Decapping and Chip-Documentation - Howto

[Noopy]

Stripping everything is not too hard. And if it should be as clean as possible just keep it longer in HF (Armour Etch). The silicon etch rate of HF is very low.

...that looks like a darkfield picture...
I should try darkfield illumination next time I strip the metal layer. It could be that darkfield is better for completely stripped parts where you usually don´t have any more resonances / colors left.

[iMo]

So he etched off all the metal layers there, not only the top mesh..
My first thought was he etched off only the top mesh..
..got it..

[Noopy]

Well there is no real high resolution picture but to me it looks like he stripped everything like I did that a few times.

As far as I know there is nobody in the hobbyist area who can do a clean delayering from layer to layer.

Ken Shirriff has some nice pictures of different layers of a Pentium CPU (https://www.righto.com/2025/01/pentium-reverse-engineering-bicmos.html). I talked to him. As everybody he is not able to strip one layer after the other over the entire die. He "just" grinded one area. That is possible, sometimes you are lucky and you get the layer you want in this area and sometimes you are less lucky and you get a mixture (of course it also depends on your experience).

[iMo]

..ion etcher.. 

[Noopy]

...would be nice...  ...with some chemicals and the right equipment for them. 

[Noopy]

For reflected light microscopes you usually use special objectives (right). Olympus marks these objectives with an M (“metallurgical”). The more common “biological” objectives (left) are designed to be used with a coverglass. Metallurgical lenses, on the other hand, give you the best pictures without a coverglass. With small numerical apertures, but the influence of the coverglass is negligible. However, there is another important difference between the two lens types. The lenses in metallurgical objectives are usually additionally coated so that they reflect less light. This is very important for coaxial illumination, where the illumination of the object follows the same path as the light reflected from the object.

The two lenses above are broadly comparable. They differ only minimally in terms of their magnification and numerical aperture. Both lenses are the revised U variants.




For comparison, the same black plastic surface was photographed with both lenses. The white balance was fixed, only a small adjustment of the exposure time was made. The images have not been post-processed. The biological lens clearly shows a bright spot in the inner area. The shorter wavelengths are more noticeable. The metallurgical objective, on the other hand, shows no weaknesses.




The weaknesses are not overly noticeable on a detailed picture.




The difference becomes somewhat clearer if the color information of the images is discarded and the contrast is greatly increased.




Due to the different numerical aperture, it is difficult to compare the resolution of the objectives. However, the image quality of the metallurgical objective appears better than one would expect due to the slightly higher numerical aperture.


https://www.richis-lab.de/Howto_Microscope_MObjectives.htm

 

[RoGeorge]

This video popped into subscribed videos list, a microscopy technique similar with focus stacking, but for diffraction.  It allows an increase of optical resolution, and also to recover not only the amplitude, but also the phase of light, though the result is a monochrome picture.  Not sure if this will work for through-lens illumination, but adding the link here just for the docs.

Dramatically improve microscope resolution with an LED array and Fourier Ptychography
Applied Science
https://youtube.com/watch?v=9KJLWwbs_cQ

[Noopy]

Some news in the optical field:




This overview from an Olympus brochure for the BX53M/BXFM microscope family shows a selection of the options that can be used for observing objects. Trinocular heads that can operate two eyepieces and a camera are common. With some variants, the angle at which the eyepieces are positioned can be adjusted for more ergonomic operation. There is also a special variant for infrared microscopy. If the eyepieces are not required or if space is limited, slim cylinders containing only a tubelens can be used. Both trinocular heads and tubelenses are available in versions with an image circle of 22mm and versions with an image circle of 26,5mm. A single digit at the end of the designation indicates the revision.

In principle, all trinocular heads and tube lenses can be used with all microscopes in the BX family. Olympus does not explicitly exclude any variants. However, it has been found that the image quality varies considerably.




The U-TR30-2 head is widely used. In some cases, its predecessor, the U-TR30, which is painted black, can still be found. Although the U-TR30-2 is only specified for an image circle of 22mm, it can be used with an APS-C sensor. The outer areas of the image are not perfectly reproduced, but it can be used to create panoramas. Focus stacking further improves the quality.




Unfortunately, the options specified for a 26,5mm image circle are very expensive. Surprisingly, the more complex U-SWETTR-2 model is sometimes cheaper to find. Unlike the U-SWTR, the U-SWETTR allows you to adjust how steeply the eyepieces point upwards.




