Hi Dave. Thanks for this collaboration and I love how you are commenting this content :). For the X-Ray machine some input from my side. All the contrast and sharpening is made by software. So without that the image would be very blurry. Therefore at some point you could see noise in the image which disappeared then. At some point you said the part is upside down, and this is very tricky coz its not upside down. The thing is that by X-raying you see only a contrast image of x-radiation which is shot trough the part. I tried it and no matter if you place it upside down or not, the image looks always the same and its very confusing. This is due to the fact that we just get an image if the radiation is able to go through the material. On one side (underneath the round plate in the machine) there is the X-Ray source and the flat square part on the top inside the machine which is moving to the side is the receiver to detect the radiation and creating an image. When you need a a lot of radiation to go trough a material very thin parts are not visible any more coz they are absorbing too less radiation to give enough contrast on the image. Therefore I believe we can not see any traces in the ceramic substrate of the TCM.
There are ways to reveal the traces, but they require a low-noise system and lots of patience. The basic techniques come from astronomy and the imaging of diffuse gas clouds and very fine structures, combined with math borrowed from tomographic reconstruction and X-Ray crystallography. To get good 2D images, first build up an extremely fine 3D image, then take slices of it. Basically, the system has to "stare" at the target at multiple energies from multiple angles, then carefully combine the many signals together. There isn't nearly enough information present in any single view to get such detail during live interaction.
anybody that Dave Jones recommends gets an immediate subscribe from me - going to dive into your collection shortly. :D As to the traces, I think some of them were visible, but just as ghostly X or Y axis straight lines - they seemed to appear when in focus but be transparent otherwise. Also as to the vias, some you could see perhaps the plating material on them, while for others the barrel was visible as well - not sure if that's a contrast thing or those had more/different plating on the vias.
Email from an X-Ray Tomography Specialist: "The reason you can't see the copper traces is because they are planar to you / perpendicular to the beam. Xray detectors have somewhat limited contrast, and in order to get good penetration, you need to run high kV to get good energy (material penetration), and have decent current (flux) in order to have enough signal on your panel. If you have hard enough beam (good penetration) to get through that steel you will absolutely blow through thin pcb traces. When the full thickness of those vias is exposed to the beam, the beam has to pass through the equivalent of that thickness of copper, hence you can see it."
Seeing planar traces isn't "hard" in the physics sense, but it is rare to find an X-Ray system capable of working right at the edge of the physics. First, it can't be done quickly. If using tomography, the voxel size must be no larger than the trace thickness, and multiple passes over a range of X-Ray energies and intensities will need to be fused. However, there are tricks that can be used by an experienced operator in near-live systems (presently ~1 fps with lots of latency). The key is to accentuate localized contrast by rapidly modulating the X-Ray beam energy and intensity, then doing dynamic fusing of the multiple exposures (with no motion between the exposures in a set). In the end, it takes a low-noise system, fine and rapid X-Ray control, a deep/wide imaging pipeline, a ton of math, and lots of patience. The weakest part of any such system typically is the imager, and there are some old-school analog tools and techniques that can provide rapid response with enhanced contrast and low noise.
Makes sense. The part of the x-ray spectrum with low enough energy to be absorbed by the silicon in the quartz would already have been blocked by the crystal's case.
We use the same X-Ray manufacture at my job. I have X-Ray'd PCBs and can see the traces. However, when I X-Ray laminates and packages, they seem to be too thin to detect. What's interesting, you can see 1um bond wires with no problem. This machine saved me in a pinch to verify a solder bridge for a large BGA package once.
4:19 Dave, RUclips stopped sending email notification on 13 Aug 2020 the bastards. The only way to get notified is by push notifications in your browser. www.socialmediatoday.com/news/youtube-will-stop-sending-email-notifications-to-alert-channel-subscribers/583258/
Also, there are probably multiple solid power planes. If you set enough power to the xray source to get through the power planes then you will also happily go through the traces and not see them. You WILL see where multiple tracks overlap, like the annular rings overlap...
X-Ray tubes consist of a highly-focused electron gun aimed at a metal (often tungsten) target. The gun voltage determines the speed, the energy, of the electrons which determine the energy of the X-Ray photons emitted at the target point, ideally one X-Ray photon emitted per electron. The current of the gun then determines the rate of X-Ray photon emission, known as the X-Ray flux. The X-Ray detector (where an array of detectors forms an imager) is sensitive to a wide spectrum of X-Rays. Whatever is placed in the beam between the tube target and the detector will act as an X-Ray filter, much like color films can filter visible light. The X-Ray energy is varied to find peak absorptions, then the flux is varied to adequately stimulate the X-Ray imager. The X-Ray beam has the shape of a cone, so where in the beam you place the material to be imaged determines the zoom effect: Close to the X-Ray tube for high magnification, and close to the detector for 1:1 imaging. However, the closer you get to the X-Ray source, the more important the size of the spot of the electron beam on the target. X-Ray tubes made for extreme magnification require extraordinarily tight focus, and are called "micro-focus" tubes. Unfortunately, the focusing of a high-energy and high-current beam on a tiny spot creates massive amounts of local heating, leading to melting and even vaporization of the X-Ray tube's target material. For this reason, such tubes tend to be water-cooled, and the target material is moved to avoid excessive wear at any single point, all done without any effect to the X-Ray beam itself. All micro-focus X-Ray beams have relatively short lives, even with cooling and target movement, so many are designed to be overhauled on-site. Things get more fun when a multi-axis robot is placed in the beam to move the sample as desired: The robotic precision needed at very high magnifications can be on the level of single-digit microns. I had to learn all this stuff not because I worked on X-Ray sources, or on X-Ray detectors, or on X-Ray robotics. I learned this because I worked on the image processing systems needed to greatly accelerate "dialing-in" the system settings to obtain ideal contrast video, to aid the operator in planning, executing and replaying imaging runs, and to remove motion artifacts from the resulting video. The system also had to tell the operator when the X-Ray source or imager or robotics was having problems. When everything was running sweetly, our 8-axis system was controlled with a dual-joystick gaming controller, and the operator would smoothly 'fly' through the material, all axes (and zoom) working together smoothly and effortlessly. 8-axis system? There's translation in X, Y and Z, then rotation on each axis, with zoom making 7. What was the 8th axis? We added a trick: Stereo vision. If you apply an external magnetic field to the electron beam in the X-Ray tube just before it hits the target, you can move the target impact point. This is normally avoided for obvious reasons, and X-Ray tubes are magnetically shielded to preclude drift from local fields, including the Earth's magnetic field. But we poked holes in the shielding and added a small external coil that was driven by a 60 Hz square wave. Then we placed an active polarizer (basically an LCD sheet) over the monitor, and flipped it's polarization at the same rate, meaning anyone viewing the display while wearing glasses have a differently polarized filter over each eye would see a stereoscopic view into the material being imaged. The 8th axis was the strength of that field, how far the electron beam target spot would jump, and thus vary the perceived depth of the stereoscopic effect. All in real-time as the operator controlled the system. All the video was recorded, and I wrote programs to "paint" a 3D volume with what the operator had viewed, creating voxels. We also had an automated mode where the operator would define the volume, and the system would automatically scan it to create a filled 3D volumetric reconstruction. The imaging system programming contained "tricks" that would allow the system to detect the presence of finer detail, then iterate the scan at ever-higher resolutions until everything of interest had been imaged. We also had a "dumb" automated tomographic scan and reconstruction routine for doing quick-scans to help define the subsequent detailed imaging scan plan. Precisely one of these systems was built in the early 1990s (before the GPU was a thing, and a 'fast' PC ran at 100 MHz), for one of the 'TLA' government agencies. Why? Because during the 1970s and 1980s the Russians happened to leave some submarines on the ocean floor, and the US managed to gather up some of the bits and pieces, including electronics that had become encrusted with minerals and marine life, and which had undergone substantial corrosion. Yet our system was able to completely reverse-engineer any part of the system having enough material still located in its original place. It was straightforward to recreate the traces on each layer of a multi-layer PCB. We could even trace the metallization layers on the crude semiconductors used. Vacuum tubes were relatively easy, even when badly damaged by the ocean's depths. I didn't get to participate in any of the "real" use of the system. We were given carefully crafted simulations to image, then the customer would suggest changes. Then, after training some folks on the system, literally during the night the system was trucked away to "an undisclosed location", never to be seen again. We never even received requests for parts or service.
