The DS2401 is a chip that Maxim sells as a “Silicon Serial Number”. That’s pretty much what it is: a chip that contains a serial number which you can read out using a digital interface. This is useful if you’re producing a large number of identical devices but you want to give each one a unique number so you can track each individual device. This chip contains 64 bits of information, consisting of one eight-bit number indicating the chip family (for the DS2401 it’s 01), a 48-bit serial number, and an eight-bit checksum to verify the number is correct. Maxim guarantees that no two DS2401s will have the same serial number, so in theory they can produce 248 = 281,474,976,710,656 of these chips before they run out of numbers.
The DS2401 has a One-Wire interface, which is a clever serial interface designed by Dallas Semiconductor. It has one ground and one active wire, which doubles as data and power supply. The chip gets its power from this wire, and contains an on-board capacitor to store charge so it can keep working even if the data line is “low” for a certain amount of time.
The One-Wire interface is quite easy to use, and it takes just a few lines of Arduino code to read out the serial number from the DS2401. I bought three of these chips, which turned out to have serial numbers 00001D04714B, 00001D04763D and 00001D0498E7. As you can see they’re not exactly consecutive, but since they came as loose chips in a bag there’s no way of keeping them in the right order anyway.
Intel introduced the 80486 processor in 1989, as a successor to the 80386. It provided a significant increase in processing speed: a 486 was about twice as fast as a 386 at the same clock frequency. This was achieved mainly by a pipelined design and 8 kB of on-chip L1 cache. The initial 486 design also included an FPU, which in 386 systems was still an optional, separate chip.
In late 1991, Intel introduced a low-cost version called the 486SX, and renamed the original processor 486DX. The only difference between the two was the presence of the FPU: a 486SX system was much slower than a 486DX when performing floating-point calculations, but was otherwise identical. The absence of an FPU did not make much of a difference for many common tasks such as word processing, and the 486SX became very successful in low-end consumer PCs, displacing the last generation of 386 processors.
Today we’ll look at two very similar chips. Both are 486SX processors running at 25 MHz. Although the ‘SX was also available in 16, 20 and 33 MHz versions, I’ve found that 25 MHz was by far the most common. I’ve owned several systems with this processor, but since it was so easy to upgrade them to much faster DX or DX2 processors all SX chips quickly ended up in my spare CPU box.
A while ago, I found this little PCB in my junk pile:
It came out of an old webcam that I used on my PC back in the early 2000’s. Although I didn’t keep any pictures of the webcam itself, I recall that it was a Logitech Quickcam Messenger, a cheap USB camera meant to be used with MSN Messenger and other instant messenging programs. I remember using it for that purpose a few times, but once the novelty wore off it just ended up gathering dust in a corner somewhere.
From 1986 to 2008, HP sold a line of workstations and servers powered by a processor line called Precision Architecture, or PA-RISC. These systems were mainly targeted at professional users in industry and academia, and provided high performance at very high cost. Nevertheless, like all computers they became obsolete after just a few years, and I managed to find a heap of old HP equipment in the recycling bin of the local university. Today we’ll have a look at the PA7100 CPU and its accompanying “Viper” memory/IO controller (MIOC), both made in the early ’90s.
Here we see both chips side by side. The CPU has HP part number 1FT9-0002, although according to the sticker on the left it used to be 1FT9-0006. Not sure what’s going on here… In any case, it’s housed in a large ceramic PGA package with capacitors on top and a screw terminal to attach a heat sink.
The Viper has part number 1FZ6-0006. Its package is similar to the main CPU’s, but slightly larger and with fewer pins.
The 741 is easily the most iconic opamp ever made. Designed by Dave Fullagar at Fairchild as a user-friendly, general-purpose op amp, it became a huge hit with electronic designers. Billions have been produced since its introduction in 1968, by a wide variety of manufacturers. What’s perhaps even more amazing is that it’s still being produced today by TI and ST, despite being hopelessly out of date. Fifty years of development has produced a wide array of opamps that are faster, more accurate, less noisy and less power-hungry than the 741. Yet somehow, this classic part keeps hanging on, basically unchanged for over half a century.
Today we’ll have a look at the insides of a couple of different 741 chips. Although it is entirely possible for manufacturers to just copy the exact layout, especially for something as old and simple as this, it turns out that each company actually makes its own unique design.
National Semiconductor LM741CN
National has been manufacturing this device for ages, as part of their LM (Linear Monolithic) range of analog ICs. Even after National’s acquisition by TI in 2011, it has remained in production alongside TI’s own uA741.
The layout is very compact; the total die size is only 1100 by 800 microns. Some test structures are visible along the lower edge. The ID on the right edge of the die says “LM741U” and “UK”, the latter probably referring to National’s fabrication plant in Greenock. Some of the PNP transistors have circular emitters; this improves their matching properties and increases their breakdown voltage.