Notes from the vault – 0x03

Part of group NOTES FROM THE VAULT.

Started 13Sep2021, updated 20Oct2021. (Burde jeg ha skrevet dette på norsk? Eller en på hvert språk? – translate)

Two early Autronica VLSI chips

In this note I will, hopefully with a lot of help from my friends, tell the story of two VLSI chips from the early 1980s . They were made for the Autronica company in Trondheim, Norway. We might succeed in pulling this up from the historic wastebasket and moving it over to the the Internet Archive’s Wayback Machine. I was employed at Autronica all of my professional career (1976-2017), bit I neither worked with the specification nor the design of these chips. But I was on the team that implemented the second (?) “loop controller” that used the smoke detectors which contained these chips. Standard disclaimer. This note is in work.

Confidentiality

All this stuff is from around 1982. Fig.1 has been on the wall of the Autronica Collection as long as it existed (I think). Customers from all over the world were invited there. I would not think that anything here would be subject to confidentiality or any non-disclosure agreements (NDA). However, to be careful I have only published half of the area of the chip in Fig.1. The full plot is available, but is password protected. The same will go for anything else I want to me careful with, like the Autronica at the time chapter.

18Oct2021: In favour of keeping the confidentiality level: none (4 replies, including mine).

Autronica Fire and Security (https://www.autronicafire.com/en/) is presently (Oct2021) owned by Carrier (https://www.corporate.carrier.com). This is, from a formal standpoint, interesting. As one of the persons I asked about this replied:

«As for the plot, I can not see who today would want to abuse it. However, there is an American owner behind who then owns all IP of the products. But the product is old and out of production for many years. Based on this it should be possible to show the full plot.»

I guess that from a formal point of view it does not matter that it was displayed at the Autronica Collection and that I literally saved it from becoming paper waste. But in my opinion, the pragmatic view is that showing the real plot, not only half of it, would be acceptable. This would even enter into the same tradition as f.ex. Intel, Computer History Museum (CHM) and IEEE [6].

Of course, the level of detail may be of interest here. I have shown it at some practical pixel width of 4480 @ 6 MB, but even then it would take some guesswork to understand and fully reverse engineer.

The photo shows 300 pixels of this plot.

To produce the 4480 scan I initally made a 10916 width png @ 312 MB, but it would reveal no more detail than seen here.

I guess, this is the final blow against confidentiality. From one of the commenters – and we are soon forty years into the history of these chips:

Norwegian English
..Men jeg tror jeg kan svare på det med konfidensialitet. For plot/masker så er de beskyttet ved lov bare i 10 år. De kunne prinsipielt ha vært beskyttet av en NDA med lengre varighet, men neppe 30+ år. Og de har ingen økonomisk verdi lenger.
Jeg ser heller ikke noe annet som trenger beskyttelse så lenge etterpå. Det er kanskje mer viktig å dele det som historie.
..But I think I can answer the case about confidentiality. For plots/masks, they are protected by law only for 10 years. They could in principle have been protected by an NDA with a longer duration, but hardly 30+ years. And they have no economic value anymore.
I also do not see anything else that needs protection so long afterwards. It is perhaps more important to share it as a story.

The plan

The idea here is to get quite some help from those who worked with these chips. Hopefully they would want to have their names here as well.

My plan is to get some idea of The plan, the vision, the specs, the story (or stories), and perhaps some names – of people, tools and technology. Who produced them in the final end, and how many transistors or gates did they have? Will any circuit diagrams have survived? What has not been thrown into the wastebaskets and what has not sunk into oblivion? This note should not go into much detail of how the units worked, nor how the smoke detectors worked. But I would like to know why Autronica got these two units made in the first place, and why, some ten years later, they were replaced by the immensely popular PIC microcontroller by Microchip Technology, initially coded in assembler.