The robot rotational axes looked like a bunch of nested squares and C-clamps with pivots at the center and end of each member. Unfortunately, it was extremely annoying when part of the robot structure would get in the way of the beam, and we also had to deal with gimbal lock. One of the most fascinating operations the robot could perform was what we called a "tumble", where all axes were quickly repositioned to get the robot structure out of the way, with the object being examined having no net motion at the end. A few things broke while getting that debugged...
I got to use a system similar to this, a cabinet x-ray. The traces were visible and you could see inside of the die. This was used to look for soldering issues under cans and components like BGAs. I used to work for a small company that used to make cell phones called Research in Motion or better know as Blackberry. I'm guessing this is a bit too thick, and the x-ray is too strong to see anything.
There was a dropdown in top left of the UI for target source. I'm guessing you can chose different gun types, and this is a reminder that you can select high power mode on this machine
I had to mod one of those Dallas RTCs to add a standard 2035 battery holder on top and I used pictures as a reference for where to dremel the epoxy to access the pins. These X-Rays would've been immensely more useful to be able to drill more confidently not scared of being near something important. Really awesome contribution thanks CPU Galaxy (I also subscribed)
The stuff about the hydrogen furnace going boom brings back memories of visiting aerospace heat treating suppliers in my last job. They would always have at least one hydrogen retort furnace with what looked like an overgrown Bunsen burner going on the top. That was the hot hydrogen venting and burning in the air. And yes, occasionally some oxygen would find its way inside the furnace. And yes, the usual result was a rebuild of everything inside the furnace. But sometimes things would get more interesting. One supplier once had enough oxygen get in that the resulting pressure blew the lid, some 3,000 lbs of steel, off the top of the retort. It went straight up the 20 feet or so to the roof trusses, smashed them flat, then dropped back to the floor, cracking the 6" thick concrete. I was told the roof of the building had to be replaced, and part of the building nearly collapsed.
You can actually see the inner traces it’s just rather confusing to look at. The inner traces are that grid pattern, you can tell the grid isn’t just on the surface because towards the centre of the image where they overlap the traces are darker but where they are not overlapping due to perspective, they are lighter. Also, if you look at the edges while it’s at an angle, you’ll see that the top layer only has traces going left-right, the layer one below looks to be top-bottom (presumably computer layout). The bond wires on the RTC are a similar size to those traces but they’re much harder to make out because you’re trying to contrast them on the background of such a highly complex 3D structure. More modern machines and maybe even this one are capable of doing a μCT where you can then slice a 3D volume model afterwards and see the detail much easier but file sizes are huge and software is a right pain. I love the way the via’s bend! Really shows how they’re assembled layer-by-layer! And CPU Galaxy’s back catalogue is a true gem!
Dallas chip you can see the skinny DIP package they first used to put the chip in, then manually you had assembly workers take the tested chips, and bent 4 pins up, then used the lead bending jig to put the rest down into the normal DIP plane, then added the battery and crystal, soldered them on, then passed the completed assembly (now with the battery in storage mode, so the clock oscillator is turned off till first power up) down to the ultrasonic cleaning and then potting line. That is why they were assembled in a country with cheap labour, adding all those hand assembly steps to the process was likely the most expensive part, costing likely around 50c US per part.
Yeah the bent pin is obviously "how ya do it". I'm pretty sure that a mill can cut through the potting and battery without damaging the chip and the crystal, thus allowing a battery replacement. That would be a fun surgery job.
I always try to get my dad to do interviews for channels. He is/was a Chip Engineer and Worked on the first Radeon chip and the 1st Apple made Iphone CPU (A1) which was in the Iphone 3. He also worked at Pixim making sensors for security cameras and another company working on OLED age compensation algorithms (Increase power to offset dimming of the display as it ages). His last job before he retired was at LG working on custom ARM chips and the University they helped fund in Brazil.
the dies are SiO2, which is a glass, transparent to x-rays on the dallas semi clock, you don't see the die, but the metal (part of the spider) onto which the die is mounted. Quartz is also transparent to x-rays
You said, those Dallas RTCs like the 1387 shown would usually last for much longer than the specified 10 years. Sadly that is not my experience. Dallas 1287s were used in SGI Indy workstations. Already by 2000 the ones we at SGI were using for private pet projects had started to fail in large numbers so we acquired replacements. And it's already years since that replacement has failed yet again :-( Dallas has long stopped manufacturing those so it's not easy to find suitable replacementss. What a number of people have done is cutting the case open, disconnecting the internal battery and connecting an alternative instead.
Damn, imagine having a machine like that to play with. The most mundane things would become extremely interesting. Hours upon hours of fun. Maybe we should start a crowdfunding thing to get Dave one for a new '2 minute xray' segment haha
This is a negative image, meaning you only see the regions where the x-rays are absorbed by the material. X-rays are absorbed by higher Z materials, typically metals, hence why you can see Cu, Au and Ni of the UBMs and the solder balls. Silicon is transparent but, with high exposure times, multiple energy scans and multidimensional projections one can actually see the AlSi layers on CMOS processes. Also a comment on something said before: X-ray tubes do not output a specific energy but rather an energy spectrum defined by the target material. The maximum energy of the tube is set by the anode high voltage (electron acceleration voltage) while the intensity is set by the cathode current (it’s thermionic emission). The electrons get stopped at the target and the produce what is called bremsstrahlung radiation (in the x-ray range for this application).
