A few months ago I looked at a Motorola MC6800 bought on eBay that was branded with a Freescale logo, contained an Hitachi fabricated die (probably from late 80’s or early 90’s) and had a 2016 assembly date code. Reader Jeremy wanted to compare a real vintage 6800 and kindly ordered this MC6802 for me.
I can make out from the very worn and faded markings
Which I believe is a week 52 1978 assembly date code. Released in 1977 as an upgrade to the MC6800, offering an internal oscillator and 128 Bytes of RAM.
Thus assembled almost exactly 39 years ago prior to the date of this this post. It is a truly vintage part, packaged in a ceramic DIP. I popped the lid off using a blowtorch to melt the solder. I also used a strong magnet to pull the lid off (To avoid damaging the die.)The die is soldered to the lead frame, one interesting feature the ground wire has two bond wires, one going to the die, another to the lead frame, but rather than bonding directly to the gold plated lead frame it is bonded to very small piece of silicon or metal – I can’t see why they did that.As it is in the package I am able to take a die photo with the bond wires intact. The package depth limits me to the 10x objective for the die photo. The working distance of my high magnification objectives is too small to focus on the die in-situ. To zoom in on the features I had to remove the die from the package.
MC6802 Die Photo
(click on image for high resolution version)
Back in 1978 3″ silicon wafers was state of the art. With a die size of 5.55mm x 4.96mm I calculate they would have had 99 candidate die per wafer. They were probably doing well to yield 25-30 die per wafer at the time. The datasheet also mentions an MC6808 which was identical to the 6802 but without the 128 Byte RAM, so they obviously recovered some failed parts where just the RAM was defective to gross more parts per wafer.
The die has a distinctive and unusual monochrome look to it. I removed the die (By heating the package on a hotplate and carefully sliding the die off of the solder. Which proved tricky to do with a very hot die and without a jig holding the die still.
You can see the pitted/mottled texture of the oxide layers which is obscuring the poly and active areas. The latter can be tricky to see in some structures.Normally the oxides are transparent (It is effectively glass), since this was sealed in a ceramic package, it must have come out of the wafer fab looking like this. I have seen this once before on another old Motorola part. So clearly it is a characteristic of the process Motorola used. It’s unfortunate for us as it makes viewing the structures quite difficult.
The main die logo is very visible confirming this is pucker Motorola MC6802
Not so visible, hidden under the mottled dielectrics there is another die mark R1H. Could that be Ray Hirt sneaking his initials onto the chip (He is one of the designers referred to on the MC6800 Wiki page) or is R1H for something else?
One aspect of taking the die photo in-situ is that it makes identifying the pin-out much easier. Here is the die photo with pin-out labelled.You can see the internal oscillator pins on the the top left side. T he ROM that contains the instruction set on the left, the data input pins at the top, and the output pins bottom and right sides.
There’s a 2kBytes (16kb) large ROM on the chip. A NOR type ROM programmed with a masked (depletion) implant. Comparing this image to a reference I found online and it is clear
(Attribution unsure found in lecture published on-line, original maybe slide set “Digital Integrated Circuits” by Rabaey et al ©2003)
The 128 Byte (1kb) RAM is also interesting. The datasheet mentions that 32 Bytes of RAM are retainable and this is clearly the smaller section on the left. I can trace the Vcc line of that section to the Vcc/Standby pin. (The larger right side goes to the main Vcc pin.)
Counting out the bits, (the small retainable section in one half is 16Bytes or 128bits, so this image contains 4 bits (Or 4 SRAM cells)In part because it is difficult to see the active area and use of buried contacts I cannot figure out the circuit used. Given this is an NMOS process I suspect it is a resistive load SRAM cell like this one, but I am unable to reverse engineer it from the layout (Without removing the metal).
I mentioned buried contacts, or more correctly buried poly contacts. These are contacts between polysilicon and substrate, used extensively on this part. In a CMOS process you would not find them. Here is an example of two such contacts on the die
Some commentary about the process quality you can observe. In 1978 the metal lines would have been wet etched (Likely in a buffered phosphoric acid mixture). Given that the etch quality of the Aluminum tracks appears very good to meYou can see a tiny amount of notching sometimes where the metal line crosses a poly line, and there is slight mis-alignment of metal to contacts, and the metal lines look a little small for the size of the contacts. But overall though this looks like a good quality process for 1978.
Finally many thanks again to Jeremy for supplying the part, and merry christmas to all of you who celebrate it.