Hacker News new | past | comments | ask | show | jobs | submit login
I Just Poured Water on my Scanner (cpushack.com)
230 points by protomyth on Mar 30, 2015 | hide | past | web | favorite | 64 comments

I remember having to use special oils to use some high magnification microscope lenses in a process known as 'oil immersion microscopy'. I didn't read much about how it worked before but I assume it is the same process. There was no other way for me to get the magnification I needed to count the reletive number heterocyst cells in strands of cyanogenbacteria for my biology coursework.

I would suggest to the author that they buy some microscopy oils as they will have a higher refractive index to water.


If the scratches are on the glass it could work even better to use a liquid with the same refractive index as the glass. That should make the scratches go away completely.

One option is glycerin as can be seen in this video: https://www.youtube.com/watch?v=pNWCB_GoQA4 (maybe it only works with pyrex glass ?)

Another option is to mix to fluids together, one with a higher index of refraction and one with lower. That way you can tune it to your particular scanner (although I would think most are the same).

How you see if your mix has the same refractive index as the glass I haven't figured out yet.

> How you see if your mix has the same refractive index as the glass I haven't figured out yet.

You can do this by direct observation but it is not going to be 100% accurate:

You'd need some kind of etched grating for reference.

Drop the glass+grating into the mix, check for changes to the grid lines where you know the glass resides. If you can spot it then you still have a difference.

You could also simply measure the refractive indices of both. (For most microscopy glass you can find out the exact index of refraction from the manufacturer so you really only need to measure one.)

If I could put the glass into the liquid it would be fairly easy. But, isn't the glass stuck to the top of a flatbed scanner in this case? Makes it a bit hard to put it into a liquid without breaking the scanner.

In that case you might want to try to do some absolute measurements on the dimensions of the grid.

So you get (scannerglass + fluid + gridsample), then measure the grid point distance (by taking as large a distance as you can across the grid) and compare to what you know it should be. The closer the grid measurements are to the actual dimensions the closer the index of refraction of the fluid will match the glass.

I think the scratches are in exactly the same place each time, which suggests they're on the chip, unless the chip was repositioned very carefully.

The vertical portions of the scratches in the dry pic are accentuated, while the horizontal portions of the same scrates are highlighted in the wet pic. The only thing I can think that would cause that is the chip being rotated 90 degrees in relation to the scanner light.

Similarly, you can scrub some scratched up glasses with toothpaste to fill in the air gaps. Good as new (at least until it rains).

Toothpaste has some abrasives in it, I wouldn't use it on the glasses.

Toothpaste is mostly soap, so try that instead. Or mineral oil.

That doesn't work the same way, that polishes the edges of the scratches so they're not as distinct.

Is this related to the old lifehack about smearing toothpaste on your watch's sapphire crystal screen?

You can use toothpase to polish the bottom of a soda / beer can, which you can then use to focus sunlight to start a fire. I've done it.

There are better ways to start a fire.

That's using the abrasive silica in the toothpaste (aka "sand") as a very mild polishing agent.

Silica is not even close to hard enough to polish sapphire, and most toothpaste, instead of containing silica, contains chalk, which is not even close to as hard as silica. The problem with brushing your teeth with silica is that silica is not a very mild polishing agent; it's a pretty aggressive one, and it will wear away your tooth enamel pretty quickly. If those are your permanent teeth, that enamel ain’t coming back.

Silica and maybe even chalk will work fine to polish glass, though.

I have a correction. I went out and bought a tube of Colgate ("Triple Acción! Extra Blancura!") and it does indeed contain silica as an abrasive. (And Colgate is hardly some weird off-brand toothpaste.). The difference is that it’s hydrated silica, which is amorphous rather than crystalline, and is somewhat softer than normal silica. (And, as an extra special bonus, not carcinogenic like normal silica!) This is the same stuff known as "silica gel" that comes in little pouches with your electronics labeled "DESICCANT — DO NOT EAT".

Amorphous hydrated silica is still hard enough to polish glass and abrade your tooth enamel. Chalk is not, although I'm not sure it's hard enough to abrade away your tartar deposits.

Clickbait title, of course.

For film scanning, this is known as a wet-gate transfer.[1] It's standard procedure for digitizing old cine film. The working fluid used has the same index of refraction as the film backing, so scratches disappear.

The guy who's photographing ICs just needs a toy-grade digital microscope.[2]

[1] http://www.arri.com/archive_technologies/archive_challenges_... [2] http://www.amazon.com/ThinkGeek-P510002-Digital-Blue-Microsc...

As the article points out, this is not a new technique. Increasing your resolving power by immersion of the target object in a liquid medium is actually used in ICs manufacturing.