@@Peter-fy4pj eh, a silicon wafer is about half a mm thick, and that sensor is ~ a meter away from the die. The mass of air encountered by the X-rays is going to be comparable to the amount of silicon they will encounter, and the atomic mass of silicon is not so much higher than air's, it might be very difficult to get a good image at all
Doesn't bremsstraulung radiation is wide spectrum up to the maximum energy of the electrons? while this machines operating mostly at voltages below 200kv mainly use the specific spectrum given by the atomic weight of the target material (k-emission)? I remember seeing the spectrum of an x ray tube with a W target operating at 120kv and it looked kinda like the spectrum of a fluorescent light bulb except it only had two peaks and the continuous spectrum was significantly less intense.
@@teresashinkansen9402 There are three ways x-rays are being produced in these tubes: - Bremsstraulung with atomic electrons: this is continues and primarily peaked at lower energies Bremsstraulung with atomic nucleus: because of the much higher mass of the nucleus the electron deceleration is almost complete resulting in Brem photons with the maximum incoming energy. Photons from shell electron excitation: These are produced when incoming photons collide with internal electrons and push them out of their respective quantum state. An outer electron will subsequently step in to fill the gap and a corresponding photon will be emitted.The energy of the emitted photon depends on the characteristic states of the target material. All three processes are superimposed at any given point. There is a trick though that one can do to cut-off lower energies by using absorbers between the sample and the incoming radiation and selecting the target material. Because absorption is nonlinear with respect to energy, the higher energy rays are less absorbed and with a well selected target the spectrum can become highly biased towards a monochromatic source. These kind of x-ray machines though are application specific since you cannot change the incoming photon energy (kind like the hospital x-ray tubes they use for simple x-rays, CTs are of the other type)
When i work at Ericsson we use to have a X-ray similar to that one. At least at the one we had, the depth of field was ridiculously short. Like a 1/10 of a mm. And the copper in the middle layers are very think. So its really hard to see them. On the prototype production line we had a segment X-ray that X-rayed in like 100 segments, they used math to subtract the segment from each other making sort of makeshift 3D model of the circuit boards. This file was saved at a server. When a board failed, we could check the X-ray database with out even touching the board. there was some software in the machine made to detect bad soldering. But i guess it was mostly made to handle larger series. For prototyping it was not optimal.
The image is likely viewed from the x-ray detector side making the DUT look upside down to the viewer. The imaging detector is under the DUT As the kV and mA is varied, the different materials' radiopaque properties are demonstrated. This is fluoroscopy xray imaging as compared to single exposure imaging. I serviced medical xray systems for over 20 years before moving into MRI service support. I lost count of how many boards, chips, and components I xrayed and still have some old plain film images in my collection. Great video Dave! -=Doug
12W of electrical power driving the electron gun. The x-ray conversion process is probably not very efficient. Also, the generated x-rays will spray in all directions, making the actual in-use beam far less intense.
You can only see the vertical "stacks" because there are so many vias stacked on top of each other. You would have to turn the power way down to see the traces without shining through them. The big circles are the pins while the small ones are via stacks.
Been working a bit with NDT X-ray devices - 12W is not a lot of X-ray power - But since its just a little IC, you dont need a lot of power. - For welding inspection of large oil tubes, they sometimes use up to 300KV 4mA which would be 1200W - But for small soft things like IC's chips and so, 12W is more than enough
@24:42 - Regarding the DS1287, Peter Wendt (a German engineer that worked for IBM) removed quite a bit of the potting material in composing the first "mods" (to disconnect the internal battery and solder on leads for a new coin-cell) to keep the modules working on a computer system. Note that a CR2032 is a larger diameter (the '20' portion) than the Dallas module. I've also heard that the compatible module produced by "ODIN" invert the battery below the actual chip die - also note the x-ray of the DS1287 was always from above (it is the dragon head always looking at you or inverted Einstein face effect).
The reason you can't see a silicon chip is because x-rays are attenuated by a combination of atomic number and density. Silicon with an atomic number of 14 and its given density of a thin chip - means that the x-rays go straight through without stopping, hence no image. Compare this with copper (atomic number 29) or gold for pins and bond wires (atomic number 79). By the way, this type of equipment is known as Microfocus X-rays. The secret to these sharp magnified images is because the x-ray source comes from a point source of about 1 to 5 microns; giving whats known as geometric magnification.
Those are hardly the only RTC chips built that way, it was a common way to do it back in the 80's and early 90's Later, everyone decided that a chip-external lithium primary was better than an internal battery (which would die and leave you with a useless RTC chip) or an external rechargeable battery (which could leak and destroy the board). There are still the old style being made as replacements, but I would be surprised to see new designs incorporating them.
@@EEVblog mine was a bit different, two cells inside. No X-rays however, plain old depotting using a rotary excavation tool. ;-) Got to connect my external battery to the contacts, though, so everything is fine and dandy again. :-D
I think that the cutaway TCM may have been used as part of the marketing of the S/390. I certainly know we had a similar cut open TCM when I worked for IBM New Zealand so I am guessing that lots of these were made for locations all over the world. The TCMs were so expensive that the IBM NZ parts centre had an aircraft to ferry these around the country in case of failure. They were too expensive to have multiples in regional parts centres as there were so few mainframes in NZ. They kept stock of TCMs centrally and a plane needed to be used to get the parts to where they needed to go to maintain the service level repair times. Happy days!
Interesting to hear an IBM guy from NZ! I'm in NZ myself and have a bunch of the old stuff in my workshop... S/3s, 1800, 1130 etc - hope to add a couple of S/360s soon! When I lived in the UK I remember a place just off the A12 in Essex - a scrap merchant in the mid 1990s that made a living breaking up the older TCM-based systems :-(
Certainly hope it helps "level up" the channel in the search rankings. It's not just about the sub count, so please go watch his video and enagage with them by commenting and thumbs upping.
I watched for 10 seconds and liked very much, thanks for the tip. I hope your shoutout helps his subscriber count. Although I like small subscriber counts as it makes me feel more special, but, I understand, numbers matter.
Oops, subscribed to Peter's... what a F*****G interesting channel, I wonder why RUclips didn't show it on my feed, because it's clearly in my areas of interest.
I took those Dallas RTC as a planned obsolescence thing and I took one apart and mounted a cr2032 holder to it instead of its welded battery in the day.
