Skip to main content.

Swiss Physicist’s “Big Hack” for Intel 4004’s 52nd Anniversary

An international collaboration, a lot of hand-soldering, and incremental debugging
(Biggest challenge: “Digital” FETs are really analog if you look closely. 0.2 volts is a tiny threshold)

Klaus’ 4004,4002,4001x PCBs all turned on Klaus’ 4004 PCB closeups: logo and decoder

The ambition to build a giant replica of the Intel 4004 microprocessor dates back to the mid-2000s when Tim McNerney approached the Intel Corporate Archives seeking schematics for this historic integrated circuit—the first, single-chip CPU that companies could buy off-the-shelf and program themselves. McNerney also reached out to Italian semiconductor pioneer Federico Faggin, who had designed both the 4004 (in 1971) and the 8080 (in 1975) for Intel before he jumped ship to found Zilog (in 1976) and launch the rival Z80—of Radio Shack TRS-80 fame and countless Soviet era personal computers.

Fast forward three decades. The Z80 was still licensable as an IP core, powering zillions of TV remote controls, while Intel was advertising their successful Pentium II on TV with colorful “bunny suit“ dancers and “Intel inside” stickers. By then Intel’s three founding CEOs had all retired, and gone with them a long-standing rivalry with Faggin. It is truly curious that the Intel founders never forgave Faggin, even after their company became a household name through the fantastically successful 8088, 286, 386, 486, and Pentium families that powered the PC revolution.

Fortunately in 2005, McNerney got to join Faggin when, for the first time since his 1975 exit, he was invited back to Intel’s headquarters building and presented with the original schematic drawings of his 4-bit brainchild plus transparent “check plots” of the chip’s artwork preserved in the Intel Museum’s archives. If this peace offering had not taken place, the Intel 4004 35th Anniversary Project might never have gotten started. [continues below]

Klaus’ 3 PCBs on Klaus’ work bench Klaus’ 4004 PCB backlit

McNerney’s bargain with Intel seemed simple enough: Intel would release the schematics and artwork of the 4004 for non-commercial use, and in return, McNerney would build for Intel an interactive museum exhibit with a fully functional 4004 built out of individually-packaged transistors on a 41x59 cm (16x23”) printed-circuit board (PCB) using the same geometry as the original chip. He and his team soon realized how hard this challenge would be. The 4004 check plots were good enough for viewing, but had many distortions and tiny, but significant errors. After McNerney designed and Fred Huettig built the interactive parts of the museum exhibit, what Faggin accomplished in one year of sleepless nights (and no CAD software), took an international team of six engineers, all working in their spare time, two years initially plus (years later) a final two year, two person push—just as the world was emerging from the Covid era.

The keys to success were threefold:

  1. Schematic capture and verification, Fred Huettig (USA)
  2. A simulator/analyzer that compared the schematic to the chip artwork and provided “oscilloscope traces” for every internal signal, Lajos Kintli (Hungary)
  3. Someone with both the interest and perseverance to get a complex project over the finish line, Klaus Scheffler (Switzerland)

In a related, equally important effort, Brian and Barry Silverman (Canada) wrote graphical simulators and reverse engineered the Busicom 141PF desktop calculator, which was the very first software application written for the 4004. This provided a fun, practical example, and a verification testbed for the giant replica 4004. The synergy between teammates was essential to keeping the project going. Kintli's first involvement in the project was to study the Busicom assembly language firmware and add all-important comments to the code. [continues below]

Klaus’ 4004 PCB flyover Klaus’ 4004 PCB layout

Inspired by the 4004’s 50th anniversary, MRI physicist Klaus Scheffler contacted McNerney about the project—“can it be done?” was his main concern. After McNerney and Scheffler met in California, for months and months, Scheffler incrementally built up and tested his 4004 prototype, painstakingly soldering 1,782 transistors, 49 LEDs and 434 resistors by hand, while troubleshooting the giant circuit section-by-section using the “scope traces” from Kintli’s simulator. The power-up reset circuit was harder to debug. For this and for integration testing, Kintli wrote a Teensy/Arduino-based test driver that simulated the MCS-4 system bus, the external RAM and ROM, the rest of the Busicom calculator hardware, plus 4004 assembly language algorithms for computing digits of Pi.

After many Zoom debugging session with Kintli, Scheffler got his transistor 4004 running the complete Busicom calculator app at 250kHz, only 3x slower than the original 750kHz chip. At first you might think “how slow,” but IMHO this is really a respectable result. PCBs and discrete field-effect transistors (FETs) have a lot more capacitance than the microscopic FETs inside a microchip. Using low-capacitance RF FETs were key to achieve this performance. Most common, discrete transistors are high-current FETs used in switching power supplies, and have roughly 10x higher gate capacitance.

Once the prototype was fully working, Scheffler found a pick-and-place electronic assembly company in Germany to build a commercial quality 4004 PCB worthy of a museum. The latest 4002 PCB is colored a very attractive blue.

But Scheffler did not stop there. He went on to build a giant 4002 RAM. Then came the question of what to do about a 4001 ROM replica. Thousands of RF transistors get expensive, even purchased at a discount from surplus electronics dealers in Hong Kong, and he just couldn’t bring himself to waste so many transistors building a non-reprogrammable ROM PCB. So the team agreed on a compromise: use a single re-writable, vintage-compatible EEPROM chip and build just the specialized bus and I/O logic using discrete transistors. Thus was born the 4001x, which Scheffler used to complete his giant 4-bit “microprocessor.” For a hypothetical future museum exhibit, it would be easy to build a ROM array using discrete transistors, but for experimentation, being able to change the ROM contents is important.

Now that Scheffler has a fully working 4004 computer built out of discrete transistors that computes the digits of Pi, he looks forward to a demonstration of another historic, commercial application: the Kienzle-Argo 1140, the world’s first electronic taxi meter developed in Germany.

Credits

Legal Notices