About my style. I have added text in different chapters at different times. Therefore, some things might be contradictive. I think I will keep it this way, since my assumptions vs. what is correct (*) may hopefully be of some interest. If it’s bad enough I will of course try to fix it before it appears. Also see narrative style disclaimer.
(*) Correct? In English to remember is some times described as to recollect. I once heard by a neurologist on the radio that this is how the brain works. I guess that’s why pictures and scans make such a valuable collection.

Intro

From the note 194:[Aside: retrospective] – which follows a private note about my own LSI learning in the 1980s:

This was at a time when Autronica (the company I always worked for) already in 1982-83 had had two VLSI chips designed for them. An address unit and a detector unit for their first addressable smoke detector (as addressable, a first – even worldwide). These chips were used for many years. As some guys at work did the specification and testing, the VLSI implementation was outsourced to SINTEF in Trondheim. I also recollect. The original colour paper printing that they used to show to people resides by me, at least for the time being. At some moment in time the alternative was the dustbin.

Autronica Fire and Security AS moved from Lade to a new building at Bromstad in Dec2016. Some of the objects of our company museum were brought along. But in Nov2018 the “Autronicasamlingen” (Autronica Collection) was dismantled. Those of us who met did pass a scary container on the way in, we therefore tried to save what we could. This picture is from the collection in Oct2016, where the below proud printout is seen framed on the wall (yellow arrow). It was still there on the teardown day. I think this was the only part I saved. Oblivion happens even by me, I’m afraid. I went away that evening turning a blind eye. But there are rumors that the container in the end was not topped up. But what indeed survived, I think now is rather scattered around. I hope, even some at where they should belong: at the new Autronica location.

(Aside: I walked to the previous Autronica building today (14Sep2021) and shot the inset in the photo of the sign. The buildings seemed empty. The five floor main building is protected and is going to get some new use when the rest of the buildings will be demolished and new use will arise (Ringve pluss). I worked there for 40 years and have so many good memories from that collection of concrete, glass and metal – however, with lots of nice people inside.)

Where did these chips fit in?

Some of the data is from [2].

  1. Analogue detectors (1959-1979): panel BS-1 to BS-10 etc. Detectors: BE-1 etc.
  2. World’s first analogue addressable detectors: panel BS-3, detectors BJ-3, BH-3. Thick film technology was used
  3. Analogue addressable detectors containing the chips in this note (1983-1994). These were mounted on a thick-film substrate.
    1. BS-30 panel with Intel 8035 microprocessors, coded in assembler. The panel processor used a scheduler that I built. The loop processors had no scheduler
    2. 1989: DYFI® handling of the detectors in the loop controller with PL/M-51 and Real-time executive for 8051-type single-chip microcomputers. Panel with Intel 8088 coded in Modula-2 + run-time
  4. Analogue addressable detectors with PIC controller (1994-2017)
    * The first were coded in assembler, the next in C. AutroSafe coded in C and mostly VxWorks
  5. With present state of the art (2017?-?)

Now study further at [1].

Technology at the time

Fig.6 – Technology

This figure is from [5] page 245, where it’s not quoted. It’s probably from a local compendium made by someone at the Institutt for fysikalsk elektronikk at NTH or at the ELAB – mentioned on page 244.

According to this, the technology was capable of about 250.000 transistors at the start of 1982. If it took a year to develop, then that present state of the art would have been about 160.000 transistors. (All assuming logarithmic growth.) This would be the maximum for our chips. I assume the number in reality was substantially lower.

1986 LSI course at NTH

From the course LSI – konstruksjon og hjelpemidler (in Norwegian) or rather LSI – design and tools I took at NTH in 1986 I get these facts, copied from my exam paper (below, PDF is here, but you won’t get my grade!). Of the red text, I only now was able to recollect SPICE! Not strange I guess, since I didn’t follow up much. I guess, the name throwing here, as well as the names, may be of interest. Especially since some/one of them had been participating in designing the Autronica chips.

