Inexpensive episcopic MTF target?

[blekenbleu]

Full MTF measurements are not intended,
instead using ImageJ line profiles to help determine sharpening (unsharp mask) parameters.
I am familiar with slanted-edge and smoked razor blade techniques,
but my Optiphot-66 lacks back-illumination capability.
For those unfamiliar with ImageJ, it is totally free (public domain) and quite easy to use,
at least for making line intensity profiles:

... in this case, drawing a line on Smokedaddy's MTF Mapper backlit target image,
then selecting Analyze > Plot Profile (or press Ctrl+K):


However, doing the same with an opaque target, in this case a coated playing card:

... yields less pleasant results:


At least part of that can be blamed on uneven coating, but also optical dot gain.
A suitable target would have opaque smooth substrate, to combat dot gain,
along with a thin contrasting opaque coating, evenly applied.
Credit cards seem hopeless, as is the substrate for my driving license,
but coating for its barcode is at least moderately uniform and opaque:


Any suggestions?

[Lou Jost]

How about test wafers? Widely used here.

[rjlittlefield]

I suggest to mount a razor blade over a "black hole", with the razor blade angled so as to reflect any convenient light source.

A simple piece of black flocking material might be black enough for your purposes. If not, then a shallow box lined with black flocking, totally enclosed except for a small opening over which the razor blade is held.

--Rik

[Adalbert]

Hi Blekenbleu,
Razor blade, as Rik has already mentioned.
Please take a look at this thread:
viewtopic.php?p=226318#p226318
Best, ADi

[blekenbleu]

How about test wafers? Widely used here.

If "test wafer" means some integrated circuit,
I was looking for something that provides a single abrupt reflectance transition
(step function) at many scales e.g. 3:1 to 1:20...

I found an old debit card with a magnetic stripe:

.. which substrate is more opaque than drivers license (IMG_0128):

... then tried a worn safety razor blade, propped up by a coin for an edge tight against the mag stripe:

... which may be yielding a useful profile on the mag stripe side (IMG_0130):

... but overall the mag stripe edge against credit card substrate seems credible
for this combination (Canon 90D with Laowa 25mm f/2.8 2.5-5X EF @ 5x)
suggests trying an unsharp mask radius of 15-20 pixels or so.

FWIW, one end of the debit card was propped up by a coin,
to allow drawing ImageJ's line across the imaged stripe edge where seemingly most in focus.

[Lou Jost]

Seems to me that many wafers offer abrupt transitions like that, at least until you get to high enough magnifications to see the imperfections (maybe 50x+). You may need to use cross-polarized light if using epi illumination through the objective, to make sure the wafer is black. See any of Robert OToole's tests.
Lou

[rjlittlefield]

blekenbleu, your last post was made with a checkmark on "Disable BBCode".

That made all the words run together as text with visible "[attachment]" tags, while the photos appeared at the end of the post without comment.

I did an Admin edit to remove that checkmark, so that your images would appear in-line as I believe you wanted them.

Remember that there is a Preview button so that you can check the appearance of your post before hitting Submit. The preview will show whether BBCode tags are being interpreted correctly.

--Rik

[blekenbleu]

Seems to me that many wafers offer abrupt transitions like that, at least until you get to high enough magnifications to see the imperfections (maybe 50x+). You may need to use cross-polarized light if using epi illumination through the objective, to make sure the wafer is black. See any of Robert OToole's tests.
Lou

Agreed that chips offer plenty of abrupt transitions,
less certain about adequate areas of constant intensity either side of a transition.
My 5x setup has about 1450 pixels/mm and wants constant intensity areas of
about 200x200 pixels or 1/7mm squares for unsharp radius estimation.

Estimating 1/3x magnification unsharp radius would want roughly 2x2mm constant intensity areas.

[blekenbleu]

blekenbleu, your last post was made with a checkmark on "Disable BBCode".

That made all the words run together as text with visible "[attachment]" tags, while the photos appeared at the end of the post without comment.

I did an Admin edit to remove that checkmark, so that your images would appear in-line as I believe you wanted them.

Remember that there is a Preview button so that you can check the appearance of your post before hitting Submit. The preview will show whether BBCode tags are being interpreted correctly.

--Rik

Sorry, I seem to have forgotten whatever I had once figured out about posting images here.
I just now tried changing that profile setting, but cannot see that preview changed.
The post with images by URL did not embed with "Disable BBCode" unchecked,
so concluded that I really do not know what BBCode does or does not do.
Whoever said "ignorance is bliss" is not me...

Rik - thanks ever so much for your patient help; I was somehow blind to per-post options; went back and repaired the first post.

[rjlittlefield]

I just now tried changing that profile setting, but cannot see that preview changed.
The post with images by URL did not embed with "Disable BBCode" unchecked,
so concluded that I really do not know what BBCode does or does not do.

Each post has its own independent setting for "Disable BBCode". That setting affects only that one post.

The user profile setting for "Enable BBCode by default" determines how each post's "Disable BBCode" is set, only when that post gets created.

Once a post has been created, changing the profile setting will not affect the post.

I have just now edited your last post (second above here) to un-check "Disable BBCode", so it now displays OK.

I looked at your profile just now, and I see that it says "Enable BBCode by default: Yes". So, future posts should be fine. If they are not, we will debug further.

--Rik

[blekenbleu]

Supposing that 20 pixel radius is approximately correct,
that value was plugged into ImageJ's Unsharp Mask filter, which is less than ideal:

• only 2 parameters (radius and strength, no threshold)

• no undo (so must reload the image to try different strengths

On the other hand, the line for Plot Profile remains intact
before and after applying the filter, so profiling the same pixels.
Here are before and after 1:1 crop images, with corresponding profiles:

Unsharp mask w/20 pixel radius and 0.2 strength (range 0.1 - 0.9):

Lacking threshold control, unsharp mask effect on noisy images is similar to contrast boost.
Sadly, Laowa 25mm lens information is not in EXIF.

[enricosavazzi]

It depends on how inexpensive, and on the required pattern, but just 1-2 days ago I saw a few second-hand MTF test targets on eBay (chrome-plated on glass, so they require a suitable illumination and the addition of flocking/black tape on the back of the target for episcopic use). For example eBay item 384631220753.

[blekenbleu]

just 1-2 days ago I saw a few second-hand MTF test targets on eBay (chrome-plated on glass,
so they require a suitable illumination and the addition of flocking/black tape on the back of the target for episcopic use).
For example eBay item 384631220753.

Thanks for the pointer.
To kill razor blade edge specularities, flat white spray paint was ordered.
I make few transmitted light microscope images,
and the last time that I tried reflected measurements involving glass,
enough light bounced around in the glass to confound results
(in fact, I developed a model for mitigating those artifacts),
but playing with different targets might inspire some other exercises.

[blekenbleu]

Krylon matte spray dries more nearly eggshell than matte.
Despite edge-up orientation while being sprayed (booth courtesy Amazon)

.. dried edge has paint blobs.

Tipping razor for edge tight against debit card mag stripe
and tipping mag stripe to vary focus along the edge

...yields a dark/light diagonal (downscaled)

Browsing that image 5x for a spot most in focus without a paint blob:

[blekenbleu]

Editing these posts with attached images successfully is obviously beyond me.
Regardless, here is the resulting ImageJ line profile.
Lighter side of the edge is useless; dark side suggests approx 4 pixel radius: