Do you remember the clamshell iBook from around 2000? In my opinion it was one of Apples best ever products. Rugged and one of the first laptops with mass user appeal. I used one for several years, it was a bit heavy but I really liked it.
So I thought I take a look at some silicon inside one.
This is a small board that links the battery to the main PCB that is in the base nestled close to the hinge
You can see the battery connection in the bottom image (Top left). The board is clearly related to battery charging and the main power regulation to the board, but I couldn’t really figure out exactly what it does, and why it needs its own board. There are a couple chips that I thought I would take a look at, a Maxim MAX675 a 5V Precision Voltage Reference and the Texas Instruments TL494C Pulse Width Modulation Control Circuits. Both had data sheets that I was able to locate.
The datasheet says the MAX675 takes a 15V input (From the battery) and outputs a precision 5V.
In the datasheet is this image showing the metal layout. I have seen this before in datasheets, but only in very old ones. I have no idea why they show it.And here is the die photo, which shows the metal layout is identical(click on image for high resolution version)
It is fabricated on a really old single metal Bipolar process, in fact I know that it is made on a 12 μm 24V Bipolar process. I know for certain as when I was looking for the datasheet I came across this document (Maxim Product Reliability Report) published in Dec 1997. It’s a strange document to be published on the internet, it shows Maxim’s reliability monitoring data on seven different processes as of 1996. It also (again strangely) shows some details of the process including the layer thicknesses and a diagram of the process in cross section. Here is the Bipolar process diagramIt is an old Bipolar process, made on <111> substrate. Very early on in 1960’s and 70’s Bipolar devices were made on <111> orientated silicon (I think because diffusions were faster). MOS could only be made on <100> and eventually even Bipolar moved to <100> orientation. The epitaxial silicon (The layer of silicon that the active devices sit in that is grown after forming a patterned doped buried layer) is very thick 17 μm! Here is the list of mask layers given
And if you look at the bottom of the die photo you can see the mask layer identification giving us a colour key for the die photo.The only difference is between 4 & 5 we have an extra layer 45. I think 45 is the N+ emitter and layer 5 is a resistor option (Presumably the P base or N” emitter do not have the right sheet resistance (Ω/square) to make desired resistor values. You can clearly see blue resistors that match the number 5 colour.
Whilst analyzing the die in the microscope, something looked odd, then it hit me, the N+ buried layer is massively mis-aligned by about 18μm! Look at the buried layer topography around this transistor.And again here
The N+ buried layer is a very low resistance path, necessary to lower the transistor collector resistance, which is a key transistor parameter allowing current flow from emitter to collector with a low ‘ON’ resistance. You can see the pattern in the die photo because an oxide mask is used when the dopant (Either Arsenic (As) or Antimony (Sb)) is diffused in to make the 4.5 μm deep buried layer region, this creates a step in the silicon that is propagated up through the epitaxial layer (And is necessary to align the subsequent layer to the buried layer). Clearly this was a working device, but it is hard to see how when for example in the transistor above the contact down through the epi to the buried layer is not even over-lapping the buried layer. I know of a phenomenon called epitaxial pattern shift that caused distortion, but this is clearly only in horizontal plane, it appears perfectly aligned vertically. If it is epitaxial pattern shift it is shifting ~1μm for every microns silicon that is grown so ~45°. Even if it is pattern shift how did they align to it? It is quite the mystery to me.
I am going to publish the Texas Instruments TL494C as a separate post to keep the size manageable.