When the Grumman F-14 Tomcat first flew in 1970, it was a marvel. With its variable-sweep wing, twin tail, and sleek lines, it quickly became one of the most iconic jet fighters of the era — and that was before a little movie called Top Gun hit theaters.
A recent video by [Alexander the ok] details something that was far less well-documented about the plane, namely its avionics. The Tomcat was the first aircraft to use a microprocessor-driven flight system, as well as the first microprocessor unit (MPU) ever demonstrated, beating the Intel 4004 by a year. In 1971, one of the designers of the F-14’s Central Air Data Computer (CADC) – [Ray Holt] – wrote an article for Computer Design magazine that was naturally immediately classified by the Navy until released to the public in 1998.
The MPU in the CADC is called the Garrett AiResearch MP944, and consists of a number of ICs that together form a full computer. These were combined in the CADC with additional electronics to control many elements of the airplane automatically, including the weapons system and the variable-sweep wing configuration. This was considered to be essential based on experiences with the F-111 and its very complex electromechanical flight computer, which was an evolution of the 1950s-era Bendix CADC.
The video goes through the differences between the 4-bit Intel 4004 and the 20-bit MP944, questioning whether the 4004 is even really an MPU, the capabilities of the MP944 and its system architecture. Ultimately the question of ‘first’ and that of ‘what is an MPU’ will always be somewhat fuzzy depending on your definitions, but there is no denying that the MP944 was a marvel of large-scale integration.
Thanks to [Stephen Walters] for the tip.
On a similar note; I had just graduated as a BS EE and started working at Hamilton Standard. They made propellers and fuel control for the F15 (1969-1970). The fuel control was built from individual TTL logic gates wire die bonded onto carriers that were mounted on a substrate and cooled by the incoming jet fuel. The design was hand drawn, showing the logic gates along with logic statements describing the equations. This was a 12 bit CPU with 16 instructions and 12 bits of address space. Jack K wrote an assembler and all coding was in assembly language. All software ran on a PDP computer fed by paper tape. The good old days!!
And now we run our CPUs in Excel.
https://youtu.be/5rg7xvTJ8SU
Thanks for the link.
Uhh, no
Wow. Propellers. Must have been a really early iteration of the F-15.
The propellers on the F 15 were not for propulsion, they were part of the wind turbine power generating system that fed all of the avionics. At Mach 2.5 they generated plenty of power. The propellers also helped to avoid bird strikes to the jet air inlets.
:)
“They made propellers and fuel control for the F15”
I suspect he meant:
“They made propellers and, for the F15, fuel controls.”
Spot on! Reminds me of an old grammar advertisement in our city. :)
Here’s the translated version.
” ‘Let’s go! We’re eating [,] grandma!’
Using commas can safe lives.”
Using commas can sa*v*e lives.
It was an interesting era in a lot of countries. As far as I know, this didn’t have a MPU but had lots of early integrated circuits. Swedish Viggen 37 had a computor “https://sv.wikipedia.org/wiki/Centralkalkylator_37” Central Calculator 37, which integrated with the analog control system and had plenty of sensors and inputs, including radar altimeter and ILS, and fed output to the HUD, radar screen and weapons. The computor replaced one man and made it a single seater. Ironically the computor started as a navigational computor for a missile, but was improved, expanded and adapted.
“Disturbing margins and delay were evaluated in one batch of MLE (Micro Logic Elements). These factors were considerably improved when Fairchild changed the manufacturing process to RTL (Resistor-Transistor-Logic). The reliability was checked in different ways, both in laboratory and in computer tests. Our early tests and information from manufacture and other users indicated that the element could reach a satisfactory reliability. In June 1966 a total of 38,000,000 MLE-hours including 319,000 during flight had been accumulated in computer prototypes. One observed catastrophic failure after 3000 hours resulted in a failure frequency less than 0.006% per thousand hours, which was encouraging. We also examined the TTL-logic elements (Transistor-Transistor-Logic) from Texas Instrument. They were faster, had a lower disturbing margin but had to be mounted on multi-layer cards, which had a relatively low reliability at that time. We thus continued with the first choice of Fairchild who also gradually extended the MLE family. Saab was one of the biggest customers of Fairchild beside NASA in the beginning of the 1960s. Early component investigations and tests used a lot of MLE. The first 5 prototypes, delivered during 1962–1963, needed about 3000 MLE each. ”
“The computer was a real computing centre for all co-operating electronic equipment such as air data, reference platform, airframe mounted accelerometer unit, radio altimeter, navigation equipment and the attack radar. Computer inputs also came from a data panel and panels for navigation, weapons, radar and test. Output signals went to the head-up display, radar display, and other displays in the cockpit. All programs and input/output variables had to be repeated 10-60 times per second to keep track of the aircraft’s position. The computer contributed also in an overall electronic system test before take-off. The number of system variables was about 700 in about 30 program blocks. The 10 prototype computers were delivered between 1965 and 1968. Specification data of the computer are shown in Appendix 3. They were used in aircraft Saab Lansen 32 alfa and gamma and in the Viggen 37-3. A careful notation of the prototype reliability indicated at the end of 1967 an MTBF of 1000 hours (calculated 1200) during ground operation. Component reliability and soldering reliability were also satisfactory and indicated that an MTBF during
flight operation of more than the required 200 hours could be reached. ”
“In January 1969 the Swedish Air Materiel Department FMV ordered serial production of the computer CK37, 114 units. First delivery was in April 1970. The same year the number of computers was increased to 196. The final serial version of CK37 consisted of 4 units, each with a weight between 11 and 16 kg. Power consumption was about 600 watts.”