The view through a 26,5mm eyepiece is impressive.






The upper of these two images was taken with the U-TR30-2. The lower image was taken with the U-SWETTR-2. In both cases, the UIS2 objective MPLAN FL N 20x was used. These are unprocessed single images without focus stacking. The depth of field of the MPLAN FL N 20x is actually too shallow to take pictures without focus stacking. This explains the slight blurring on one side. Sharp areas were selected for further examination.

Both images are also available in higher resolution:
https://www.richis-lab.de/images/howto/17x04XL.jpg (10MB)
https://www.richis-lab.de/images/howto/17x05XL.jpg (10MB)




The image above is taken from one corner of the image captured with the U-TR30-2. The image quality is poorest in the corners. A slight loss of sharpness can be seen. There is also some chromatic aberration. However, as described, this image quality is perfectly acceptable for working with. Panoramas can also be created. Only occasionally do the interfaces between the individual images produce artifacts.




Surprisingly, the image quality of the U-SWETTR-2 is hardly any better. It is assumed that the additional degree of freedom of the eyepieces comes at the expense of image quality. Perhaps this is why there is now a fifth revision of the U-SWETTR.




The U-SWATLU tubelens replaces the trinocular head and is a relatively inexpensive way to make optimal use of the 26,5mm image circle.




Without the trinocular head, it is no longer possible to view the image directly. Furthermore, it is no longer possible to remove an eyepiece in order to adjust the aperture diaphragm of the Köhler illumination. You have to note down the position of the aperture lever for the different objectives so that you can then make the adjustment blindly. It is advisable to use a caliper for this.




In addition to the larger image circle, the U-SWATLU tubelens has another advantage. It allows more light in the near-infrared range to pass through. This is a major advantage when performing infrared microscopy.






The UIS2 objective MPLAN FL N 10x was used to take these images. The upper image was taken with the U-TR30-2, and the lower image with the U-SWATLU.

Both images are also available in higher resolution:
https://www.richis-lab.de/images/howto/17x10XL.jpg (12MB)
https://www.richis-lab.de/images/howto/17x11XL.jpg (11MB)




As with the 20x lens, a certain amount of blurring can be seen with the U-TR30-2. Chromatic aberration is also evident again.




The U-SWATLU tubelens significantly improves image quality. The image is sharper and almost free of chromatic aberration. The last remaining color fringes are only visible at very high zoom factors.






These two test images were taken with the UIS objective MPLAN FL 10x. The upper image was taken with the U-TR30-2 head, while the lower image was taken with the U-SWATLU tube lens.

Both images are also available in higher resolution:
https://www.richis-lab.de/images/howto/17x14XL.jpg (12MB)
https://www.richis-lab.de/images/howto/17x15XL.jpg (12MB)




The UIS and UIS2 objectives are comparable. With the U-TR30-2 head, blurring and chromatic aberration are equally apparent.




The U-SWATLU tubelens also significantly improves image quality in terms of sharpness and chromatic aberration. In contrast to the UIS2 objective, however, the UIS objective retains slightly more chromatic aberration.


https://www.richis-lab.de/Howto_Microscope_Tubelens.htm

 

[magic]

After some success with decapping chips in colophony last year I have been doing various experiments with "wet" methods, i.e. boiling chips in liquids.

I tried molten paraffin wax at 380°C, but that only removed a bit of the surface. Also mineral oil and molten vaseline, I kept them at ~400°C for 30~60 minutes but in those cases there wasn't even surface damage, only some cracking and bulging which looked similar to chips heated to very high temperature in dry air. So it definitely looks like the colophony process involves some chemistry, not just temperature.

On the other hand, I found that ordinary cooking oils (tried rapeseed and olive) can work. Which made me very happy, because oil is a room temperature liquid, so easier to use and easier to clean up. Unfortunately, I found that my ST TL062 chips are resistant to it. I will see if I can find some way around it, if not then it's back to colophony again.


Here's the oil procedure, which so far worked on several small (SO8/SO14) chips from China and one Microchip IC. I haven't tried larger and thicker packages yet.

I pour 1~2ml of vegetable oil into a 16x160mm borosilicate test tube and drop the chip inside. Too much oil seems to work worse, but I don't have enough data to really know. I also tried a shorter 14x80mm test tube and it worked, but I had to be careful with heating to avoid losing vapor to the atmosphere. I heat it on a spirit burner with enough flame to cover the whole rounded bottom of the test tube but not much more than that. It seems that in this process heating intensity doesn't make a huge difference, but this too is based on very limited data.