Hovedemner

Alternative realiserings-metoder, strategier, NMOS og CMOS teknologier, prosessering, grensesnitt mot leverandør, simulering med SPICE2, TEGAS5 og RTSIa, utlegg med CALMP, testvennlig konstruksjon, testmetoder, teknisk/økonomisk avveining.

Forelesere / lecturers
Forskningssjef H.M. Bayegan, A/S Elektrisk Bureau
Siv.ing. Frank Berntsen, Nordic VLSI
Forsker H. Danielsen, ELAB
Førsteamanuensis Arne Halaas, NTH
Forsker O. Marvik, ELAB
Ingeniør Jan Meyer, Nordic VLSI
Professor Einar J. Aas, NTH
Direktør Oddvar Aaserud, Nordic VLSI

Ansvarlig faglærer / responsible
Professor Einar J. Aas, NTH

Main topics

Alternative realization methods, strategies, NMOS and CMOS technologies, processing, interface to supplier, simulation with SPICE2, TEGAS5 and RTSIa, layout with CALMP, test-friendly construction, test methods, technical / economic assessment.

1st chip (1982)

This is a plot printed out from some program on some computer. I guess the guy who printed it used a flatbed plotter with ink. This probably is the addressing chip / “adressepakken”. It basically counted pulses of a certain length and restarted on another length, allowing smoke detectors on a two wire loop (two wires out – detectors – two wires returned) to be individually handled. Loop, so that one broken wire may be detected while all detectors still are fully operational.

Fig.1 – UTLEGG AV AUTRONICAS FØRSTE KUNDESPESIFISERTE HALVLEDERKRETS JANUAR – 1982. Kun 50% synlig! (Layout of Autronica’s first customer specified semiconductor chip. January – 1982. Only 50% seen!)

Aside: As much as I personally would have liked to see this chip reverse engineered if it’s possible, by someone (and published here!?), I have only shown half of the surface in the diagram above. The upper right half is copied as a triangle, rotated and pasted on the lower, left. The full plot is seen here – but it’s a password protected zip file for the time being (16Sep2021).

The drawn frame of the document above is 28.9 * 29.4 cm. The text is seen below the plot, and below the figure. The terms customer specified, is a direct translation from kundespesifisert. But I guess custom defined, application-specific integrated circuit (ASIC) or even what we did use at the time: Very large-scale integration (VLSI) also cover. (VLSI started with MOS transistors in the 1970s).

Strengthening the assumption that this is the addressing chip is the fact that we may see seven similar parts on the top (the seventh is not seen in the 50% visible scan which is now present). This points toward seven D-type flip-flops, which would do 28 divide [0..255] on eight pins, or 27 divide [0..127] on seven pins, if the first of the seven was used as a conditioner only. See Wiki-refs.

The latter is most probable since the address range was [0..127] of Autronica (but before 1989 only [0..99] was used by the detector range).

2nd chip (fragments)

It may be that this also is the chip shown above, since I can also see seven equal components lined up on this chip. But these could also be the resistor ladder for the analogue to digital conversion, if the unit contains any such. Since analogue values are converted to time, then maybe an integrator took care of A/D conversion. However, it does look like the connection tabs are different. (Press Fig.3 and then the ⓘ and then press View full size.) So, until anything is proven to the opposite, I assume this is the 2nd chip.

Fig.2 – 2nd chip, what’s left after all these years floating around in my drawer at work (photo)

I have scanned this and added it in my favourite graphical text editor where I overlaid a circle and found that this was a 3″ wafer (25.4 mm * 3 = 76.2 mm diameter). (It’s Pages for macOS.)

I wish I hadn’t been so careless with it, which used to be a full disk. Or maybe even two..

Fig.3 – 2nd chip (detail) (Long lines every 5 mm, smallest division 0.5 mm)

I wonder how these were cut and bonded at the time. I don’t wonder how they do it today because I think it’s closer to what they do with a magic stick. But then, isn’t it pretty magic that they did this back then, long before the young who today think they are the first to live in the modern, were born?