I know it’s hard for food grade production X- Rays to see low mineral glass at approximately 6 micro sieverts. Not sure about quartz or silicon. Used to validate food grade X-rays for compliance on metal and glass test wands. QRMP standard backup owner on inline product production for a while at a big company who’s famous for round formed potato products and crinkle cut french fries lol. They also didn’t see aluminium at all.
Isn't lead (as in tin/lead solder) rather famously x-ray opaque? If the vias are filled as part of the stack-up process, that could go a long way toward explaining why we can see them.
CPU Galaxy has 4.36K Subs, as of today(24th Dec 2020) Ive just subbed, i can see his channel going crazy...Ha ha ha. I hope so. That would be great if Dave has helped increase CPU Galaxy's channel size. After all EEV is pretty BIG(77K, on 24/12/20
Just imagine what we can do now....😳 that's always what pops into my mind. Look at cold war tech, and how amazing it was, and that some of it is still classified. Imagine what the latest experimental r&d effort from IBM, Nvidia, AMD, or intel would look like under a cutting edge gamma spectroscopy setup...
I'm guessing only the gold vias & pads (like our bones) are dense enough to block the X-rays, while they just power through the substrate & silicon dies & copper like they aren't even there...
assumming the 12w is output power well think of the efficiency of x ray machines which is typically quite poor like 99% heat 1% energy... afaik x ray stuff is cooled with liquid nitrogen for a reason...
Yes, very interesting. Not kidding. I guess one could change the battery in the Dallas chip. No need to remove the old one, just disconnect it and add an external battery. Ta-daaa! Thanks a lot! As you say, winner-winner, chicken dinner!
you can only see the vias because they are aligned with the xray beam, so the beam becomes weaker on this positions. if you tilt towards 45° at the beginning in the video you can see how the vias are becomming weaker and do smear. further more the vias are i supose thicker than the traces so even if tthe traces would be aligned with the beam you would probably not see them very well. because the chip is made of silicon and all the different p-type or n-type areas are silicon too with only different doping, and doping means it has extremly low concentrations of impurities (1-100ppm), furthermore you have some THIN oxide layers and all these different materials have no changing properties in respect to x-rays because only the atom-core size and amount of these atoms in alignment to the xrays have an effect on the xrays. so in other words, you can only see kontrast made by different chemical elements and enough "concentration" in this x-ray (absorption?) machine. on diffraction machines e.g. you would have completly different effects....
Hi dave! Could you do a video about safety mechanisms/ precautions with wall power at some point? I'm mainly a hobby engineer right now and since i lack in experience, i'm really scared of going up a notch to the 230V AC. I can't find any videos talking about this matter...😪
Hey Dave, try to get in touch with FrizchensFritz on Twitter. He does non destructive DIE shots. Might be really interesting to see inside those CPU DIEs.
It would be interesting to see, those objects xray images at different angles, captured and the using some Photogrammetry software piecing them together into a 3D model :D (I can do the Photogrammetry Bit if needed)
Looks like Dave had a lot of chrismas fun already with this stuff. IBM AS 390 is something I can remember from those early 90 days ... back then when I had to deal with punch cards and later with a bill that killed the departments budget cause my program did run very well - on that budget ... I needed 6 hours and the monthly budget was gone ... but after 6 months or so I also had found a few hundred millions of savings back then. But the first 6 months all went crazy with the bill arrived which the new employee has caused alone while every other costs were full in limit. 15 months later they could sleep better, the budget was intact again ... and I had left out of the blue tired of discussing my work and possibilities.
Hi Dave. Thanks for this collaboration and I love how you are commenting this content :). For the X-Ray machine some input from my side. All the contrast and sharpening is made by software. So without that the image would be very blurry. Therefore at some point you could see noise in the image which disappeared then. At some point you said the part is upside down, and this is very tricky coz its not upside down. The thing is that by X-raying you see only a contrast image of x-radiation which is shot trough the part. I tried it and no matter if you place it upside down or not, the image looks always the same and its very confusing. This is due to the fact that we just get an image if the radiation is able to go through the material. On one side (underneath the round plate in the machine) there is the X-Ray source and the flat square part on the top inside the machine which is moving to the side is the receiver to detect the radiation and creating an image. When you need a a lot of radiation to go trough a material very thin parts are not visible any more coz they are absorbing too less radiation to give enough contrast on the image. Therefore I believe we can not see any traces in the ceramic substrate of the TCM.
Thanks for doing this, it turned out great!
There are ways to reveal the traces, but they require a low-noise system and lots of patience. The basic techniques come from astronomy and the imaging of diffuse gas clouds and very fine structures, combined with math borrowed from tomographic reconstruction and X-Ray crystallography. To get good 2D images, first build up an extremely fine 3D image, then take slices of it. Basically, the system has to "stare" at the target at multiple energies from multiple angles, then carefully combine the many signals together. There isn't nearly enough information present in any single view to get such detail during live interaction.
Immediately subbed to your channel. I'll be looking forward to learning more about CPU's with you.
anybody that Dave Jones recommends gets an immediate subscribe from me - going to dive into your collection shortly. :D
As to the traces, I think some of them were visible, but just as ghostly X or Y axis straight lines - they seemed to appear when in focus but be transparent otherwise. Also as to the vias, some you could see perhaps the plating material on them, while for others the barrel was visible as well - not sure if that's a contrast thing or those had more/different plating on the vias.
I could put it in a CT scanner, but I am not sure how bad artifacts will be
Email from an X-Ray Tomography Specialist:
"The reason you can't see the copper traces is because they are planar to
you / perpendicular to the beam. Xray detectors have somewhat limited
contrast, and in order to get good penetration, you need to run high kV
to get good energy (material penetration), and have decent current
(flux) in order to have enough signal on your panel. If you have hard
enough beam (good penetration) to get through that steel you will
absolutely blow through thin pcb traces. When the full thickness of
those vias is exposed to the beam, the beam has to pass through the
equivalent of that thickness of copper, hence you can see it."
Seeing planar traces isn't "hard" in the physics sense, but it is rare to find an X-Ray system capable of working right at the edge of the physics. First, it can't be done quickly. If using tomography, the voxel size must be no larger than the trace thickness, and multiple passes over a range of X-Ray energies and intensities will need to be fused.
However, there are tricks that can be used by an experienced operator in near-live systems (presently ~1 fps with lots of latency). The key is to accentuate localized contrast by rapidly modulating the X-Ray beam energy and intensity, then doing dynamic fusing of the multiple exposures (with no motion between the exposures in a set). In the end, it takes a low-noise system, fine and rapid X-Ray control, a deep/wide imaging pipeline, a ton of math, and lots of patience.
The weakest part of any such system typically is the imager, and there are some old-school analog tools and techniques that can provide rapid response with enhanced contrast and low noise.
Makes sense. The part of the x-ray spectrum with low enough energy to be absorbed by the silicon in the quartz would already have been blocked by the crystal's case.
