Skip to main content

Phono stylus inspection pt. 2

In the first part of the article, we covered most of the theoretical limits of our proposed optical system that would be able to resolve contact surface of a phono stylus. If you haven’t read that, there’s a good chance you won’t understand half of what’s coming up, so make sure that doesn’t happen. I tried to simplify the mathematical expressions as much as possible, so there’s only basic algebra to work through. Still, there’s a lot to juggle, especially considering all the interactions between the parameters.

Luckily, we have great tools and programming languages to simplify things and make them more intuitive. Below is my take on that – it’s an optical calculator that helps with adapting a microscope objective to your camera for macro-photography. I originally wrote it as a quick helper tool for assembling my current optical inspection setup. It worked pretty well, so I decided to share it here. And before anyone asks – yes! It works not only for phono styluses but for bugs and bees too!

Optical System Calculator


Light

Sensor

Objective

Outputs

Input Field Explanations

Light

  • Wavelength: Slider to adjust wavelength of light illuminating object.
  • Wavelength (μm): Wavelength converted into micrometers.
  • Color: The perceived color corresponding to the selected wavelength.

Sensor

  • Type: Monochrome is set by default. This means only light intensity is sampled and sensor resolution is the same for all wavelengths. Color – assumes worst case Bayer filter influence on color sensor resolution when using monochromatic light.
  • Format: Select the type or size standard of the camera sensor to update sensor width and height.
  • Resolution: Select standard sensor resolutions to update sensor width in pixels.
  • Sensor width (mm): The physical width of the sensor in millimeters.
  • Sensor height (mm): The physical height of the sensor in millimeters.
  • Sensor width (pixels): The horizontal resolution of the sensor in pixels.
  • Pixels per Rayleigh distance: Number of pixels covering the diffraction-limited spot size (Rayleigh criterion).
  • Pixels per Circle of Confusion: Number of pixels spanning the diameter of the Circle of Confusion, relating to perceived sharpness.

Objective

  • Type: Finite is assumed by default.
  • Magnification: The optical enlargement provided by the objective lens.
  • Aperture (NA): Numerical Aperture; defines the light-gathering and resolving power of the lens.
  • Field Number: Diameter of the observable field in the intermediate image plane, in millimeters.
  • FOV object side (mm): Field of view on the sobject side, in millimeters.
  • Nominal Tube length (mm): Standard distance between the objective and eyepiece for focus and nominal magnification.
  • Used Tube length (mm): The actual distance used in the setup. For infinity-corrected objective this is tube lens focal length.

Output Field Explanations

Sensor

  • Sensor diagonal (mm): The diagonal size of the sensor in millimeters.
  • Pixel size (μm): The size of each pixel on the sensor, in micrometers.

Optics and Resolution

  • FOV image side (mm): Field of view diameter in the intermediate image plane.
  • Tube magnification: Magnification introduced by the mismatch of nominal and used tubes.
  • Crop magnification: Effect of the sensor size compared to a full-frame reference.
  • Real magnification: Combined magnification considering objective and sensor.

Aperture and Resolution

  • Effective aperture: The effective aperture at sensor, adjusted for the tube magnification and Numerical Aperture.
  • Optical resolution (μm): Smallest resolvable feature on the sensor, based on the wavelength and numerical aperture.
  • Sensor resolution (μm): Smallest resolvable feature based on pixel size and number of pixels per Rayleigh sampling distance.

Focus and Depth

  • Circle of Confusion (μm): Maximum blur diameter still perceived as sharp.
  • DOField (classical) (μm): Depth of field using geometrical optics.
  • DOField (wave) (μm): Depth of field considering wave optics and diffraction.
  • DOFocus (mm): Depth of focus at image plane where the image remains sufficiently sharp, when moving only sensor.
  • ΔTube Tolerance: Tolerance to variation of Tube Length, where we can refocus image and not loose any sharpness. Doesn’t apply to infinity-corrected objective.

Houston, we have a problem!

I don’t know about you, but the first time I played with this tool above, I was shocked! Even the lowest magnification objectives, like 2X 0.05 NA, have a depth of field of 220 μm. That means only 0.2 mm of our object’s depth can be captured in a single image. When we get into the magnification territory of our interest, like 10X or 20X, we find usual NA of 0.25–0.65 that produce depths of below 10 microns. And if we try to use near-UV with 0.65 NA, we end up with below 1 micron of depth of field!

