EEVblog #1358 - $250,000 IBM Processor X-RAYED

[EEVblog]

Collaboration with CPU Galaxy!
https://www.youtube.com/c/CPUGalaxy
The $250,000 IBM ceramic hybrid TCM processor module is X-RAYED!
BONUS x-ray of a classic DIP chip!

TCM Technical PDF: https://eevblog.com/files/ibm-system390-air-cooled-alumina-thermal-conduction-module.pdf



Original video:

[Kleinstein]

X-ray show heavy materials dark and light materials (e.g. silicon, quarz) bright.

So the vias are probably so visible because of the Mo-paste.  The trances are kind of visible, as the dark lines between the square grid. The traces are just a little dim and thus show up mainly were many are on top of each other. individual lines are probably a little to faint. Copper is still a relatively light element.

The class 2 ceramic caps usually contain some barium based materials and this shows up so dark.

One can adjust the contrast to different materials a little via the voltage. However with the very fine focus this can be limited, and the difference is not that large. So silicon and quarz and also alumina are really difficult to see. The absorption goes up with density times a something like the atomic number (not a simple function). So 1 kg of lead is quite a bit more effective in shielding than 1 kg of iron.  So heavy stuff shows up a lot darker.

The power for the x-ray is rather low - somewhat comparable to what a dentist would use (though only for some second or so).
The X-ray machine is really amazing, not because of the power, but because of the resolution. This is not just the sensor, but also the tube to make such a fine spot. So it's some 14 W in a spot of only a few 10 µm² size. X-ray optics is rather limited, so much of it is more like a pin hole camera.

The part shown moving in the xray machine is likely the detector. The tube is likely from below.

[dmsc]

That's a very good X-Ray machine!

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."

That's only half true.

The X-Ray is like any other electromagnetic wave.

The energy (in kV) sets the wavelength of the X-Ray light - so a bigger energy penetrates denser materials easily, the power (in A) sets the flux, so more power is the same as more light in a standard illuminator. At the shown energy levels (more than 100kV), most metals are semi-transparent, so the X-Ray light simply pass through the copper as light passe through glass.

Now, imagine the PCB made from glass, the vias would be visible because the light *reflects* on the vias if the light angle is low enough - the same is shown here with the X-Ray. Also, if there is some lead in the vias, lead is much more denser than copper, so it also will be a lot darker in the X-Ray image.

Most industrial X-Ray machines also have filters in the X-Ray generator to filter out the low energy light - the most used filter is simply a copper sheet, this will absorb the low-energy light and produce a much better X-Ray beam. You normally can change the filter to select different X-Ray wavelengths easily. I suspect that to actually see the traces you should select about 70kV energy and remove the filters, then the copper will be much less transparent than the other PCB materials.

[schmitt trigger]

Thanks for introducing us to CPU Galaxy.
He deserved better than a paltry 2.5k subscribers

[EEVblog]

Thanks for introducing us to CPU Galaxy.
He deserved better than a paltry 2.5k subscribers

Getting close to 10k subs now!

[b24clark]

Thank you for the interesting video of the IBM TCM technology.  The IBM multi-layer wiring and interconnection technology went through an evolution of form factors.  The technology you x-rayed was a hybrid technology between the first alumina ceramic technology and the advent of the most advanced, glass ceramic with copper wiring technology.

The substrate you X-ray-ed was an alumina substrate with top and bottom evaporated and etched copper metallurgy, passivated with an organic insulator.  The insulator appears yellow.  The metallurgy of the body of the white appearing alumina substrate is molybdenum.

You spent some time on why the 'traces cannot be seen.'  I'm assuming here that you mean the many many hundreds of meters of wiring that personalize each substrate part number, matching the unique personalization of the semiconductor chips that population of the top surface.

Here's the reason you cannot see the personalization wiring: 

All personalization wiring is done within a "plane pair" consisting of a mesh voltage layer, an X direction wiring layer, a Y direction wiring layer and another mesh voltage layer. 

The mesh voltage layer "wiring" are directly above the personalization wires to make the current return path of any signal as short as possible and to maintain the integrity of the wiring impedance.  So the density of wiring in the wiring "channnels" all stack up through 40 to 50+ personalization layers and associated voltage mesh planes and are therefore virtually invisible to the x-ray imaging.

Hope this helps.

[tocsa120ls]

The newest post from Ken Shirriff contains a cross-sectioned ceramic substrate: (nice highres pictures)

http://www.righto.com/2021/01/examining-technology-sample-kit-ibm.html

Following the 'don't fix if not broken', the PCB carrying these modules is also like 10-layers deep.