We use the same X-Ray manufacture at my job. I have X-Ray'd PCBs and can see the traces. However, when I X-Ray laminates and packages, they seem to be too thin to detect. What's interesting, you can see 1um bond wires with no problem. This machine saved me in a pinch to verify a solder bridge for a large BGA package once.
4:19 Dave, RUclips stopped sending email notification on 13 Aug 2020 the bastards. The only way to get notified is by push notifications in your browser. www.socialmediatoday.com/news/youtube-will-stop-sending-email-notifications-to-alert-channel-subscribers/583258/
Also, there are probably multiple solid power planes. If you set enough power to the xray source to get through the power planes then you will also happily go through the traces and not see them. You WILL see where multiple tracks overlap, like the annular rings overlap...
X-Ray tubes consist of a highly-focused electron gun aimed at a metal (often tungsten) target. The gun voltage determines the speed, the energy, of the electrons which determine the energy of the X-Ray photons emitted at the target point, ideally one X-Ray photon emitted per electron. The current of the gun then determines the rate of X-Ray photon emission, known as the X-Ray flux.
The X-Ray detector (where an array of detectors forms an imager) is sensitive to a wide spectrum of X-Rays. Whatever is placed in the beam between the tube target and the detector will act as an X-Ray filter, much like color films can filter visible light. The X-Ray energy is varied to find peak absorptions, then the flux is varied to adequately stimulate the X-Ray imager.
The X-Ray beam has the shape of a cone, so where in the beam you place the material to be imaged determines the zoom effect: Close to the X-Ray tube for high magnification, and close to the detector for 1:1 imaging. However, the closer you get to the X-Ray source, the more important the size of the spot of the electron beam on the target. X-Ray tubes made for extreme magnification require extraordinarily tight focus, and are called "micro-focus" tubes.
Unfortunately, the focusing of a high-energy and high-current beam on a tiny spot creates massive amounts of local heating, leading to melting and even vaporization of the X-Ray tube's target material. For this reason, such tubes tend to be water-cooled, and the target material is moved to avoid excessive wear at any single point, all done without any effect to the X-Ray beam itself. All micro-focus X-Ray beams have relatively short lives, even with cooling and target movement, so many are designed to be overhauled on-site.
Things get more fun when a multi-axis robot is placed in the beam to move the sample as desired: The robotic precision needed at very high magnifications can be on the level of single-digit microns.
I had to learn all this stuff not because I worked on X-Ray sources, or on X-Ray detectors, or on X-Ray robotics. I learned this because I worked on the image processing systems needed to greatly accelerate "dialing-in" the system settings to obtain ideal contrast video, to aid the operator in planning, executing and replaying imaging runs, and to remove motion artifacts from the resulting video. The system also had to tell the operator when the X-Ray source or imager or robotics was having problems.
When everything was running sweetly, our 8-axis system was controlled with a dual-joystick gaming controller, and the operator would smoothly 'fly' through the material, all axes (and zoom) working together smoothly and effortlessly.
8-axis system? There's translation in X, Y and Z, then rotation on each axis, with zoom making 7. What was the 8th axis?
We added a trick: Stereo vision. If you apply an external magnetic field to the electron beam in the X-Ray tube just before it hits the target, you can move the target impact point. This is normally avoided for obvious reasons, and X-Ray tubes are magnetically shielded to preclude drift from local fields, including the Earth's magnetic field. But we poked holes in the shielding and added a small external coil that was driven by a 60 Hz square wave. Then we placed an active polarizer (basically an LCD sheet) over the monitor, and flipped it's polarization at the same rate, meaning anyone viewing the display while wearing glasses have a differently polarized filter over each eye would see a stereoscopic view into the material being imaged.
The 8th axis was the strength of that field, how far the electron beam target spot would jump, and thus vary the perceived depth of the stereoscopic effect. All in real-time as the operator controlled the system.
All the video was recorded, and I wrote programs to "paint" a 3D volume with what the operator had viewed, creating voxels. We also had an automated mode where the operator would define the volume, and the system would automatically scan it to create a filled 3D volumetric reconstruction. The imaging system programming contained "tricks" that would allow the system to detect the presence of finer detail, then iterate the scan at ever-higher resolutions until everything of interest had been imaged. We also had a "dumb" automated tomographic scan and reconstruction routine for doing quick-scans to help define the subsequent detailed imaging scan plan.
Precisely one of these systems was built in the early 1990s (before the GPU was a thing, and a 'fast' PC ran at 100 MHz), for one of the 'TLA' government agencies.
Why? Because during the 1970s and 1980s the Russians happened to leave some submarines on the ocean floor, and the US managed to gather up some of the bits and pieces, including electronics that had become encrusted with minerals and marine life, and which had undergone substantial corrosion. Yet our system was able to completely reverse-engineer any part of the system having enough material still located in its original place. It was straightforward to recreate the traces on each layer of a multi-layer PCB. We could even trace the metallization layers on the crude semiconductors used. Vacuum tubes were relatively easy, even when badly damaged by the ocean's depths.
I didn't get to participate in any of the "real" use of the system. We were given carefully crafted simulations to image, then the customer would suggest changes. Then, after training some folks on the system, literally during the night the system was trucked away to "an undisclosed location", never to be seen again. We never even received requests for parts or service.
The robot rotational axes looked like a bunch of nested squares and C-clamps with pivots at the center and end of each member. Unfortunately, it was extremely annoying when part of the robot structure would get in the way of the beam, and we also had to deal with gimbal lock. One of the most fascinating operations the robot could perform was what we called a "tumble", where all axes were quickly repositioned to get the robot structure out of the way, with the object being examined having no net motion at the end.
A few things broke while getting that debugged...
I got to use a system similar to this, a cabinet x-ray. The traces were visible and you could see inside of the die. This was used to look for soldering issues under cans and components like BGAs. I used to work for a small company that used to make cell phones called Research in Motion or better know as Blackberry. I'm guessing this is a bit too thick, and the x-ray is too strong to see anything.
So I once heard a relation of biomimicry in x-ray systems and lobster eyes. A cubic lattice of tubes to focus a beam, is that a thing?
awesome input, thank you
Wow cool stuff :)
I just started watching this video and looks like the EEVblog viewers have already begun to work their magic. He's up to 4110 subs already!
great, i said i hope that happened. he must have mega money's worth of electronics in that video..wow.
Almost 9000
Sticker on the side of the xray computer .. "Ja High Power Target ist installiert" .. seems like a legit, not a toy warning :-)
There was a dropdown in top left of the UI for target source. I'm guessing you can chose different gun types, and this is a reminder that you can select high power mode on this machine
@@getrektfn5725 get a job dude.
Thanks for highlighting his channel. Had no idea it existed and is right up my street. Happy Christmas to you and everyone here.