Above, you can see the actual image captured by my camera through the 20X 0.4 NA microscope objective. Zoomed in part is showing the DOF – part of the object that is in focus. Wait, did we just go back to the beginning of the first article and see nothing again? Well, I’ll be damned…

But not all is lost. While it is still true that it is practically impossible to capture the stylus wear surface in sufficient detail and in one shot, we can try to cheat physics and create what I call an “Impossible Image.” Or “Image Impossible” for those who think adding a French name makes everything sound more cool – like croissants or haute couture!

Unfortunately, toasts and fancy pants will not get you there – you need algorithms and computing power, so probably not the most useful place for French words after all. We can use a technique called “Image Stacking,” and anyone who is into macro-photography reading this might say, “Duh… we’ve been using this stuff for decades!” Well, good for you. I, on the other hand, just discovered how cool it is and you can imagine my huge sigh of relief knowing I won’t have to crop and stitch every single image by hand. Or write my own software to do that! As that would be side-project number 101 on my list.

How does this work? We take multiple photos of the same subject at different focus points and combine them into one image. This creates a final picture with more depth of field and sharper details throughout. So, it’s really quite simple, right? Well, who am I fooling… By now, you know I wouldn’t be writing this if it was that simple.

Stepping the steps

First major problem we already identified. We need to take a lot of photos at a very small step size. How much and how small? Let’s say we want to capture just a tip of stylus that’s 0.1mm with 20X 0.4NA objective. Using our handy calculator we find a DOField of 3.3 μm so that’s 100/3.3 ~ 30 photos to take. What if we want all stylus with cantilever? That’s usually about 1mm or 300 photos! You get the idea – it’s really a lot of photos and we need to be able to step our gear (or object) with accuracy of no less then 1 micron.

Now I come from industrial automation background and I have an idea of what it takes to machine something with accuracy of 1 micron. Believe my when I say a ton. Literally, we need as much mass and rigidity to keep things stable. Luckily here we are dealing with static loads and object that are moving very slow, so we can get away with some simple sliding bearings and small 0.9deg steppers.

However, if you think this setup will be cheap – guess again. Most popular options like the StackShot Macro Rail Package or WeMacro Rail will cost you around $800 and $400, respectively. You can probably cut that in half if you have masochistic tendencies and skip the automatic stepping, going for a manual knob instead. And I thought the audio hobby wasn’t cheap!

Another approach is to DIY something, and I was just about to go on a shopping spree for CNC parts when I realized I already own a CNC router! Another Duh moment. What was left to do was buy bellows to allow tube distance adjustments and 3D print a bunch of adapters. And voilà – la CNC de la photographie!

I will also add that if someone decides to buy this CNC router to only do macro photo – don’t, it’s a really bad idea. I have upgraded it with linear rails, ball screws on all axis and 0.9deg stepper on Z axis, otherwise you can forget about 1 μm step repeatability.

Choosing the objective

Finally! The moment I was waiting for – shopping for the microscope objectives. I checked my account for sufficient funds, made a list of what would be “nice to have”. Like covering a range from 5X to 20X or more, having good working distance of at least 10mm so I could fit cartridge on it’s side and usual “mid-level” NA’s.

I soon found out that I had overestimated my funds by quite a margin. It seems I’ll need to reevaluate my mortgage just to afford them all! Sigh… Why, whenever I start having fun, does everything have to be ruined by some nonsense like money, loans, and financial commitments?

OK, how about lowering expectations and sampling the market? What about under $100 objective? Sure, it means buying from China with no warranty and a shady return policy – so it’s basically burning your money. But hey, at least we can have fun doing that, right?

I diversified my samples as much as possible: a 20X AmScope from the USA, a 10X AmScope from a European importer, and a 5X LMP Apo directly from China. Don’t be fooled by AmScope’s “all ‘merica, yeah!” marketing – these are the same generic lenses from China you’d find on eBay or AliExpress. I also threw in a used ZEISS Plan for good measure – just for fun. As a bonus: the 10X, 20X, and ZEISS are finite objectives, meaning I won’t have to buy a tube lens straight away. Sweet!

Stacking the stack

While waiting for my “sample” objectives to arrive, I started researching image stacking software options. Surprisingly, there aren’t many – Zerene Stacker, Helicon Focus and Photoshop. Technically, Photoshop isn’t dedicated stacking software, but it can handle simple stacking tasks using scripts. It’s also the only one with a subscription plan, meaning you never truly own the software – you just pay for it indefinitely. Sorry, but I’ll never subscribe to that (and that’s how you know you’re getting old kids).