Fig.4 – 2nd chip’s thinness

It is 0.35 mm thick. Thin, but not compared to the huge 300 millimetre wafers of thinness 40 µm as of 2021 (Infineon). A human hair is 17 – 181 µm according to Wikipedia so I guess it’s ok to say that it is “thinner than a human hair”.

Autronica at the time

Some of the data is from [2] and [3]. See Notes from the vault – 0x03 (I) (password protected)

The stories

I start with the person I first got contact with, and then continue in the same order:

Åsmund Tiller

Engineer at “the lab”. Summary of some mails:

Norwegian English
Jeg har nok ikke så mye å bidra med om selve prosessen. Jeg deltok til en viss grad i planlegging / spesifikasjon og senere i utprøving. Opprinnelig besto de adresserbare detektorene av to kretskort / tykkfilm-substrater. 1. Dette var en produksjonsmessig og logistikkmessig kostbar løsning. Disse to kortene skulle ha kontakt via et sinnrikt fjær-system i sokkel og detektorhode. 2. Dette var en dårlig teknisk løsning som skapte mye driftsproblemer. Utviklingen av “kode-brikken” (ved hjelp av Nordic VLSI) gjorde det mulig å samle all elektronikk på ett kort, noe som løste begge problemer nevnt over. I probably do not have much to contribute about the process itself. I participated to a certain extent in planning / specification and later in testing. Originally, the addressable detectors consisted of two circuit boards with thick film substrates. 1. This was a costly and logistically expensive solution. These two cards should have contact via an ingenious spring system in the base and detector head. 2. This was a poor technical solution that created a lot of operational problems. The development of the “code chip” (using Nordic VLSI) made it possible to gather all electronics on one card, which solved both problems mentioned above.
Her er et eksempel på hva som kunne skje før vi fikk denne forbedringen: Autronica skulle på oppdrag fra Riksantikvaren overvåke et antall stavkirker. Jeg ankom Urnes stavkirke sammen en servicemann for å igangkjøre anlegget. Der ble vi møtt av en stolt installatør som anså seg ferdig med jobben og hadde revet stillasene. Han hadde imidlertid fått noen deler til overs – det var adresseenhetene som skulle sittet i detektorsokkelen. Det var bra ingen bivånet jobben med å klatre opp til alle detektorene for å utbedre mangelen Here is an example of what could happen before we got this improvement: Autronica was to monitor a number of stave churches on behalf of the National Heritage Board. I arrived at the Urnes Stave Church with a serviceman doing the initial start-up. There we were greeted by a proud installer who considered himself finished with the job and had torn down the scaffolding. However, he had some parts left over – it was the address units that were to sit in the detector base. It was good no one attended the job of climbing up to all the detectors to rectify the problem
(Jeg spurte om han husket Jan Meyer, som jeg må ha pratet med i forbindelse med faget jeg tok. Jeg husker at han satt på kontoret sitt på NTH og jobbet med design på en “grafisk skjerm”.) (I asked if he remembered Jan Meyer, whom I must have talked with in connection with the subject I took. I remember him sitting in his office at NTH doing design on a “graphical screen”.)
Jeg husker navnet Jan Meyer, ja og møte i lokaler på Flatåsen. Det var nok ganske tidlig, og komponenten ble i markedsføringen omtalt som den “første kommersielle VLSI i Norge I remember the name Jan Meyer, yes and meeting in premises on Flatåsen. It was probably quite early, and the component was referred to in marketing as the “first commercial VLSI in Norway

Nordic VLSI [4] had their first location at Flatåsen in Trondheim. I also remember going with them at a meeting there at a that time, but I cannot remember what the occasion was.