I had to mod one of those Dallas RTCs to add a standard 2035 battery holder on top and I used pictures as a reference for where to dremel the epoxy to access the pins. These X-Rays would've been immensely more useful to be able to drill more confidently not scared of being near something important.
Really awesome contribution thanks CPU Galaxy (I also subscribed)
With these x-rays, it looks like it could be possible to sand or mill one side of the chip to get to the battery contacts.
@@anomaly95 it’s what I did with a dremel to disconnect the internal battery but I didn’t know what else was in there
The stuff about the hydrogen furnace going boom brings back memories of visiting aerospace heat treating suppliers in my last job. They would always have at least one hydrogen retort furnace with what looked like an overgrown Bunsen burner going on the top. That was the hot hydrogen venting and burning in the air. And yes, occasionally some oxygen would find its way inside the furnace. And yes, the usual result was a rebuild of everything inside the furnace. But sometimes things would get more interesting. One supplier once had enough oxygen get in that the resulting pressure blew the lid, some 3,000 lbs of steel, off the top of the retort. It went straight up the 20 feet or so to the roof trusses, smashed them flat, then dropped back to the floor, cracking the 6" thick concrete. I was told the roof of the building had to be replaced, and part of the building nearly collapsed.
You can actually see the inner traces it’s just rather confusing to look at. The inner traces are that grid pattern, you can tell the grid isn’t just on the surface because towards the centre of the image where they overlap the traces are darker but where they are not overlapping due to perspective, they are lighter. Also, if you look at the edges while it’s at an angle, you’ll see that the top layer only has traces going left-right, the layer one below looks to be top-bottom (presumably computer layout). The bond wires on the RTC are a similar size to those traces but they’re much harder to make out because you’re trying to contrast them on the background of such a highly complex 3D structure. More modern machines and maybe even this one are capable of doing a μCT where you can then slice a 3D volume model afterwards and see the detail much easier but file sizes are huge and software is a right pain.
I love the way the via’s bend! Really shows how they’re assembled layer-by-layer!
And CPU Galaxy’s back catalogue is a true gem!
Yeah, that makes sense actually, I should have realised that. Was completely confused by the square pads IBM uses.
Dallas chip you can see the skinny DIP package they first used to put the chip in, then manually you had assembly workers take the tested chips, and bent 4 pins up, then used the lead bending jig to put the rest down into the normal DIP plane, then added the battery and crystal, soldered them on, then passed the completed assembly (now with the battery in storage mode, so the clock oscillator is turned off till first power up) down to the ultrasonic cleaning and then potting line. That is why they were assembled in a country with cheap labour, adding all those hand assembly steps to the process was likely the most expensive part, costing likely around 50c US per part.
Yeah the bent pin is obviously "how ya do it".
I'm pretty sure that a mill can cut through the potting and battery without damaging the chip and the crystal, thus allowing a battery replacement. That would be a fun surgery job.
I always try to get my dad to do interviews for channels. He is/was a Chip Engineer and Worked on the first Radeon chip and the 1st Apple made Iphone CPU (A1) which was in the Iphone 3. He also worked at Pixim making sensors for security cameras and another company working on OLED age compensation algorithms (Increase power to offset dimming of the display as it ages). His last job before he retired was at LG working on custom ARM chips and the University they helped fund in Brazil.
the dies are SiO2, which is a glass, transparent to x-rays
on the dallas semi clock, you don't see the die, but the metal (part of the spider) onto which the die is mounted.
Quartz is also transparent to x-rays
You said, those Dallas RTCs like the 1387 shown would usually last for much longer than the specified 10 years. Sadly that is not my experience. Dallas 1287s were used in SGI Indy workstations. Already by 2000 the ones we at SGI were using for private pet projects had started to fail in large numbers so we acquired replacements. And it's already years since that replacement has failed yet again :-(
Dallas has long stopped manufacturing those so it's not easy to find suitable replacementss. What a number of people have done is cutting the case open, disconnecting the internal battery and connecting an alternative instead.
Damn, imagine having a machine like that to play with. The most mundane things would become extremely interesting. Hours upon hours of fun. Maybe we should start a crowdfunding thing to get Dave one for a new '2 minute xray' segment haha
I'd have to crowd fund the machine and a space to put it in... (and also get a license to have it, that's a thing here apparently)
This is a negative image, meaning you only see the regions where the x-rays are absorbed by the material. X-rays are absorbed by higher Z materials, typically metals, hence why you can see Cu, Au and Ni of the UBMs and the solder balls. Silicon is transparent but, with high exposure times, multiple energy scans and multidimensional projections one can actually see the AlSi layers on CMOS processes. Also a comment on something said before: X-ray tubes do not output a specific energy but rather an energy spectrum defined by the target material. The maximum energy of the tube is set by the anode high voltage (electron acceleration voltage) while the intensity is set by the cathode current (it’s thermionic emission). The electrons get stopped at the target and the produce what is called bremsstrahlung radiation (in the x-ray range for this application).
I fully agree. The tube seemed to be set at 100ish of keV. He should try to lower it until one, hopefully, starts to see the dies.
@@Peter-fy4pj eh, a silicon wafer is about half a mm thick, and that sensor is ~ a meter away from the die.
The mass of air encountered by the X-rays is going to be comparable to the amount of silicon they will encounter, and the atomic mass of silicon is not so much higher than air's, it might be very difficult to get a good image at all
Doesn't bremsstraulung radiation is wide spectrum up to the maximum energy of the electrons? while this machines operating mostly at voltages below 200kv mainly use the specific spectrum given by the atomic weight of the target material (k-emission)? I remember seeing the spectrum of an x ray tube with a W target operating at 120kv and it looked kinda like the spectrum of a fluorescent light bulb except it only had two peaks and the continuous spectrum was significantly less intense.
@@teresashinkansen9402 There are three ways x-rays are being produced in these tubes: -
Bremsstraulung with atomic electrons: this is continues and primarily peaked at lower energies
Bremsstraulung with atomic nucleus: because of the much higher mass of the nucleus the electron deceleration is almost complete resulting in Brem photons with the maximum incoming energy.
Photons from shell electron excitation: These are produced when incoming photons collide with internal electrons and push them out of their respective quantum state. An outer electron will subsequently step in to fill the gap and a corresponding photon will be emitted.The energy of the emitted photon depends on the characteristic states of the target material.
All three processes are superimposed at any given point. There is a trick though that one can do to cut-off lower energies by using absorbers between the sample and the incoming radiation and selecting the target material. Because absorption is nonlinear with respect to energy, the higher energy rays are less absorbed and with a well selected target the spectrum can become highly biased towards a monochromatic source. These kind of x-ray machines though are application specific since you cannot change the incoming photon energy (kind like the hospital x-ray tubes they use for simple x-rays, CTs are of the other type)
When i work at Ericsson we use to have a X-ray similar to that one.
