electronupdate - 2018-04-02
When looking at a old mini computer CPU card I kept running across a chip from Motorola marked MC8641.... no data sheets were readily available on the internet. I suspect this may have been a custom chip for DEC. In this video I tear down the chip to figure out what it is. More details on my blog at: http://electronupdate.blogspot.com/2018/04/pdp-1124-mystery-chip-mc8641.html
Love seeing dies of the visible vintage.
I love your content. So unique and interesting.
Is this how the Chinese copy ICs?
I took apart a genuine 2N3055 and placed it under a microscope. Compared it with one that I bought from Banggood.
The one from Banggood, had a die about have the size of the genuine and it even looked different.
Can you do a video about fake ICs and transistors from China?
You don't have to go far to find them. I did the same tests on LM317 and found them to only work up to 1A instead of the genuine which can do 1.5A !
Do they go for smaller dies to be able to make more of them from a single raw wafer?
Thanks again for such interesting videos.
Yes, smaller dies are cheaper because they can make more per wafer.
Yeah, I added the last line to catch the Sherlocks.
Awesome!! Thank you I was bored but now it is getting interesting!
thanks, very interesting. I would like to decap too (to see what went wrong in dead IC's and fi mosfets, but the chemical stuff needed is a bit frighting (and hard to find over here)
Fascinating, the reverse engineering of designs like this is great to watch and it really helps to watch your process like this one, thank you.
Again, superb conten! Reverse engineering at silicon level is so rare. You really deserve more subs.
You have a very unique skillset!
Fascinating. Thank you.
Great video can't wait for the next one.
Very fascinating, and excellent delivery of information
have you ever thought about tearing down a very modern chip like an i7?
Nice in depth analysis. 👍
Could you please show how you decap an IC? I've a microscope and would like to check out a few dies.
I've tried the heat method (no 1$ calculator needed, because I got enough old ICs laying around to try it out with ;) ). I tried it with a heat gun (a small one for soldering and a big one that goes up to 600°C). Unfortunately that's not hot enough to remove the plastic (only weakens it, but not enough to be removeable without damaging the die). I also tried a blow torch and, yes, half the times I can expose the chip. But the heat usually messes up the small traces (they are just aluminium) and the chip doesn't look nice anymore (I'd like to get a few pictures out of it).
Hmmm, did you try heat and a scraper?
In a previous video, @electronupdate featured a cheap method via epoxy suspension and sandpaper. https://youtu.be/GQOW7si5-cw [He uses a surface plate, but if you don't have some Starrett granite lying around, I'd imagine any glass which was flat-pour manufactured (e.g. a standard silvered mirror) should be flat +/- a thou between any two points up to a foot, which is enough room to lap against. Glass is fluid enough that it will creep over time, so using an antique mirror might not be the best idea..]
Go slowly and try to apply continuous pressure. Every few minutes, rotate the chip by 45 degrees, so you minimize your chances of biasing your wear to one part of the IC. I'm not sure if 600 grit will be fine enough to resolve your feature components. I'd buy some cheap diamond lapping abrasive off ebay at various sizes (try ranging from 15-300 micron). You can buy premade paste compound or make your own with the raw abrasive and a solvent. Should be good enough for up to two metal chips I'd imagine.
Chemically, the standard method is using nitric acid. Depending on the composition of the potting material, you will have to experiment to find the proper amount of heat applied and/or the duration. A fume hood is basically required. Neoprene (*not nitrile*) lab gloves are what you want to use, along with full Neoprene and/or PVC full-suiting. Even with a fumehood, you need proper respiratory PPE (cartridged full-coverage masks, we used 3M's special line that was specifically certified for what we were using -- red fuming nitric acid-- concentrations, though it'd probably be good up for white fuming I'd imagine) I'd duct any exhaust fumes to funnel through an aqueous KOH or NaOH solution, because I'd feel like a dick just ducting it out into the environment without scrubbing beforehand to a netural ph, but I just wiki'd it, and nitric acid apparently autoprotolyzes fairly easily into NO2, NO3, and H2O so that step might be completely extraneous [ the moooooore you know ].
