Another chip from the clamshell iBook from 2000. Another Maxim power chip, this time it is from the main PCB.
Its a MAX785C which I am pretty sure is a dual (3.3V & 5V) PWM buck regulator for laptop computers.
I could not find a datasheet for the MAX785 but I did find one for the MAX786 which as you will see is almost certainly a minor change, or a slightly different version of the MAX785.
For the MAX786 Maxim were once again kind enough to publish a chip topography 🙂 and you can see it matches the MAX785 die photo and pad layoutHere is my MAX785 die photo with the pads annotated, you can clearly see the symmetrical and identical 3.3V & 5V PWM supply blocks with the 3.3V on the left.
Referring to the Maxim Reliability Report I found whilst researching the earlier Maxim chip I can say with confidence the MAX785 was made on the Maxim SG5 process that looks something like this (The transistors here are pretty drawn pretty ugly IMO)Its a CMOS process with a PNP transistor, zener diode and a Chrome/Si precision resistors. I have annotated the previous image with the layersThe layer 2 (PNP base drive) is/was unusual, they are making vertical PNP transistors. In a standard CMOS process you can build lateral PNP transistors using the regular process. However the gain of the lateral transistor is normally very low (As the base is defined lithographically you cannot make the base very narrow, unlike a diffused vertical base.) Here is a large Bipolar transistor on this chip which is sandwiched between an two arrays of them to make high current drive transistors.
Most of the large output transistors in the PWM blocks are I believe multi-finger MOSFET devices, here is a zoom image of one of them (It s hard to see the polysilicon gate as they have stitched metal lines along the gate, you can see the single gate contact at the very bottom right of the image.)
Update: Laser trimmed resistors?
Frank commented that he could see some laser trimmed resistors, this was a very eagle eye observation! A bit of background, most silicon resistors are made with polysilicon or diffusions, and there are a number of variables that limit their accuracy such as thickness, width, dopant concentration, amount of dopant electrically activated during thermal processing. It is typical that a resistor value is at best +/-10%. For this reason most designs require only differential or ratio accuracy, and here the variables cancel out and you get very accurate matching. Well over 90% (Perhaps 99%) of analog ic’s make do with these resistors. Occasionally a design needs an accurate absolute resistor, for these, a few processes (Like this Maxim SG5 process) offer a precision thin film resistor. These are made from thin (Typically 20nm-100nm) metal layer like Chrome used here. I believe they can be made with +/-1-2 % accuracy. Sometimes a design needs even more absolute accuracy and for that they laser trim the thin film resistors, using a high power laser to ablate the metal track usually on a probe station where they measure the desired signal before and after trimming.
It is a very complex serpentine that enables a wide variation in resistor value depending on where cuts are made. After staring at this a bit and thinking some more, I don’t think they are laser trim cuts. The cut material looks too clean for laser ablation. What I think has happened is they designed the part with a un-cut serpentine resistor, and then evaluated the initial prototypes making various laser cuts, and then used the results to change the resistor mask. In high volume, this would be much cheaper than trimming every die on a probe station.