This is because resolving power is directly proportional to numerical aperture, which is in turn proportional to the refractive index of the medium between the imaging device and the target object.

As a result, it works both ways: you can use it to increase resolution while either sending or receiving beams of light. For instance, the same technique is also used in light microscopy.

In the end, this is a clever and straightforward hack.

You are talking about immersion lithography. It replaces the air gap from the final lens of the lithography system with water which has greater IOR. AMD has been using it for a while, probably before the 45 nm node. Strangely enough Intel was not using it until the 32 nm node. It's the same hack with oil immersion lenses, basically, for the optical microscopes.

If you didn't watch this video before... https://youtu.be/NGFhc8R_uO4?t=15m50s

This is a great presentation on the evolution of transistor manufacturing processes. I didn't know how a transistor worked before watching this video, yet was engaged by the technical narrative of the semiconductor industry overcoming one physical constraint after the other.

Pretty sure it predates even that. Some of the earliest lenses, especially before the ability to grind glass accurately, were water based.

> Pliny the Elder (23-79AD) describes how glass balls filled with water could set clothes on fire when placed in line with the sun. Both Pliny and Seneca the Younger (3 BC-65 AD) also describe the magnifying effect of a glass globe filled with water.

> Pliny the Elder (23-79AD) describes how glass balls filled with water could set clothes on fire when placed in line with the sun.

Really? Sounds like a job for Adam and Jamie.

Fluid mounting things to scan them is something that photographers have been doing for a while. The higher-end Epson flatbed scanners (which is really a low-end film scanner) have fluid mount options [0] while fluid mounting for drum scans [1] (high-end film scanners) is the norm unless you get a flexframe model from Hasselblad which doesn't use glass (so no need for a fluid). Plus wet mounting your scans helps with the Newton rings issue.

[0] http://www.photo-i.co.uk/Reviews/interactive/Epson%20V750/pa... [1] https://www.youtube.com/watch?v=aCHS5yiF4Mc

I'm not convinced that the flextight (talking about what used to be Imacon, right?) options are a complete substitute for wet mounting, because wet mounting will reduce the appearance of scratches on the film itself, but eliminating the glass doesn't fix that particular problem.

I'm sorry, I should have posted more clearly. Yes, you will reduce film scratches (and Newton rings and glass scratches) if you wet mount. You will only reduce Newton rings and glass scratches if you use an Imacon/Hasselblad since there is no glass to scratch or produce Newton rings. So, yes, I do think that a wet mounted drum scan will still get you the best quality.

This is super common amongst film photographers - it's a wet scan. Works wonders. Some of the fanciest "drum" scanners out there work this way exclusively.

"The markings are laser etched directly onto the surface of the silicon die."

That's the package, not the silicon die.

Assuming someone downvoted because they didn't understand or felt I didn't explain enough... That chip in the picture (the intel QG80331M500, the chip in question) is a flip chip package; instead of using wire bonds to connect to pads on the die, it has solder balls like a BGA package. The chip is soldered down to a carrier PCB and an underfill glue/sealant is applied around it, so the only thing exposed is the bare back of the die. There is no plastic package to laser-engrave, the engraving is done right on the silicon.

I was hoping to see scans of the actual silicon, but I guess that requires more exotic equipment. It seems to me that imaging the chips would be easier and better using a DSLR and a good macro lens.

Zeptobars has some nice shots of silicon dies. As a random example http://zeptobars.ru/en/read/Nordic-NRF24L01P-SI24R1-real-fak...

I estimate that those images are taken with equivalent of about 60000 DPI. Even with that sort of resolution you can't really resolve the details in the smaller features.

In comparison typical macro lenses have 1:1 magnification, so e.g. D800 DSLR with 36MP sensor would have theoretical maximum of roughly 5000 DPI, and I would expect the reality to be quite a bit harsher than that.

I was thinking the same. They could set it up with a tripod and couple of lights like this [0].

[0] http://globalflyfisher.com/gallery/digitizing/setup.jpg

You'd need to strip back the casing to see the actual circuits.

This article was on the home page for a while with the title "Immersion scanning", a pretty dull one. Thanks to whoever changed it to "I just poured water on my scanner"! Makes a much more intriguing and appropriate title.

Now if only the headline had been "This person just poured water on their scanner and you won't BELIEVE what happened next. Find out why scanner manufacturers HATE this one weird technique!"

But yeah, I do agree it's a catchier headline than "Immersion Scanning".

I disagree. It now sounds like 100 other clickbait titles - "The 10 best ways to lose weight!", "The one true way to scan that the manufacturer doesn't want you to know!"

I assumed it was an intentional tip of the hat to the "I just put my laptop in the oven" hack.

I don't understand. Why not just use a decent resolution digital camera to take a picture of the chips?