Definitely not something you should try at home. Decapping an IC uses using boiling, strong acids. Hot nitric acid being one of them. That's the kind of stuff that will happily explode, thermally burn you and kill you as it eats through your eyeballs. Another acid that sees use is (hot) Hydrofluoric acid. That's the kind that will happily poison you to death if you splash a bit on your hand or inhale the fumes. And melt your eyeballs. And give you thermal and chemical burns.
Take a college chemistry course first so you know how to handle strong acids in a safe manner.
To decap an IC, here is a method: https://youtu.be/eoRVEw5gL8c
do a vic II chip from a c64! or a sid chip!
Looks like it might have been used as a buffer, here's the data sheet I found: http://www.analog.com/en/products/amplifiers/operational-amplifiers/high-voltage-op-amps-greaterthanequalto-12v/ad8641.html#product-evaluationkit
You should reverse engineer some of the new CEM3340 Synth VCO IC clones that would be really really interesting.
Heck yeah. I have some defective 80’s era Yamaha VCO, envelope and filter chips that I would love to see reverse engineered. The envelope control chips should be interesting.
Aww hell yeah! Exactly what I wanted!! How'd you identify the NPN vs PNP and the BCE configuration? I'm probably pushing my luck, but could you do an analysis of say, a real simple, self contained system --- e.g. a 1970s Seiko 35SQ vintage? Or maybe a couple of 7400 series chips that wouldn't be too much trouble (latches?). I'd kill for you to tackle a 4004 or 68k but that'd probably take hundreds of hours to produce :[. Either way, SUPER COOL. Thanks man!!
The emitters are all tied together, so let me guess... Is it an ECL logic chip?
No, not ECL but TTL. The 5 transistors are being used as a 4 output gate, one input to a single transistor acting as a clamp to ensure it is fast turn off, with a base resistor and a pull down resistor, and then 4 open collector outputs in the chip to act as gating functions. Most likely part of the internal enable signal, which is about the only part of the chip that needs 4 separate outputs to feed the logic. 4 parts because there will be logic level changes on the open collector gate from the other sides, the open collector is pulling the internal signal down to disable the logic level from changing when disabled.
With the large area needed for polysilicon resistors you also often had a need for high value small resistors, so those were often implemented with a JFET instead, as you can run that pinched off and get a high resistance channel that gives a very roughly constant current, though it is not too accurate resistance wise, tolerances are in the range of +-50%, but it is small, and on chip most will be very close and track. Otherwise you designed in proper current sources and sinks, as the area they used was generally smaller, though you did have limitations, but the active ones made better pull up and pull down over resistors in most cases, though you would always have a saturation voltage close to the supply rail.
Given that the IC is on a digital board, and is only 16 pins, an analysis of the electrical properties can probably reveal its function without decapping it.
At the very least, the power and ground pins should be quite easy to identify.
Your say that "of course" the easiest way to find out what the chip does is to de encapsulate it. It's that always the case?
Tadesan yes. decap is the gold standard to determine a chips function. black box probing can sometimes be sufficient, but decap is the only way to get the netlist
This is really not much more than a 7400 quad 2-input nand gate.
IC 8641 is a bus driver receiver for Unibus with following
pins: INA=pin 14, OUTA=pin 13, INB=pin 11, OUTB=pin 10, INC=pin 2, OUTC=pin 3,
IND=pin 5, OUTD=pin6, ENB1(neg)=pin 7, ENB2(neg)=pin 9, BUSA=pin 15, BUSB=pin
12, BUSC=pin 1, BUSD=pin 4 (all bus signals are negated). Ref. PDP11bus
handbook page 71.
lol
BMR Studio - 2018-04-02
Early morning I’m watching semiconductor reverse engineering.... I need a doctor:)
UFOhunter - 2018-04-02
Lets make an appointment together and get half off :)
Pentti Kantanen - 2019-07-16
Congratulations, you are smarter than most people :-)