200.000 instructions per second, 48 basic instructions. Memory 8192 words, 28 bits with 2 parity bits (about 25 kbyte). Variable part 1536 words, the rest write protected. Cycle time 2,8 microsec. The biax memory had a similar specification. 64 analogue in/out signals with converter, 0.2 % accuracy within –10,24 to +10,24 volt. Conversion times about 200 microsec. 450 bits of binary in/out-signals, mainly collected in 13 bits word, represented by 0 or +12 volt. All signals are protected by in/out filter.
Computer systems when compared to the mechanical / analog systems they replaced are amazing. There is a McDonald /USAF documentary from the early 60s , directed at pilots and maintenance crews, explaining the operation of the various systems of the F-4 Phantom. One system is built to regulate the rudder operation. It uses bellows affected by airspeed and density as well as weights and extended levers that operate based on g-load to change the amount the rudder is allowed to move during various flight profiles. Anyone familiar with computers and modern sensors could understand how a few pounds of electronic equipment could be used to duplicate (and exceed the capabilities of ) the work done by the hundreds of pounds of mechanical/analog devices used on prior aircraft. Add to that the ability to change parameters via settings and software makes the switch to computers and electronics amazing!
Agreed, but credit to the genius of the mechanical guys that designed the system the electronics replaced. Wow. Makes the electronics seem simple.
Just more proof the United States military always got the best shit first
Even internet
Imagine what kind of technology they have in 2024
After looking at old reactor cpus and icbm control computers, and those already old if yall got em.
Gel Packs.
What did he mean by this
Star Trek tech
“old reactor cpus and icbm control computers”
If it works, don’t fix it, especially when controlling such critical things.
Hey, the U.S. military has enough money to fuel their jets with liquified unicorns!
I am curious where Lee Boysel’s Phase IV Systems AL1 factors into this,
https://ece.engin.umich.edu/stories/lee-boysel-the-early-history-of-microprocessing
Since it was used to demonstrate the tech he published in papers prior to make the process and techniques essentially public domain. It was used in court also to demonstrate that TI nor Intel could lay claim to the first microprocessor or CPU.
Boy that takes me down a Wikipedia rabbit hole. I think sometimes if Frank Wanlass and Lee Boysel had teamed up, Intel woulda been fairly slow to catch up. Wanlass did some great things at General Microelectronics and General Instrument. If Wanlass and Boysel had worked together,… who knows how quick LSI woulda grown and squeezed out TTL and the TI 74xx era.
Ken Shiriff actually has a fairly compact but pretty complete rundown of the competitors for ‘first Microprocessor’, and in specific the CADC’s relation to the AL1 in his TMX1795 article:
https://www.righto.com/2015/05/the-texas-instruments-tmx-1795-first.html
I agree with his argument that the CADC isn’t really a ‘microprocessor’ as commonly defined, being a multi-chip system and not even really having a general purpose ALU. The argument that the AL1 isn’t either (microcoded and usually multi-chip, the TI trial demo system notwithstanding) is a bit more arguable.
The CADC and the 4004 had multi-chips, however, the 4004 required 59 TTL chips to run a decent application. The CADC did not require any external logic.
My dad was an electrical engineer for Grumman. Worked on the F-14, the A-6 and the LEM. No idea if he had anything to do with this thing in particular, but thank you for the story regardless. Brings back good memories.
Considering Garrett AiResearch is a company that made/makes turboprops engines and such, that’s wild. [yes I know they made more than just those… https://en.wikipedia.org/wiki/Garrett_AiResearch ]
If it was space and air flight Garrett AiResearch was involved. Amazing company and amazing design teams.
As someone who extensively researched the history of “the first microprocessor” I will be the first to admit that I hate even the idea of “firsts” and the concept of the lone inventor hero. In the modern world this notion is largely kindling for mythology. Modern breakthroughs are developed by teams and built on a tower of innovations going back to the Industrial Revolution. Ray Holt is certainly a pioneer and deserves his place in the annals of history. The F-14 CADC was a remarkable computational chip-set that was way ahead of its time. And when you look closely, even Federico Faggin had legitimate competition from Texas Instruments for the title of first microprocessor. No one person can stake a flag at the top of a mountain. Even Intel stopped making the claim. It was just not worth the cost of paying lawyers to wear “asbestos business suits”—the question too muddy. And in the end it didn’t matter. F-14s kept the skies safe. TI sold millions of pocket calculators. Intel dominated the PC market for decades. Faggin went on to found Zilog and champion the optical mouse. Chances are excellent that one of the TV remotes in your home has a Z-80 inside. Oh, and yes, hotels did put a microprocessor in every doorknob.
I absolutely agree. I don’t even normally namedrop engineers in my videos since most breakthroughs are the hard work of huge teams, and built upon the work of many before them. I decided to in this case since Ray wrote much of the source material for the video. I wanted viewers to leave with the realisation that ‘first’ usually isn’t well defined and rarely matters. In this case, my main takeaway was ‘hey, this chipset was really advanced for the time!’
There was even a competitor for ‘first’ made just down the road from me in Scotland (the PICO1/GI250), but that’s another story for another time.
Nice comment Tim.
Reminded me that the 8086 had an optional math coprocessor, the 8087, released a decade later.
The 8086 was released in 1978. The 8087 in 1980. Not quite a decade.
These IC chips were mounted on multi / micro wire controlled impedance circuit boards, another amazing invention of the time.
Not sure of your comment. Please explain more.