I don't see how that will help.

Wouldn't it be harder to try to get a good picture of the chip in such an angle that you could read the etching?

I assume this angle will most likely be "from the top", such that it basically turns into a scanner, and then it's easier to do it with an actual scanner.

But with the camera you have control over the lighting, and can move the lighting such that the etching has more contrast.

I'll just assume that they tried the camera and the scanner works better.

Edit: So maybe you do have better control, but what if the etchings are so small that there's no significant contrast change when viewing from different angles?

I'm just basing my speculation on the fact that these guys probably have tried a lot of things and this is most likely the best way they've found.

It would be interesting to hear from them what were the results of other setups.

A camera on a tripod and a light source would give infinite flexibility as to angle and illumination. Once optimized, the setup could be left as such.

Because taking accurate reproductions of small flat objects boils down to three things: zero distortion, even illumination, flat focus plane. Even if you can find a well-corrected macro lens, building the physical setup is a major PITA and you'll still have ugly shadows all over. Scanners have been getting these things right since forever and for next-to-nothing; a 15 year old $300 film scanner easily beats any DSLR film scanning attachment you can find, and not only quality-wise - but especially in terms of ease of operation. If you don't believe me, try "scanning" a document using your camera.

Besides being a very interesting hack, this made me laugh:

"(and contained as to not get IN the scanner, as that does NOT help the process)"

Not sure why but this seemed really funny in the technical context.

That's a really cool technique! Though it looks like his scanner glass is scratched to hell and back (I wish they'd make the glass more scratch resistant.) My exact scanner type, too. I would say it's definitely fast at 300dpi; but it takes well over 90 seconds for 600dpi (full scan, chip scan would be faster of course), never dared try 1200dpi.

Also too bad I can't use use this trick when scanning cardboard boxes and manuals.

> Though it looks like his scanner glass is scratched to hell and back

I don't think the scratches are on the scanner, since they're in the same location on both images and the CPU had to be moved to be placed in water.

I tried to compare the two to rule that possibility out, but it's possible I'm mistaken. That would be good though, I'd hate to think they were scanning things with all that scratch damage getting on every picture.

The scratches are dominantly vertical in dry scan and horizontal in wet scan. This in my mind would indicate that scratches are on the scanner glass and the chip was rotated between the scans.

Is that his scanner glass that's scratched, or the chip?

The scratches don't seem to match between the scans, so it seems that it is indeed his scanner glass that is scratched.

I just did an overlay of the two images and most of the visible scratches in the "before" version have matches in the "after" version. There are other scratches that are visible in the "after" image that often obscure the "before" scratches, so it can be difficult to tell. But maybe there are scratches on both the chip package and the glass?

I don't see resolution improving all that much, only contrast, as a quick and dirty photoshopping can show:


The contrast of the engravings on just the chip is what was improved the most, not the resolution itself.

And speaking only for myself, there are still a number of markings in your photoshopped image which I am unable to read, which can be clearly read on the immersed image.

That's a solid improvement, but I bet they could get better results with some other machine vision tricks. Diffuse on-axis lighting [1] would probably be a good approach for this application; it's good at making subtle markings on flat surfaces stand out.

[1] http://www.microscan.com/en-us/products/NERLITE-Machine-Visi...

Some holographers use Xylene as an index-matching fluid. Other suggestions here including decalin and sugar water:


I'd be curious whether you'd be able to reduce the noise (not the same as lensing, obviously) if you used tape like this: https://www.youtube.com/watch?v=nRoL2q-tU-Q

I've never had to scan things like your process so I have no clue; just a thought.

This sounds a lot like how an oil is used to look at slides in medical labs.

For example, it may be standard procedure to smear a sample along the slide, place it on a stainer, retrieve it later, put some oil on it, and the lens of the actual microscope will come in contact with the oil.

Fixing it on the top surface is good, but there's still another air-gap between the bottom of the scanner glass and the head. A bit harder to fill that gap with water though.

Dont chips get destroyed when powerful light (like the one in the scanner) hits them? Or is the point of scanning chips just to get their specs on file and not to preserve them?

There are some kinds of chips that can be reset by certain kinds of light (http://en.wikipedia.org/wiki/EPROM). Apart from that, packaged ICs aren't generally damaged by light that isn't at a laser-like intensity.

I wonder -- does scratch removal software help in such cases?

It seems to do wonders with certain cases of slide scans, so I wonder if it helps with these kinds of wet-mount scans as well.

Wet mount scanning can do a world of good. Some examples: http://www.scanscience.com

Pretty slick trick I must say...

aaaaaand it's gone.

It's just to get rid of the air gap between the glass and the chip. The intent isn't to break the scanner.

Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact