Re: CONFOCALMICROSCOPY Digest - 27 Jul 2017 to 28 Jul 2017 (#2017-167)

[Daniel White-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 Zdenek, and all,

*** Disclaimer - I work for GE Heathcare as applications specialist for OMX
3D-SIM ***

Regarding your description of the OMX Blaze 3D-SIM  light path (fast stripe
generator):
I just wanted to add a little detail and perhaps clear up / emphasise a
couple of ideas for the larger confocal community.

There _IS_ a diffraction grating to generate the stripes:
Not some kind of mirrors or prism beam splitters to generate the 2 or 3
beams required.
The only beam splitters are in the optical block filtersets, autofocus etc.

The difference between OMX v4 and OMX SR with Blaze, and other basic SIM
systems, all using a grating,
is that the OMX grating is fixed in position, there is only 1, and it never
moves! Good for stability.

The stripes are moved laterally by phase shifting the diffracted laser
beams with glass windows on galvos - which is where much speed comes from.
This also means the 1st order beam positions at the back of the objective
lens can be adjusted very quickly on the fly,
in order to optimise the image plane stripes' linewidth to be as small as
possible for each laser illumination wavelength:
No need to change the grating or accept worse than optimal SIM resolution
for eg. red and far red, when going fast and live.

It also means that the optimal stripes pattern focus can be tuned
opto-mechano-electronically instead of mechanically moving a large heavy
grating,
This make it possible to fine tune calibrate and then optimise pattern
focus (top phase in OMX speak)  very precisely, reproducibly, and stably,
for each of the three stripes angles, and for all illumination wavelengths
during a multi colour fast SIM experiment.

The Stripes are rotated using a galvo mounted mirror pointing to 3 pairs of
beam rotating fixed mirrors. Again good for stability and speed.

I think you can still say the grating is imaged onto the sample.... but it
is a bit more complicated than in a simple system.
The grating image is effectively magnified onto the sample differently on
the fly for each illumination laser wavelength,
in order to optimise for smallest possible line spacing for each colour
channel sequentially during a z-slice.

So I am not 100% sure what that does to the arguments you nicely
explained... if we are in case 1 or 2?

But it does mean OMX SR Blaze v3, 3D-SIM / TIRF-SIM / 2D-SIM illuminator
can make very high contrast stripes at the "resolution limit",
for each illumination laser and quickly switch
a)  angle of stripes
b)  stripe lateral position (phase)
and also
c)  illumination wavelength stripes linewidth optimisation on the fly
during eg. the z stack, and time series.
Multiple cameras also means no slow filter wheel changes, and also means
fast channel switching.

I like that this discussion is here on the confocal list, because from my
perspective,
3D-SIM is just "confocal done right".

Abbe was around before we knew about fluorescence illumination modulation
or else his resolution estimation equation would be the SIM case of

d =  lambda / (4*NA)

(not 2*NA)

For me, 3D-SIM is what comes after and succeeds a conventional point scan
confocal - its much more efficient:
For properly mounted samples, it does much of what a confocal tries to do,
but much better,
 in terms of speed over a large FOV, resolution and contrast.
(if not in terms of deep penetration in a scattering sample, or simple
homogenous high background rejection
but both of those are sample prep problems, the hammer of the confocal
pinhole is used to squash)
Tip: clear the fixed sample into glass like RI medium and then 3D-SIM
probably works great at depth,
or for living samples use high % iodixanol (optiprep) to raise the RI as
high as you can,
 see:

https://elifesciences.org/articles/27240
)

