After a bit of browsing on eBay I found several “new” versions of my favourite classic op amp, the 741. We’ve already seen about twelve different layouts so far, but today we’ll see that there are still more out there.
Starting with the oldest, we go back more than four decades to 1974. Texas Instruments at that point sold the SN72741, adapting the original Fairchild part number to their own nomenclature: SN stands for “Semiconductor Network”, meaning “integrated circuit”, while the numbers indicate the product series and temperature range (0 to 70 °C in this case; there was also an SN52741 with a wider temperature range).
After dissecting a whole batch of 555 timer ICs I thought it would be an interesting project to make a working copy of the 555’s internal circuit out of discrete components, in the same physical space as the original IC. I’ve done this before with the 741 opamp, and the steps are similar. First, I drew the schematic in KiCAD:
I used the original circuit as designed by Hans Camenzind, as a tribute to his design but also because it uses the smallest number of transistors among all different designs I found while dissecting the various 555 ICs. Discrete transistors are larger than discrete resistors, so the original design saves space compared to newer versions that include several more transistors. Note that it’s the opposite situation when you’re designing an actual IC: integrated resistors are usually the largest parts in your layout.
If there’s a classic analog chip even more iconic than the 741 op amp, it has to be the 555 timer. Released just three years after the 741, it similarly took the world by storm, selling billions of units over five decades. Quite unlike the 741, which established op amps as a common IC type, the 555 has remained largely in a class of its own. There are many ICs that can generate square or triangle waves, but I can’t think of any chip that can function as a one-shot, a flip-flop, Schmitt trigger, or one of a million different oscillator types like the 555 can.
Designed by Hans Camenzind in 1972, its story is described in detail in Camenzind’s own book Designing Analog Chips. I highly recommend reading it (available on paper or as a free download) if you’re interested in analog IC design. In Chapter 11, Camenzind shows the schematic of the original 555 timer:
Regular readers of this blog will have noticed that the 741 op amp features quite often. Apparently I’m not the only one with an interest in this venerable amplifier; the Wikipedia page on op amps contains a detailed description of the 741, several books and web sites describe and even celebrate the chip’s history, and you can buy a kit to make a large-size discrete replica from a company called Evil Mad Science Labs. I bought one of these because I thought it looked rather cool.
On the picture above you can see the 741SE compared to an original LM741. I got the SMD version of the kit, although you can also buy one that uses through-hole components and is shaped like a giant DIP package. Still, even the SMD version is enormous compared to the real chip, which got me thinking: would it be possible to make a discrete equivalent of the 741 in a space equivalent to an actual DIP chip?
I recently read a discussion on an electronics forum where someone had trouble getting an IR2104 to work correctly. He had bought these from a shady online store and could not get the correct signals to come out. I offered to analyze the chips, and one of the contributors to that discussion very kindly sent me a couple of them.
The IR2104 is a half-bridge MOSFET driver, which is used to drive the FETs in circuits like DC-DC converters and class-D power amplifiers. It’s made in a high-voltage CMOS process and is capable of driving the high-side FET at voltages up to 600V. The original designer and manufacturer is International Rectifier (IR), one of the first manufacturers of diodes, transistors and power management ICs. Currently the IR2104 is manufactured by Infineon, after it acquired IR in 2015.
A while back I dissected the Maxim DS18B20 temperature sensor, along with a counterfeit one. Today we’ll have a look at a couple more 18B20s from different manufacturers, all of them Chinese. I found a web store in China that sold me the XSEC SE18B20, the Novosense NS18B20, the 7Q-Tek QT18B20, the GXCAS GX18B20W and the UMW (Youtai) DS18B20. Prices varied from 65 cents to about two Euros apiece.
To get an idea of their performance, I put one of each on my Arduino board and placed it outside, where it was quite cold:
The good old 741 op amp is way more versatile than I first thought. After dissecting two sets of varying designs (here and here) I managed to get my hands on even more different varieties, from all over the world, from the 1970s to the 21st century.
The Raspberry Pi Pico is the latest in the Raspberry Pi series of single-board computers. Introduced in January of 2021, the Pico marks a significant change from the earlier series: instead of having a Broadcom system-on-chip, the Pico’s heart is a custom IC with a dual-core ARM Cortex-M0 processor at 133 MHz. Combined with its tiny form factor and lack of interfaces like HDMI, it feels more like a supercharged Arduino than a small PC.
This is the Pico. It contains the main CPU in the centre, 16 Mbits of flash memory next to it, a DC/DC converter on the right, and a USB connector to program the whole thing. I/O pins are scattered around the edge to enable easy interfacing to its environment.
The 486DX series were the complete, full-performance versions of the 486 processor line. Unlike the low-cost 486SX, the DX versions contain an FPU, which significantly speeds up floating-point calculations compared to the earlier 386. In 1992 Intel delivered another huge leap in performance by simply doubling the clock frequency of the 486 core from 33 to 66 MHz. Motherboards at the time could not cope with such high speeds, so the bus clock remained at 33 MHz. Although 40 and 50 MHz speeds were also available, the 66 Mhz version was by far the most common and remained the processor of choice for most high-performance 486 computers until Pentium systems became mainstream around 1995.
The DS18B20 is a digital temperature sensor that communicates through the One-Wire protocol. Like the DS2401 that we looked at earlier, it is housed in a three-pin TO92 package and can be read out using just one pin (plus ground). It’s made and sold by Maxim Integrated, although still labelled “DALLAS” on the package. Introduced in 1999, the DS18B20 was the successor to the DS1820 and provided a higher resolution (up to twelve bits, instead of the DS1820’s nine).
This part has become wildly popular over the past few years, because it is reasonably cheap, quite accurate, and very easy to use with microcontrollers. It is widely available online, but its popularity has, unfortunately, also given rise to an industry of fake DS18B20s. Although they usually work, there is no guarantee that they meet the specs listed in Maxim’s datasheet.
These fake chips are made by small semiconductor companies that have designed a drop-in replacement for the Maxim part, copying the exact functionality. There’s nothing wrong with second-sourcing a part like that (it’s how many semiconductor companies started in the first place), but marking them with the Dallas brand and selling them as if they’re genuine Maxim parts is of course illegal. Most likely it’s not the manufacturers that apply the fake label, but shady companies that buy large stacks of second-source chips, re-label them and sell them as if they’re the real thing.