After experimenting with demos of Zerene Stacker and Helicon Focus, I found that both produce equally impressive results. Using the same settings and algorithms, I couldn’t spot any noticeable differences in the output images. This just goes to show that both software packages are excellent choices.

Helicon Focus Demo

What sold me on Helicon Focus was its ability to export a depth map as a 3D rendering. That’s probably my coolest discovery of 2024. You take lots of photos along one axis, and suddenly, you see your object in 3D. It completely blew my mind – it’s really nuts!

The pricing isn’t terrible either (currently on holiday sale for 205€), and guess what? You actually get to own the software, complete with all its feature updates. That feels almost unheard of in 2025!

Sampling the samples

With all this fun, I almost forgot that the lenses finally arrived. It’s time to see what a lens under $100 can do. But first, we need to find a target – something with periodic features and high contrast. There are dedicated targets for resolution checking like 1951 USAF Resolution Test Targets and it’s derivatives, but looking at lines is not fun and remember – we are having fun here!

The next best thing I could think of was a silicon wafer from a chip. Now that’s something interesting to examine! Since I’m in the lab, there are also tons of donor boards of all sorts that can be scavenged for ICs. From there, it’s just a matter of using a heat gun and some pliers to pop the die out like a cookie.

Say hello to a piece of flash memory! It’s a stack of 10 photos taken with 4 μm steps. I somehow forgot to photograph the IC it came from, but it’s quite an old one, so the features of the memory cells are clearly visible. The image above was captured using an AmScope 20X 0.4NA objective and a Micro Four Thirds sensor (4592 x 3448). Although this lens is marketed as “Achromat,” only about 60% of the field of view is corrected for chromatic aberrations, resulting in an effective field number (FN) of just 9 mm.

Pixel Perfect Example

AmScope 20X 0.4NA FN Center 100% Crop

AmScope 20X 0.4NA Field Curvature

Above is a 100% center crop – a section of the image displayed at its native resolution (1:1 pixel ratio) without resizing, compression, or any other adjustments. I added two measurements to indicate the scale and assess the resolution. As you can see, we can discern a low-contrast line of 0.82 μm against the background, which is remarkably within a theoretical 0.84 μm limit for a 0.4NA lens, as far as I can tell. That’s really impressive for a $35.99 objective!

I must admit, I’m really just eyeballing this “resolution.” If we want to be serious, we need to use a real test target and measure the Modulation Transfer Function (MTF). Then I’d have to unpack the whole concept of line-pair frequency and contrast modulation, which would require a Part III of this article. But since this is an audio site, I think we’ve ventured deep enough into this optical stuff for now, so I’ll leave it here.

Field curvature is terrible, so the real object FOV for even a perfectly flat surface is very small on a single image. But hey – it’s not a Plan Achromat, so at least nobody tried to falsely advertise it as such.

Well this was unexpected  – ZEISS is total nein nien nein! Terrible aberrations and vignetting on all corners and even looking at the thumbnail image it’s obvious that we are missing a lot of details. So it’s really just a “Plan”  without any chromatic corrections and effective FN is just 4.7 mm.

Pixel Perfect Example

ZEISS Plan 20X 0.45NA FN Center 100% Crop

ZEISS Plan 20X 0.45NA Field Curvature

Field curvature is corrected pretty well, as expected, but for our purposes, it’s really not important. Since we are using image stacking, it doesn’t matter that some parts of the image are out of focus at certain distances – we will still cover them by stepping. In a real microscope, you’re limited to only one distance per view and it’s a real problem.

Even though its NA is 0.45, the real resolution is much worse than AmScope’s 0.4. We can barely discern the central lines and are missing a lot of sharpness.

With 10X magnification, we can now fit an entire row of memory cells, but the problem remains the same – an effective FN of just 8.3 mm. Sure, we can fit more into those 8.3 mm, but it’s still only about 40% sensor coverage for a Micro 4/3 camera. I’m starting to feel like all these cheap Chinese objectives are designed for 1/2.3″ sensors with 8mm diagonal. They are commonly found in your typical ‘microscope camera’ sold everywhere.

Pixel Perfect Example

AmScope 10X 0.25NA FN Center 100% Crop

AmScope 10X 0.25NA Field Curvature

Even though I can easily discern theoretical 1.34 μm features, I don’t like this image at all—it’s very soft and blurry, much more so than the 20X image, which doesn’t make much sense. Are they sending lower-quality lenses to us poor Europeans and reserving the good ones for the USA? Or is this just an unlucky sample? Hard to say with a sample size of one. At least the field curvature is nearly ideal across the entire sensor range, which is great for microscope use, though irrelevant for our purposes.

