optics wise, is it important to ensure that
the separated images are incident on the final tube lens parallel to the
optical axis? And simply shifted laterally to achieve the spatial offsets?
As in this cartoon illustration: http://www.andor.com/scientific-cameras/
multi-wavelength-imaging/optosplit-iii#prettyPhoto[pi_gal]/1/

Or is generating the spatial offsets using a combination of tilt/tip of the
dichroics and/or mirrors also adequate? In this case the different beam
paths might not all be perfectly parallel to the optical axis.

I am just thinking through my head the alignment/optimization of our home
built image splitter.

Thanks again,
Adam
"