After 2~3 minutes, the whole test tube fills with dense fog and temperature peaks at 420°C. Then the fog clears up, plenty of foam appears on the surface and temperature gradually drops to 390°C over another 2~3 minutes. Foam seems to be caused by presence of solid objects - I tried heating oil without anything and it didn't foam much, then I dropped a piece of sharp glass inside and it started to foam. Foaming seems to decrease somewhat when the chip is "done", but I don't have enough experience yet to say if this is a reliable indicator.

After 5~10 minutes the chip should already be visibly smaller, soft and possible to take apart by hand. But I don't want to take chips apart by hand, so I keep boiling. At 10~15 minutes the oil becomes dark and opaque - enough to block light from behind the test tube. This is an indicator of progress, simply boiling oil without any chip inside doesn't show this effect even after 30 minutes.

After 20 minutes the chip should be completely dissolved. I wait for the oil to cool down and stir a little to keep the die and metal pieces from sticking to internal walls. Then dilute the oil with equal volume of petroleum ether, give it a good shake and pour it out. Find the die, pick it up with tweezers, wash it, done.


With oil, it seems that allowing vapor to escape is detrimental - it only reduces the amount of liquid without significantly increasing boiling point. Though I need to try it again to be really sure [I tried next day and it helped]. With colophony, I found that crazy heating and allowing some fumes to escape may increase the boiling point even to 450°C. Perhaps that's the trick, but my best result so far is a very soft but still not completely dissolved chip after 5 minutes.

[RoGeorge]

tube fills with dense fog and temperature peaks at 420

Sounds like a bong! 

[Noopy]

Very interesting! I would have guessed that mineral oil does more than cooking oil. Thanks for sharing! 

Someday I will give DMSO a second try.

It seems if we leave the acids aside all decapping methods which dissolve the epoxy can´t dissolve every flavour.

Sounds like a bong! 

 

[magic]

Sounds like a bong! 

But smell like a mix of burnt oil and burnt plastic for some reason
And probably no less carcinogenic than tobacco.

It seems if we leave the acids aside all decapping methods which dissolve the epoxy can´t dissolve every flavour.

Last year I tried colophony on several SOIC/TO92 ICs and transistors from various vendors and it worked on all of them. But in my process, colophony turns into a sticky and hard to clean tar, so I looked for alternatives. I think I will check if this problem can be dealt with by diluting with solvents while it's still liquid and I will continue experiments with smaller test tube and more fire to see if it helps to distill off low-boiling components of the mixture and only leave the heavy stuff at high temperature.

The goal is a setup similar to that shown by Sacodepatatas, which supposedly takes 4 minutes to completely dissolve all plastic (if I understood correctly).

[Noopy]

Did you really reach ~400°C? On a second thought that sounds quite high for a natural oil...

[magic]

I'm pretty sure it no longer is natural oil

It seems that regardless of initial oil quantity, container dimensions and flame intensity (within reason), oil initially reaches 420°C and then apparently decomposes into some other "stuff" which pulls its boiling point (and therefore mixture temperature) down to 390°C within a few minutes.

In a long test tube, all the low boiling components condense and return, keeping temperature down. In fact, after long time it stabilizes around 380°C. Right a moment ago I reached one hour boiling 4ml of oil (with no chip) in the 16x160mm tube on slightly stronger flame than yesterday. About 70% still remains, thermal behavior was practically the same as before and while it became dark brown, it's still transparent enough to see bubbles. For the last 30 minutes it oscillates between 370°C and 380°C - rises when the low boiling components evaporate, drops when they condense and drip down. By increasing flame some more I managed to push it up to 380~390°C.

Things are different in a shorter tube, because the low boiling vapors escape. Today I managed to decap that TL062, using 4ml of oil (to compensate for expected loss) in 14x80mm tube on similar "slightly stronger" flame. This time temperature only dropped to 400°C and began to rise again: 415°C at 10 minutes, 430°C at 15 minutes, then stabilized at 437°C. Oil was disappearing quickly and I had to stop after 25 minutes. The chip was almost completely dissolved, and the remaining chunk was soft.

But the bonding pads got dissolved as well

So I don't know if this is very much better than, say, sulphuric acid. I mean, fair enough, it's easier to get and safer at room temperature. And on chips which work without crazy temperature it doesn't seem to damage the pads, but not all chips work like that.