To summarize this finite objective overview, I’d say I’m really happy with my AmScope 20X 0.4NA purchase. For macro-photography, it’s probably a no-go due to poor sensor coverage even on micro 4/3. But for technical inspection, where you’re only focused on small portions of the image? It’s excellent value for the money.

AmScope 20X 0.4NA crops at different tube lengths

It can also be stepped up / down ±20 mm, giving a total DOFocus of 40 mm without much loss in fidelity. That’s close to the theoretical 34 mm predicted by the calculator above. Since this is a very subjective evaluation, having it’s “subjective constant” – I’ll say it fits with observation quite well.

The other two objectives are completely unusable. They might be fine for microscope use, where your samples are already a bit fuzzy, but other than that, they’re garbage.

Into the infinity

Finally, it’s time to see what’s up with that “infinity space” and evaluate my 5X 0.15NA LMP DIC Semi APO. “LMP” here stands for “Long Working Distance Metallurgical Plan” and “Semi APO” means a semi-apochromatic lens with improved color correction compared to standard achromatic lenses. DIC is “Differential Interference Contrast”, a technique used in microscopy to enhance contrast in transparent samples. Usually DIC lenses have better optics and works for regular applications too. 

I had high hopes for this one, as this is not my first purchase from the “Scientific-Optics-Lab-OEM Factory Store.” I bought my stereo microscope there, and I’m happily using it to this day. Despite the cheeky name, it’s really one of the biggest optical factories in China, with its own side store. Rumors have it they make lots of optical parts for big names like Leica and Zeiss, but what do I know?

Just one problem though – I don’t have a tube lens. I could just use any photo lens with a focal length close to the required tube distance in reverse. But all my astro-photo toys range from 20mm up to 130mm only. So I started digging into what macro-photo people are using, and I found the photomacrography forum. It’s a really great resource by the author of Zerene Stacker, Rik Littlefield. Tons of good info, and I wish I had found it sooner.

It looks like the majority of posters using infinity-corrected objectives are using a macro add-on lens by RAYNOX. This is a Japanese company I must admit I’d never heard of, but apparently, they make really good quality optics. The model M-150 has an effective focal length of ~208mm, so it’s perfect for an objective with a 200mm tube length.

With some 3D-printed adapters, I rushed to take the first photo but realized I had almost run out of target space. Still – there it is above – almost a full flash memory chip in one picture. And I have to say, it’s a disappointment. Yes, it’s really semi APO – in the sense that all corrections are applied equally across all of the 25 mm of specified FN – but the quality of those corrections leaves much to be desired. The image is very soft and lacks sharpness.

Since our flash memory no longer has a repetitive pattern in the corners, I found a new target. Meet a webcam sensor! I have no idea what model it is, and pardon the dust, but it measures 1/3″ (4.8 mm x 3.6 mm). It comes from an era when 720p was “cutting edge,” so the pixel pitch is 4.8/1280 = 3.75 μm. If you zoom in enough, you can actually start to see the pixels! So this objective is close to its theoretical 2.24 μm resolution, but not quite there, I’m afraid.

This reminds me of a joke:

Engineer and scientist are participating in a survival challenge. First mission: start a fire.

The engineer grabs some rocks, sticks and some dry grass and in 10 minutes got a roaring fire going. Meanwhile, the scientist is meticulously crafting a research proposal titled, “The Effects of Combustion on Wild Environments”… and is still waiting somewhere in the jungle for the grant committee’s approval to buy matches!

Pixel Perfect Example

5X 0.15NA LMP DIC Semi APO Corners and Center 100% Crop

5X 0.15NA LMP DIC Semi APO Field Curvature

As can be seen from the above crops, there is no loss in fidelity across the entire 21.6 mm sensor, so it has at least an FN of that size (25 mm is declared). Field Curvature is reasonably good, but some vignetting is present. These days it’s probably most simple digital correction there is so not a problem really.

5X 0.15NA LMP DIC Semi APO tube length 160-240mm

With only 0.15NA the tube length can be varied between 120 mm and 280 mm, producing magnifications from 3X to 7X without any noticeable loss in image quality. Probably even more as the theoretical limit is 1700mm. Obviously I can’t test that with my setup.

All in all, I wouldn’t have much to complain about if this objective were specified at something like 0.08 NA, but as it stands – it’s not a great value for 70 €.