>
> Date:    Fri, 28 Jul 2017 21:34:12 +0200
> From:    [hidden email]
> Subject: Re: Illumination source coherence requirement for SIM
>
> *****
> 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.
> *****
>
> Let me just add, that it depends a lot on the way you create your pattern.
> If you use fancy beamsplitters to get your two or three spots in the
> objective BFP (the "GE OMX SIM" arrangement, for example), then the
> temporal
> coherence (and, conversely, the spectral width) is of utmost importance. In
> this case the spatial frequency of the pattern depends linearly on
> wavelength, so if you want a good pattern contrast across, lets say, 1000
> lines in your field of view, you need linewidth much smaller than 1/1000 of
> the central wavelength. Easy with lasers, impossible with LEDs.
>
> On the other hand, in the more traditional setup (e.g. Nikon N-SIM), where
> you are imaging the pattern into the sample (as Lu and Marco mentioned) the
> frequency of the pattern is given by the grating (or SLM patter) period, so
> the contrast is high even with broadband source (note that for the shorter-
> wavelength components the pattern frequency is less than the maximum
> possible, and the long-wavelength components may miss the BFP aperture
> completely, leading to zero contrast). Or, from another point of view: the
> path difference between the two (or three) beams in the BFP is always zero
> in this case. Also, the +/-1 "spots" in the BFP are no longer spots, but
> little patches of rainbow...
>
> Hope it's not too confusing.
>
> Disclaimer: no commercial interest.
>
> Disclaimer: I have not confirmed this experimentally. I still may be wrong.
> Has anyone tried "broadband" SIM, e.g. with superluminescent diodes (not
> really useful for fluorescence excitation due to red-NIR wavelengths), or
> supercontinuum lasers?
> Best, zdenek
>
>
>

[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 Daniel,
thanks for clarifying. Yes, it makes more sense to use grating to split the
beam. I must have confused the OMX approach with some drawings from
unrelated papers. Sorry for that!

Honestly, I don't know what would be the effect of reduced spatial or
temporal coherence of the light source on the pattern contrast. But assuming
that keeping the path lengths of the individual beams equal (to within one
wavelength) was not one of the design goals (especially if you use the
galvos to fine-tune the position of each beam individually), it won't work
with incoherent illumination. Not that it would matter... It's supposed to
work with lasers, and it works well.

I agree, that in non-scattering sparsely-labelled samples SIM works great.
But in less-than-optimal samples confocal is, and will be, the golden
standard of 3D imaging, as the risk of artifacts is far lower that in any
other method.

And thanks for pointing out the OptiPrep RI matching for live samples. It
seems even cheaper than Histodenz, might be very useful in lightsheet
imaging... 

Best, zdenek


---------- Původní e-mail ----------
Od: Daniel White <[hidden email]>
Komu: Confocal Microscopy List <[hidden email]>, zdedenn@
gmail.com
Datum: 17. 8. 2017 10:11:06
Předmět: Re: CONFOCALMICROSCOPY Digest - 27 Jul 2017 to 28 Jul 2017 (#2017-
167)
"



Hi Zdenek, and all, 




*** Disclaimer - I work for GE Heathcare as applications specialist for OMX
3D-SIM ***




Regarding your description of the OMX Blaze 3D-SIM  light path (fast stripe
generator):

I just wanted to add a little detail and perhaps clear up / emphasise a
couple of ideas for the larger confocal community.




There _IS_ a diffraction grating to generate the stripes:

Not some kind of mirrors or prism beam splitters to generate the 2 or 3
beams required. 

The only beam splitters are in the optical block filtersets, autofocus etc. 




The difference between OMX v4 and OMX SR with Blaze, and other basic SIM
systems, all using a grating, 

is that the OMX grating is fixed in position, there is only 1, and it never
moves! Good for stability. 




The stripes are moved laterally by phase shifting the diffracted laser beams
with glass windows on galvos - which is where much speed comes from.

This also means the 1st order beam positions at the back of the objective
lens can be adjusted very quickly on the fly,

in order to optimise the image plane stripes' linewidth to be as small as
possible for each laser illumination wavelength:

No need to change the grating or accept worse than optimal SIM resolution
for eg. red and far red, when going fast and live. 