[Noopy]

We need a chemist! 

Another freaky idea:
High temperatures speed up a lot of reactions. Would be interesting if there are more common substances which could dissolve epoxy if you heat them more than normaly possible.
I´m thinking about things like acetone in a pressure cooker. Perhaps aceton would dissolve epoxy at 200°C.
Thinking about such things scares the hell out of me. Definetly not safe to do. 

[Noopy]

Do you have some pictures for us?

[magic]

It's no rocket science, here's the "low temperature" (380°C) oil process which works on some chips but not on others.

16x160mm borosilicate test tube with about 2ml of oil, mounted with a wire on a "third hand". Nicely adjustable.
The burner is DIY: small bottle, capacitor can and a cosmetic cotton pad cut into spiral - these make excellent wicks.
I fuel it with 92% denaturated spirit and tweak the wick so that flame roughly covers bottom of the test tube.
The burner stands in a wide metal can so that potential spills of burning liquids won't spread all over the place.

Not shown: fume extraction. Not necessary for the process to work, but recommended

Thermometer is optional at this point because the process seems repeatable. I use fiberglass insulated thermocouple in a glass drinking straw melted and sealed at one end. Look up how to make glass ampoules from eyedropper pipettes if unsure. Soda lime glass is good enough and much easier to melt than borosilicate.

After 20 minutes turn it off, cool down, dilute with solvent and pour into a small bowl. Diluted oil is transparent enough that finding the die is relatively easy (unless you boil it for two hours). Here you can see it near the center with some wires still attached.

What's the deal with boiling for two hours? Yesterday I boiled oil with no chip for one hour and then dropped a chip in it to see what happens. Turns out that oil seems to degrade and lose effectiveness with time, and it took much longer than usual to decap that chip and oil was very dark and opaque at the end. Solution: use fresh oil for each run and don't boil too long.

[Noopy]

Nice! 
Do you also have a picture of a die decapped like this?
If you want to do bigger packages I assume it would be necessary to cut the epoxy around the die?

[magic]

With bigger packages cutting useless fragments away may be necessary, because this "low temperature" process is rather slow. Should be easy with DIPs, less so with TO220 etc.

I currently don't have die images and haven't posted anything new this whole year at all. In my infinite wisdom I have dismantled my epi-illuminator to build a better one and you can guess what happened I will see if I can put it back together this weekend, thankfully the most important bits are already completed.

[magic]

Without the trinocular head, it is no longer possible to view the image directly. Furthermore, it is no longer possible to remove an eyepiece in order to adjust the aperture diaphragm of the Köhler illumination. You have to note down the position of the aperture lever for the different objectives so that you can then make the adjustment blindly.

Actually, you can see the aperture if you have a substage condenser - just remove the specimen and place a thin paper disc on top of the condenser's iris diaphragm. I suppose any small wide angle camera focused at infinity could work instead of a condenser, at least up to some NA, but condensers are practically made for this.

The microscope should be telecentric to ensure that reflected light returns back to the objective and doesn't go sideways, so the objective projects the aperture towards infinity. Similarly, iris is projected by the condenser to infinity, so image from infinity appears in the iris.

[Noopy]

I have the BX51M which just has an epi illumination, no trans illumination.

I have measured the position of the diaphragm knob for all of my objectives. That´s ok for me and actually is faster than checking the appearance of the aperture. 

[magic]

Ah, never mind then. Why does the stage have a hole in it? I thought it's one of those universal ones, with top and bottom light.

I remember seeing a fairly inexpensive Bausch&Lomb Balplan scope which had it all - brightfield, darkfield and transmitted. Almost got it, but I gave up because those things use absolutely proprietary optics and other objectives can't even produce an image there.

[Noopy]

I got it like this. I don´t know whether they change the substage before or if this table is the normal one for a BX51M. There is a hole in the substage but the substage holder is a massive part withouth a hole.

Unfortunately good microscope optics cost still a lot of money... 

[mawyatt]

Unfortunately good microscope optics cost still a lot of money... 

Agree!!

BTW we are completely closing down our photographic optics lab and selling off all the lenses and fixtures, even our prized Mitutoyos (5X, 10X, 20X all Plan Apos and two Plan 50X NUV)!!

Best

[Noopy]

BTW we are completely closing down our photographic optics lab and selling off all the lenses and fixtures, even our prized Mitutoyos (5X, 10X, 20X all Plan Apos and two Plan 50X NUV)!!

Where and how "cheap"?