Merlin and computers in Brazil

In 1995 I had this extremely brief history of the Merlin Smalltalk computers I had developed. I will go into more details in posts here over the next few days, but I will start with some context about computers in Brazil.


History is complicated and what seem to be opposing narratives might be simultaneously true. So take the following as my opinion and not too seriously.

For the first 300 years Portugal kept Brazil on a very short leash. Nothing could go in or out directly, but only through the mother country. And almost nothing could be made locally. Printing presses, for example, were forbidden so it was hard to spread subversive material like in the North American colonies.

Then the Portuguese court had to be saved from Napoleon by the English and the price they demanded was direct access to Brazil’s market. The English took over and were able to successfully suppress local development throughout the 19th century through purely economic means.

This history led to the “colonial mentality” where the foreign is always considered better than the local even if it is not objectively the case. Some features of this mentality are being extremely picky (“I will get this French hat, this Swiss clock, this German toolbox…”) and an intolerance for the natural development cycle. The mentality is present in the middle class and up while the middle class and down is limited to access of local things (still true in this Internet age due to the language barrier) some of the attitudes do trickle down. So if you get a gold medal you are the most fantastic person in the world, but if you get a silver medal you are an incompetent fool who should have given up earlier rather than embarrass us all. There is no notion that the guy who came in last today might get a gold medal in eight years with the proper support or might just be a footnote in history without it.

A small group is very much against this. I will call them “nationalists” though rooting for the underdog is a major part of their motivation and that happens in other contexts as seen in movies like Tucker and Arrow.

Though local industry was finally started in the early 20th century, the real turning point was the building of a steel foundry by the Americans in World War II in exchange for Brazil joining them (even though a dictator with large Italian and German populations would naturally favor the other side). When the military returned from deposing dictators in Europe the first thing they did was depose their boss.

The new democracy had excess cash from sending supplies for the war while not getting bombed, so the population went into an import frenzy which the American industries restarting the production of civilian goods were more than happy to supply. This is known as “the Yo-yo economy” since most of what was coming in were not exactly essential goods.

In the late 1950s there was a reaction to that with a new state company Petrobras to control all aspects of oil exploration and the start of local automobile production by foreign companies Volkswagen, Chevrolet and Ford (joined by Fiat in the late 1970s).

In 1964 a military dictatorship took over and, in a very ironic move by someone dedicated to fighting communism, established many more state owned companies which dominated the economy. The years from 1969 to 1973 are known as the “Brazilian economic miracle” with an average growth of 10% per year. That all came crashing down with the oil crises and the sharp increase of the interest rates of the foreign money that had financed that growth. The trend of replacing imports was taken to extremes.

There is a popular thesis in the Internet that innovation can only happen where there is freedom. The thing about a dictatorship is that if you can convince the right person, an idea that otherwise would have remained academic speculation forever can actually get implemented. The problem with that is that most ideas are bad and the result might be starving half of your population to death. But once in a while you get interesting things like replacing gasoline with ethanol as fuel for nearly all cars.

The 1970s was the nationalists’ time to shine since they didn’t have to overcome the colonialist mentality but just befriend the right general. They made Brazilian aircraft, a space program, the largest hydroelectric plant in the world, a Formula 1 car, were developing a nuclear submarine and the atom bomb and many things like that. When Jimmy Carter halted weapons exports to Brazil due to human rights violations the country created the world’s third largest arm industry from scratch in just four years.

Tomorrow I’ll put computers in this context.


Fascinating! Thanks for laying the historical and political foundations, @jecel.

In 1917 IBM (then still CTR) arrived in Brazil, followed in 1919 by Burroughs. After a major role in the 1920 census, IBM established a proper local presence in 1924, the same year Burroughs did the same. IBM’s factory in Rio de Janeiro started operations in 1939.

The first computer in Brazil was a Univac 120 imported in 1957 by the city government of SĂŁo Paulo to keep track of the water distribution system. In 1958 Burroughs started local manufacturing of electric calculators while in 1961 IBM started making its 1401 in Brazil.

Still in 1961, José Ellis Ripper, Fernando Vieira de Souza, Alfred Wolkmer and Andras Vásárhelyi decided that their end of course project at ITA (Aeronautical Technology Institute) would be a working computer. They were able to get their hands on one of the first batch of transistors ever made in Brazil, but were shocked at how different the characteristics of one device were relative to another. So they carefully tested each of the 1500 transistors and then designed the circuits for their computer based on that. It had only 8 words of 8 bits each with a light attached to each bit since the idea was to be educational rather than practical.

The computer was named Zezinho (the Brazilian name for one of Donald Duck’s nephews) but it didn’t last long as the next few classes reused its components in their own projects.