It also means that the optimal stripes pattern focus can be tuned opto-
mechano-electronically instead of mechanically moving a large heavy grating,

This make it possible to fine tune calibrate and then optimise pattern focus
(top phase in OMX speak)  very precisely, reproducibly, and stably,

for each of the three stripes angles, and for all illumination wavelengths
during a multi colour fast SIM experiment. 




The Stripes are rotated using a galvo mounted mirror pointing to 3 pairs of
beam rotating fixed mirrors. Again good for stability and speed.  




I think you can still say the grating is imaged onto the sample.... but it
is a bit more complicated than in a simple system. 

The grating image is effectively magnified onto the sample differently on
the fly for each illumination laser wavelength,

in order to optimise for smallest possible line spacing for each colour
channel sequentially during a z-slice.




So I am not 100% sure what that does to the arguments you nicely
explained... if we are in case 1 or 2? 




But it does mean OMX SR Blaze v3, 3D-SIM / TIRF-SIM / 2D-SIM illuminator can
make very high contrast stripes at the "resolution limit",

for each illumination laser and quickly switch 

a)  angle of stripes

b)  stripe lateral position (phase)

and also 

c)  illumination wavelength stripes linewidth optimisation on the fly during
eg. the z stack, and time series. 

Multiple cameras also means no slow filter wheel changes, and also means
fast channel switching. 




I like that this discussion is here on the confocal list, because from my
perspective, 

3D-SIM is just "confocal done right".




Abbe was around before we knew about fluorescence illumination modulation 

or else his resolution estimation equation would be the SIM case of 




d =  lambda / (4*NA)




(not 2*NA)




For me, 3D-SIM is what comes after and succeeds a conventional point scan
confocal - its much more efficient:

For properly mounted samples, it does much of what a confocal tries to do,
but much better,

 in terms of speed over a large FOV, resolution and contrast.

(if not in terms of deep penetration in a scattering sample, or simple
homogenous high background rejection

but both of those are sample prep problems, the hammer of the confocal
pinhole is used to squash)

Tip: clear the fixed sample into glass like RI medium and then 3D-SIM
probably works great at depth,

or for living samples use high % iodixanol (optiprep) to raise the RI as
high as you can,

 see:


https://elifesciences.org/articles/27240
(https://elifesciences.org/articles/27240)


)







 
"
Date:    Fri, 28 Jul 2017 21:34:12 +0200
From:    [hidden email](mailto:[hidden email])
Subject: Re: Illumination source coherence requirement for SIM

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

Let me just add, that it depends a lot on the way you create your pattern.
If you use fancy beamsplitters to get your two or three spots in the
objective BFP (the "GE OMX SIM" arrangement, for example), then the temporal
coherence (and, conversely, the spectral width) is of utmost importance. In
this case the spatial frequency of the pattern depends linearly on
wavelength, so if you want a good pattern contrast across, lets say, 1000
lines in your field of view, you need linewidth much smaller than 1/1000 of
the central wavelength. Easy with lasers, impossible with LEDs.

On the other hand, in the more traditional setup (e.g. Nikon N-SIM), where
you are imaging the pattern into the sample (as Lu and Marco mentioned) the
frequency of the pattern is given by the grating (or SLM patter) period, so
the contrast is high even with broadband source (note that for the shorter-
wavelength components the pattern frequency is less than the maximum
possible, and the long-wavelength components may miss the BFP aperture
completely, leading to zero contrast). Or, from another point of view: the
path difference between the two (or three) beams in the BFP is always zero
in this case. Also, the +/-1 "spots" in the BFP are no longer spots, but
little patches of rainbow...

Hope it's not too confusing.

Disclaimer: no commercial interest.

Disclaimer: I have not confirmed this experimentally. I still may be wrong.
Has anyone tried "broadband" SIM, e.g. with superluminescent diodes (not
really useful for fluorescence excitation due to red-NIR wavelengths), or
supercontinuum lasers?
Best, zdenek


"


"