Pål Fløtten

Developing engineer at “the lab”. Summary of a mail:

Norwegian English
Det var Jan Meyer som digitaliserte den første kodebrikken. Jeg kontrollerte og godkjente tilstandsmaskinen som var resultat av digitaliseringen. Etterpå måtte jeg gå gjennom utlegget i detalj for også å godkjenne det før det gikk videre til prosessering . Det var ingen datastøtte den gang, så det var hardt arbeid. Jeg overså et signal som skulle ha vært invertert i første utlegg, så vi fikk en ekstra runde før vår første brikke var oppe og gikk. Prosjektet var meget vellykket. It was Jan Meyer who digitised the first chip. I checked and approved the state machine that was the result of the digitisation. Afterwards, I had to go through the layout in detail to also approve it before proceeding to processing. There was no computer support at the time, so it was hard work. I ignored a signal that should have been inverted in the first layout, so we got an extra round before our first chip was up and running. The project was very successful.
Meyer holdt til i Industribygget i Innherredsveien i Trondheim, kanskje  sammen med Aaserød. Jeg tror han lånte designutstyr på SINTEF i den tiden. Dette var før Nordic. Nordic oppsto før utvikling av Autronica brikke to. Det skulle bli en røykdetektor, men den ble aldri ble ferdigstilt. Brikken hadde for mange feil som de ikke fant ut av. Det var Ivar Fiskvik som hadde fått den nesten umulige oppgaven å håndtere denne på Autronica. Meyer worked in the Industribygget (The Industrial Building) in Innherredsveien in Trondheim, perhaps together with Aaserød. I think he borrowed design equipment from SINTEF at that time. This was before Nordic. Nordic originated before the development of Autronica chip two. It was supposed to become a smoke detector, but it was never completed. The piece had too many errors that they did not resolve. It was Ivar Fiskvik who had been given the almost impossible task of handling this at Autronica.

Nordic people

18Oct2021: see Notes from the vault – 0x03 (II) – password protected

More

Next note in the series → is also about an Autronica matter (but it’s in Norwegian, translate)

References

Wiki-refsDigital dividers. NTHNTNUSPICE

  1. A piece of Norwegian industrial history, by Autronica Fire and Security. See https://www.autronicafire.com/en/about-us/the-story/
  2. .. “og det var liv laga“. “En beretning om Autronica gjennom 35 år – for ansatte og andre med nær tilknytning til Autronica.” By Harry Amundsen (1992). In Norwegian
  3. .. “og livet går videre“. “En beretning om Autronica gjennom 40 år – for ansatte og andre med nær tilknytning til Autronica.” By Harry Amundsen (1997). In Norwegian. The last page is shown in Notes from the vault – 0x04
  4. Jan Meyer, Oddvar Aaserud, Frank Berntsen and Trond Sæther started Nordic VLSI in 1983 (Nordic Semiconductor since 2004). Started in Trondheim: A real technology adventure. Since the start in Trondheim in 1983, Nordic Semiconductor has grown to almost 500 employees who develop technology we all use in everyday life. See here. The original is in Norwegian: Startet i Trondheim: Et ekte teknologi-eventyr. Siden starten i Trondheim i 1983, har Nordic Semiconductor vokst til nesten 500 ansatte som utvikler teknologi vi alle bruker i hverdagen. Adresseavisen 09Nov2017. See https://www.adressa.no/brandStudio/2016/05/20/Et-ekte-teknologi-eventyr-12753860.ece
  5. Verktøy og vitenskap. Datahistorien ved NTNU av Ola Nordal. Tapir akademisk forlag, Trondheim 2010. ISBN 978-82-519-2610-2. In Norwegian. (“Tools and science. The computer history at NTNU”)
  6. THE SURPRISING STORY OF THE FIRST MICROPROCESSORS. You thought it started with the Intel 4004, but the tale is more complicated. By Ken Sherriff . IEEE Spectrum (Aug2016). See https://spectrum.ieee.org/the-surprising-story-of-the-first-microprocessors