By 1969 computer and parts imports had grown to US$13 million. And as happy as the Navy was to take its new ships for a spin, some within its ranks grew extremely worried about the Ferranti Argus 600 that controlled their missiles. If they had any problems, would they have to stop a war so a technician could come from England to fix it? They tried to convince BNDES (National Bank for Development) to finance the creation of a domestic computer industry in the name of national security.

In 1971 the first integrated circuit (in ECL) is made by JoĂŁo AntĂ´nio Zuffo at the microelectronics lab (LME - created in 1968) of the Polytechnical School of the University of SĂŁo Paulo (Poli-USP). Meanwhile, the laboratory for digital systems (LSD, also created in 1968) had invited IBM researcher Glenn Langdon to give a graduate course on computer architecture. Impressed with the results of the final test, he suggested that the students combine the best features into a single design and that USP should get them money to actually build it.

In 1972 IBM started making the S/370 145 in its new factory in SĂŁo Paulo, along with tape drives, controllers and soon after terminals and printers.

As the computer was coming together at LSD-USP, they saw a newspaper article saying that the Navy was looking for some group to build a computer to disprove the popular opinion that this technology was beyond Brazilians. The article said that Unicamp, a rival university, was going to build the White Swan computer (the Navy’s mascot) so they promptly named their own effort the Ugly Duckling (Patinho Feio).

The computer was very PDP-8 like but with an 8 bit datapath, 12 bit address (4KB of core memory) and 16 bit instructions. It used 450 TTLs in 45 printed circuit boards. In the big inauguration in July with the state governor and the press, a photographer tripped on the power cable and the demo program that had been painstakingly toggled in was lost (which I wouldn’t expect with core memory). Nobody was too upset and LSD got the contract with the Navy backed group to build the hardware (with the Potificial Catholic University of Rio de Janeiro being hired for the software).


I’ll continue the story later, but I would like to mention that MAME now has an emulation for the Ugly Duckling computer which you can run in your browser but all text on that page is in Portuguese.

By 1976 computer and parts imports had grown to US$118 million and the installed base was 5964 computers, 4105 of which were minicomputers (corresponding to only 8.5% of the market by value). The market shares in revenue were:

  • IBM 63.6%
  • Burroughs 20.2%
  • Honeywell Bull 4.2%
  • Univac 3.6%
  • Olivetti 1.7%
  • NCR 1.5%
  • DEC 1.8%
  • others 3.8%
1 Like

Thanks for the link to Felipe Sanches’ emulator! I see he has a thread about the Ugly Duckling on this other forum of interest:

(Most of the forum is in Portugese but this thread is in English.)

1 Like

I was just reminded about this topic while reading a discussion about something else. I see that the author of this topic is the person who mentioned Merlin:

There can’t have been many ARM2-based machines developed outside Acorn:

1 Like

I still plan to continue this thread, but am not very good with deadlines (still trying to finish the children’s computer I started in 1983…).

When I first read about the ARM in Byte magazine I really wanted to use it for the follow-on to the 68000 based Merlin 2. In early 1987 there was a short article about a complete chipset in some electronics magazine and I became even more interested. But due to getting burned waiting for Motorola to released the chipset for the 68000 (this has been discussed in other threads - “UK’s Amiga”) I had adopted a policy of not using any chips that were not already being used in some other product. That made me design the Merlin 3 with the 68020 instead.

In November of 1987 I learned about Acorn’s Archimedes, but by then Merlin had become a software only project for the PC-AT. I bought the chipset anyway and got the VLSI data book in early 1988. I wrote an assembler for the ARM and a friend wrote an ARM simulator. We wrote a subset of a PC simulator in ARM assembly and ran a benchmark showing that the ARM2 could simulate a PC with a performance similar to the original 4.77MHz PC (which was still selling reasonably well at the time). That was to try to convince my partner that we should restart the hardware side of the project. That didn’t work so the parts sat in a drawer for a few years.

In 1992 I did restart the Merlin 4 project, but instead of buying new components I just used what I already had though they were pretty obsolete at the time.

About other ARM2 projects, I think Apple used it in a graphics accelerator and Herman Hauser in his Active Book.


Thanks for providing a list of Smalltalk Computers. The slides for your Smalltalk Computers, Past and Future presentation complemented that page well. It shows some alternative approaches to the kind of computing we have now that are interesting and thought-provoking. I’m sure that many of the hardware implementations could be done in FPGAs now – I’m sure some of them must have been done already.

Any of those old Smalltalk processors would easily fit into a reasonably low cost modern FPGA. I doubt there would be any bits to be found to run on them, however. In most cases there wasn’t much in the first place, with academic projects only being completed enough to run simple benchmarks to publish a paper. Some of my own projects from back then are in tapes I can no longer find any machine to run.

We do have the full listing of the Dorado microcode for Smalltalk and most of the other bits from various Smalltalk-80 implementations, but the pieces that are in BCPL are missing, I think.

1 Like