At least at the one we had, the depth of field was ridiculously short. Like a 1/10 of a mm. And the copper in the middle layers are very think. So its really hard to see them.
On the prototype production line we had a segment X-ray that X-rayed in like 100 segments, they used math to subtract the segment from each other making sort of makeshift 3D model of the circuit boards. This file was saved at a server.
When a board failed, we could check the X-ray database with out even touching the board.
there was some software in the machine made to detect bad soldering. But i guess it was mostly made to handle larger series. For prototyping it was not optimal.
The image is likely viewed from the x-ray detector side making the DUT look upside down to the viewer.
The imaging detector is under the DUT
As the kV and mA is varied, the different materials' radiopaque properties are demonstrated.
This is fluoroscopy xray imaging as compared to single exposure imaging.
I serviced medical xray systems for over 20 years before moving into MRI service support.
I lost count of how many boards, chips, and components I xrayed and still have some old plain film images in my collection.
Great video Dave!
-=Doug
That's one of the coolest x-rays I've ever seen.
Engineering of that TCM takes the expression "state of art" to a whole new level.
"Ja,High Power Target ist installiert" the missing space behind the comma is driving me nuts.
LOL!
No budget left after buying such a maschiene I guess, Labelmaker tape is expensive
"Yes, the High Power Target is installed".
When zooming into that chip and stretching out the layers it reminds me of a spaceship going into hyperspace! So friggen cool!
Well, this channel has rapidly become one of my new favourite channels on RUclips.
You are right. He has a seriously underrated channel. I just watched half a dozen of his uploads. Good stuff. Thanks!
Quartz is fairly transparent to X-rays (some diffraction/scattering but probably negligible in this configuration).
12W of electrical power driving the electron gun. The x-ray conversion process is probably not very efficient. Also, the generated x-rays will spray in all directions, making the actual in-use beam far less intense.
You can only see the vertical "stacks" because there are so many vias stacked on top of each other. You would have to turn the power way down to see the traces without shining through them. The big circles are the pins while the small ones are via stacks.
Been working a bit with NDT X-ray devices - 12W is not a lot of X-ray power - But since its just a little IC, you dont need a lot of power. - For welding inspection of large oil tubes, they sometimes use up to 300KV 4mA which would be 1200W - But for small soft things like IC's chips and so, 12W is more than enough
@24:42 - Regarding the DS1287, Peter Wendt (a German engineer that worked for IBM) removed quite a bit of the potting material in composing the first "mods" (to disconnect the internal battery and solder on leads for a new coin-cell) to keep the modules working on a computer system. Note that a CR2032 is a larger diameter (the '20' portion) than the Dallas module. I've also heard that the compatible module produced by "ODIN" invert the battery below the actual chip die - also note the x-ray of the DS1287 was always from above (it is the dragon head always looking at you or inverted Einstein face effect).
Thanks man!
I also subscribed to the CPU Galaxy
Nice shoutout, love his channel and he definitely deserves it
You don't see Dave so excited everyday!
You've almost doubled his sub count. Nice
6 days later and he's tripled they guy's subcount. He's now over 10k!
13K right now. I just recently watched another video made by him about fake CPUs. He truly deserves it.
CT scanners use A Lot more power, Up to 84Kw of power 140KV to 80KV at 10-700mA for the systems I am familiar with.
The reason you can't see a silicon chip is because x-rays are attenuated by a combination of atomic number and density. Silicon with an atomic number of 14 and its given density of a thin chip - means that the x-rays go straight through without stopping, hence no image. Compare this with copper (atomic number 29) or gold for pins and bond wires (atomic number 79).
By the way, this type of equipment is known as Microfocus X-rays. The secret to these sharp magnified images is because the x-ray source comes from a point source of about 1 to 5 microns; giving whats known as geometric magnification.
Those are hardly the only RTC chips built that way, it was a common way to do it back in the 80's and early 90's Later, everyone decided that a chip-external lithium primary was better than an internal battery (which would die and leave you with a useless RTC chip) or an external rechargeable battery (which could leak and destroy the board). There are still the old style being made as replacements, but I would be surprised to see new designs incorporating them.
Pity the view is from a camera looking at his monitor... Can we not get some content from the direct footage?
An hour in and he got 4k ! Way to go Dave and thanks for the new sub
The RTC was actually more exciting : )
That was a very cool bonus!
@@EEVblog mine was a bit different, two cells inside. No X-rays however, plain old depotting using a rotary excavation tool. ;-) Got to connect my external battery to the contacts, though, so everything is fine and dandy again. :-D
The X-ray for this is just insane.
Cpu galaxy just got a big boost in subscribers.
I think that the cutaway TCM may have been used as part of the marketing of the S/390. I certainly know we had a similar cut open TCM when I worked for IBM New Zealand so I am guessing that lots of these were made for locations all over the world.
The TCMs were so expensive that the IBM NZ parts centre had an aircraft to ferry these around the country in case of failure. They were too expensive to have multiples in regional parts centres as there were so few mainframes in NZ. They kept stock of TCMs centrally and a plane needed to be used to get the parts to where they needed to go to maintain the service level repair times. Happy days!
I believe Peter did the machinging himself and cut it open.
Interesting to hear an IBM guy from NZ! I'm in NZ myself and have a bunch of the old stuff in my workshop... S/3s, 1800, 1130 etc - hope to add a couple of S/360s soon!
When I lived in the UK I remember a place just off the A12 in Essex - a scrap merchant in the mid 1990s that made a living breaking up the older TCM-based systems :-(
tint it green and you have a perfect matrix screensaver .. trippy .. great stuff !
LOL - he's up to 3.24K subs already.
3.81k at 23.15 GMT, 23/12/2020, 17 mins after your comment :)
4.26k now...what a boost, the RUclips algorithm is gonna love this guy lol
Certainly hope it helps "level up" the channel in the search rankings. It's not just about the sub count, so please go watch his video and enagage with them by commenting and thumbs upping.
@@EEVblog I'm watching it now, 1.5" CRT on a retro PC. Great production quality videos.
6k now
Used Xylon x-ray machines for inspection of inconel castings for jet engines. Big upgrade over film!
Just fantastic as always
And please get Edward in the amphour
Subscribed to CPU galaxy because I love me some X ray old CPU pornography! Excellent collaboration!
Daaamn, that's AWESOME AF having people X-ray stuff like that!
Only 15 hours and he now has over 6k subscribers. I’m one of them, such awesome content. Thanks.
CPU Galaxy Subscriber Count: 2,46k Subscribers
24 hours later.
6.8k Subscribers
CPU galaxy is really great.
I watched for 10 seconds and liked very much, thanks for the tip.