Time for the commitments

So, at the end of the day, I had one usable objective and a lot of regrets. But hey, at least these weren’t the $3k kind of regrets! I’ll probably keep the 5X lens for myself, as it’s only necessary for capturing an entire cantilever-stylus assembly, where maximum resolution is more of an aesthetic concern.

I thought, Right, enough testing – let’s take some real stylus photos! But I hit another roadblock almost as quickly as that thought ended. When taking photos from above, everything is fine – working distance isn’t an issue. But the moment I tried to rotate the cartridge 45° and find focus – disaster struck! There’s just no way to make it work with 0.17mm working distance due to interference from the objective body. Sigh…

This project is really starting to take its toll on me. Not the “let’s forget everything and go off-grid” kind of toll, but close to it. Even disassembling the objective’s front guard doesn’t help. Sure, some “naked-body” cartridges like Lyra’s will work, but that’s not good enough. So I need a real 20X Long Working Distance objective and at this point I wasn’t in the mood to buy any more of that Chinese garbage.

After spending way too much time reading macro-photo forums, I narrowed my choices down to two options. Mitutoyo seems to be the No. 1 weapon of choice for most users, but to my eyes, the Olympus LMPLFLN (now that’s an abbreviation!) is equally impressive in terms of image quality – at least based on the samples I found. Mitu has much longer Working Distance (WD) at 20mm vs 12mm for Olympus, but even 10mm is enough for me.

When it comes to WD – it’s primarily a matter of cost and not physical constrains. Achieving a larger WD without compromising resolution requires proportionally larger lens diameters to maintain the angular width of the entrance cone, and these larger lenses demand stricter control of aberrations, driving up their price significantly.

In the end, it all boiled down to which one I could get a better deal on. And the winner is… (drum roll!)

It’s undoubtedly a significant investment, but as the saying goes, you’ve got to break an egg to make an omelet. For the price of this objective, I could probably eat omelets for the rest of my life – so let’s not delay any further and see if it’s as good as it is expensive.

The working distance is an absolute treat! It easily accommodates even the most massive cartridges, with plenty of room for maneuvering. Safety has also improved significantly – since this CNC table can exert up to 10 kg of force when stepping, the chances of turning someone’s cartridge into tofu are now much lower. But does it resolve?

Testing of the money bag

Oh yes it does! Suddenly, I had no regrets about my purchase. Don’t mind the scratch in the top-right corner – that’s from my ill-fated attempt to clean the target sensor with a cloth (definitely a bad idea).

Olympus LMPLFLN 20X 0.4NA target 3.75μm pixels

Individual pixels are sharp and clearly visible, and even the spaces between them are discernible. That’s detail on the order of 0.3 μm or 300 nm – crazy!

Pixel Perfect Example

Olympus LMPLFLN 20X 0.4NA Corners and Center 100% Crop

5X 0.15NA LMP DIC Semi APO Field Curvature

There is absolutely no loss in resolution in the corners, indicating that aberrations are extremely well-controlled for a 21 mm FN. Next, we need to test how this objective performs on a full-frame sensor, as it’s spec’ed at 26.5 mm. The field is flat, and the wrinkles you see are a result of stacking artifacts, not the lens.

Olympus LMPLFLN 20X 0.4NA on a Full-Frame sensor

No problems covering FF sensor, but if we are pixel-peeping here, then at the corners there is some loss of sharpness. Still, it exceeds it’s specified 26.5mm.

Olympus LMPLFLN 20X 0.4NA vs. AmScope 20X 0.4NA center crops

Olympus LMPLFLN 20X 0.4NA with different sensors

Above, we can see a clear illustration of the cost of good sensor coverage and long working distance. If we compare only the central crops of each image without considering the context, the difference between the AmScope and Olympus objectives doesn’t seem significant. However, the price difference is 60 times! That’s the premium you pay for larger glass with superior corrections.

Also, please bear in mind that this might just be a lucky sample of my exact AmScope lens, and your results may vary. You simply can’t expect consistent quality at this price point.

Next, we can test how well our calculator holds up. For a 0.4NA lens, the predicted sensor pixel limit is ~8.3 μm, corresponding to 2 pixels per Rayleigh distance of 16.56 μm. Conveniently, this matches the exact pixel size of my old Canon 5D MkI. However, as you can see, with only 2 pixels per sample, we can no longer discern the spaces between target pixels. This is why it’s a theoretical minimum, and we should aim for at least 2.5–3 pixels per sample. Going beyond that provides no visible improvement in image quality.