I hope your shoutout helps his subscriber count. Although I like small subscriber counts as it makes me feel more special, but, I understand, numbers matter.
Oops, subscribed to Peter's... what a F*****G interesting channel, I wonder why RUclips didn't show it on my feed, because it's clearly in my areas of interest.
Thank you 😊
I took those Dallas RTC as a planned obsolescence thing and I took one apart and mounted a cr2032 holder to it instead of its welded battery in the day.
I have no idea why I love his stuff, but I do.
I know it’s hard for food grade production X- Rays to see low mineral glass at approximately 6 micro sieverts. Not sure about quartz or silicon. Used to validate food grade X-rays for compliance on metal and glass test wands. QRMP standard backup owner on inline product production for a while at a big company who’s famous for round formed potato products and crinkle cut french fries lol. They also didn’t see aluminium at all.
If I could post pics I’ve scanned a few random items. ;)
Isn't lead (as in tin/lead solder) rather famously x-ray opaque? If the vias are filled as part of the stack-up process, that could go a long way toward explaining why we can see them.
CPU Galaxy has 4.36K Subs, as of today(24th Dec 2020) Ive just subbed, i can see his channel going crazy...Ha ha ha. I hope so. That would be great if Dave has helped increase CPU Galaxy's channel size. After all EEV is pretty BIG(77K, on 24/12/20
You missed a digit so out by a factor of 10. He has 773k subs 😁
@@MeppyMan sorry. read it wrong..i can be lazy sometimes. after another tough week at work. like a zombie gas fitter.lol
Just imagine what we can do now....😳 that's always what pops into my mind. Look at cold war tech, and how amazing it was, and that some of it is still classified. Imagine what the latest experimental r&d effort from IBM, Nvidia, AMD, or intel would look like under a cutting edge gamma spectroscopy setup...
Cool I was a subscriber of CPU Galaxy before watching this video :D
I'm guessing only the gold vias & pads (like our bones) are dense enough to block the X-rays, while they just power through the substrate & silicon dies & copper like they aren't even there...
Repeat after me: Mo-lib-den-um.
I would say it's more like Mo-lib-de-num
Lovecraftian creature name notation: Mo'Lib-Denum
Moly-dbe-num.
I went and subscribed, criminally underrated
A medical chest xray is somewhere in the order of 100-125kV and 50-80mA.
Amazing video!
By the way, green screen matches eye color, LOL
yes please that guy would make an interesting amp hour
Guess the silicon is too close too the ceramic pcb to be visible since you really just see the different density of materials
Please make a video on Rishabh multimeters
assumming the 12w is output power well think of the efficiency of x ray machines which is typically quite poor like 99% heat 1% energy... afaik x ray stuff is cooled with liquid nitrogen for a reason...
some beautiful sparc mcms in there
subd
Wishing you and your family a very happy Christmas Dave x
You either see the traces on the first layer or maybe 4, or all vias all the way through, not both.
Yes, very interesting. Not kidding. I guess one could change the battery in the Dallas chip. No need to remove the old one, just disconnect it and add an external battery. Ta-daaa! Thanks a lot! As you say, winner-winner, chicken dinner!
So beautiful dave thank you so much
Have you ever seen space hardened components? Like the hardware put on the modern mars rovers?
We've got GE Phoenix x-ray at work which is amazing. I wish I could look into this processor by myself.
A tripling in subscribers in 3 days, not too shabby lol
If you wanted to see the individual traces you would have to pass a current through them, and use SEM.
Oh my god you get into some crazy s*** my friend it's been a long time but you still as impressive as ever I see
Dave, I would love a high res photo of the RTC.... it would make a great poster for my office wall...
$$$?
Awesome thank you sir
age of IBM's superscalar processors
3:35 these 4 chips in the right look like DLPs from projectors?
Also just subscribed to CPU Galaxy :)
Thank you 😊
You are right they are texas instrument DLP
you can only see the vias because they are aligned with the xray beam, so the beam becomes weaker on this positions. if you tilt towards 45° at the beginning in the video you can see how the vias are becomming weaker and do smear. further more the vias are i supose thicker than the traces so even if tthe traces would be aligned with the beam you would probably not see them very well.
because the chip is made of silicon and all the different p-type or n-type areas are silicon too with only different doping, and doping means it has extremly low concentrations of impurities (1-100ppm), furthermore you have some THIN oxide layers and all these different materials have no changing properties in respect to x-rays because only the atom-core size and amount of these atoms in alignment to the xrays have an effect on the xrays. so in other words, you can only see kontrast made by different chemical elements and enough "concentration" in this x-ray (absorption?) machine. on diffraction machines e.g. you would have completly different effects....
How does this beast perform in cinebench?
Hi dave! Could you do a video about safety mechanisms/ precautions with wall power at some point? I'm mainly a hobby engineer right now and since i lack in experience, i'm really scared of going up a notch to the 230V AC. I can't find any videos talking about this matter...😪
The square box with fan, is the DR detector, not the micro focused xray head!
Viewing this on the 24th, guess this is my christmas present. :D
Thanks for finding CPU Galaxy. Amazing channel. Subscribing.
2:07 what is "RUclips Certified"?
OK, the density is absolutely insane.
Работал на Yxlon Cougar несколько лет назад. Трубка у него обслуживаемая, катод можно менять.
@10:27 Isn't he focusing on a spot with no silicon? On the visible spectrum image it looks like the silicon chips are darker.
Hey Dave, try to get in touch with FrizchensFritz on Twitter. He does non destructive DIE shots. Might be really interesting to see inside those CPU DIEs.
I wish I had one of those so I could do my own brain scans and stuff.
It looks upside down because the xray sensor is under the device being inspected and the x-ray emitter is above it.
What Dave said was the source is actually the detector; the source is underneath. We've got an older Cougar in our lab.
9:56 there is no die there. Just seconds before you've pointed that the camera on the right shows the target and there is no die under the cross.
Holy shit, that x-ray machine is some real sci-fi stuff...
It would be interesting to see, those objects xray images at different angles, captured and the using some Photogrammetry software piecing them together into a 3D model :D (I can do the Photogrammetry Bit if needed)
You can see viers but can't see the tracers dave maybe turn up the x-ray power
Looks like Dave had a lot of chrismas fun already with this stuff.
IBM AS 390 is something I can remember from those early 90 days ... back then when I had to deal with punch cards and later with a bill that killed the departments budget cause my program did run very well - on that budget ... I needed 6 hours and the monthly budget was gone ... but after 6 months or so I also had found a few hundred millions of savings back then.
But the first 6 months all went crazy with the bill arrived which the new employee has caused alone while every other costs were full in limit.
15 months later they could sleep better, the budget was intact again ...
and I had left out of the blue tired of discussing my work and possibilities.