Olympus LMPLFLN 20X 0.4NA with different sensor to tube lens distance (Raynox normal (no text) and reverse (text))

Above is a one-in-two gif animation comparison: the same Olympus lens on a micro 4/3 sensor, tested with different sensor to tube lens distance and switching between a normal and reversed Raynox lens orientation. There isn’t a huge difference between the normal and reversed orientations of our tube lens, but the reversed setup looks slightly sharper to my eyes, so it stays that way.

A more interesting result is why images at a shorter-than-optimal focal distance appear sharper. The Raynox lens has a focal distance of approximately 208 mm – I verified this by adjusting the sensor-to-lens distance until objects at infinity (like trees outside my window) came into focus. However, to my eyes, the crops taken at 180–190 mm look significantly better. Thinking there might be a mistake, I repeated the experiment with a different target.

Olympus LMPLFLN 20X 0.4NA with different sensor to Raynox tube lens distance at 208mm (no text) vs. others (text)

Above are center crops of a pixel target. The first frame shows images taken at varying Raynox-to-sensor distances, while the second frame represents the perfect focus at 208 mm for all cells. This setup allows us to compare individual cells without visual distractions. And honestly, I still find 188 mm slightly sharper, but the results are much closer now.

The takeaway here is that there’s some tolerance for the sensor-to-tube-lens distance, and small deviations won’t significantly impact image quality.

Let’s cheat a little bit

We concluded Part I with two key insights into how the wavelength of our illumination influences visible results. First, decreasing the wavelength improves the optical resolution of the objective, but we loose the DOF as demonstrated in the calculator above.  Second, and not depicted there, is the effective elimination of chromatic aberrations. With monochromatic illumination (a single wavelength), all light rays converge at the same focal points, ensuring sharper image.

The easiest way to produce such illumination is by using LEDs. For the experiment above, I used an RGB strip, allowing us to observe the anticipated effects at the beginning, middle, and end of the visible spectrum.

Again, I’m just eyeballing this, but on the single frames above, you can clearly see how depth-of-field is getting shallower with decreasing wavelength of our illumination.

Well, looking at the animation above, it seems that physics does work! What a revelation. Wait, but does that also mean the Earth isn’t flat? Guess I’ll have to cancel that edge-of-the-world sightseeing trip…

The results are truly impressive, leaving no room for interpretation – resolution significantly improves with shorter blue light. Of course, as always in physics, there’s no free lunch. The trade-off is a monochromatic image with shorter DOF, which might be perfectly fine for most technical photography but is far less appealing for biology or general macro photography.

Epilogue

This project was incredibly tough and interesting at the same time. I was on the verge of quitting, unwilling to spend any more time, money, or effort on it. But somehow, I persevered, and up until now, I’ve taken plenty of photos of pixels and memory cells. Finally, it’s time to capture some real images!

AT95E retiped stylus (adjusted for sharpness)

Here’s what I had on hand: my lab workhorse, the AT-VM95E, which I retipped after discovering it was glued at an incorrect angle. This was probably my third practice subject for retipping, so it’s not exactly my proudest work. But wow – the level of detail is staggering! This stylus has maybe 1-2 hours of use, so no contact surfaces have formed yet.

AT95E retiped stylus (RAW after stacking)

AT95E retiped stylus 3D visualisation

And above is the unprocessed result – straight out of the image stacker. It took a whopping 283 images, consuming 5.2 GB in total, with a step size of 2.4 μm, to produce this single image.

Next, I grabbed something more challenging – my DL-103. It has a very finely polished conical stylus that is totally translucent, yet the surface is very smooth and without any texture. So, in terms of photographing and stacking difficulty – this is a nightmare level.

No problem whatsoever – it stacked perfectly without any need for retouching. And believe me this was not the case with other lenses. Lens flaring and reflections within the internal optical system produced lot’s of image artifacts that covered the focus and produced stacking artifacts.

Side-shots like these inevitably require a bit of retouching. This happens because light bends or diffracts around the edges of the stylus, and at such a shallow depth of field, rays from behind the stylus can get included in the stack. As a result, there’s some ghosting around the edges, and naturally, the stacking software isn’t too thrilled with it.

But here it is – the long-awaited contact surface of the stylus in all it’s glory. This particular cartridge has about 100 hours on it, so the surfaces are well-defined by now. Even without exact measurements, it’s clear there’s still plenty of life left.

And that’s when it hit me – why not turn this into a service? I presume not many people would want to go through the same process I did, spending around $3k just to assess the wear on their stylus. And that’s my dear readers is how the optical stylus inspection service was born. Thanks for sticking for so long!