A BRIEF HISTORY OF TTT 1965-90 Lars H Zetterberg In the middle of the sixties the scholl of electrical engineering at KTH went into a period of change and expansion. In 1963 the two promiment professors, Erik Löfgren in Radio Technique and Torberm Laurent inTelegraphy andTelephony retired. According to a parliament decision these two professorships were the same year replaced by three new ones, mamely in Applied Electronics, CommunicationTheory and in Com- munication Systems.Also, during the same period the parliament decided on an imcrease of the student emtramce which was part of the general expan- sion of higher education at that time. I was appointed as the first holder of the chair in Commumicatiom Theory in February 1965, amd later that year Lars Kristiansson became assistamt professor (lektor). This was a mew type of position at the umiversities and it was essential in establishing our educational program. Most of my professiomal life I had spent at FOA which made me acquainted with several new techmical topics such as Information Theory amd the beginning of Coding Theory, imitiated by Claude Shamnon at Bell Laboratories. Another strong influ- ence came from Norbert Wiener and S.O. Riceandothers concernimgopti- mal statistical filterimg amd statistical amalysis of detectiom and modulatiom schemes for commumicatioms and radar. My studies of stochastic pro- cesses were stimulated by my comtacts with Ulf Grenander, then at Stock- holm Umiversity, which led to a year of study in USA. The moderm circuit theory 1 learned from Nyquist, Darlington and others. It was very different from what 1 learned from my predecessor "Tobbam" Laurent. Much effort was spent the first years on developing a mew program, and Lars and 1 were given much freedom by the EE section. The result was the four courses, see insert with notation of attendance taken from our yearly report 1968. Looking back at this program it seems very ambitious, and it is sur- prising that students accepted it as they . did without protests. Talking to former students they agree that it was tough, but they liked our program because it was new and the contents seemed meamimgful. Gradually the course program has changed in contents and disposition. Im particular one should mote the introductiom of a course on Signal Processing, startimg 1972. It emerged from our research om the analysis of electroencephalograms and was very much tied to the installment of our computer PDP-11 in 1970171. Almost every year the installation was upgraded amd expanded. A course on Fiberoptical Commun cation Systems was added 1980 and three years later ome on Data Compression.They lasted for a few years, reflecting the research interests at that time. A course on Satellite Communications was given for omly one year. A more persisting change was the replacement ofCircuit Theory by two courses, one in Linear Systems amd one in Analogue and Digital Filters 1981 -82. The teachimg staff included Kris- tiamssom, Mossberg and myself Kristiamsson left 1975-76 to become professor at Chalmers and then Torbjörm Ström took over as "lektor". He was 1983 replaced by Ragnar Ottoson. Later Franz Cech joined the staff, also as "lektor". Most of the research can be arranged under the following headimgs: . information theory and communication systems. . signal theory and signal processing . system theory and circuit technique . digital technique and computer software An impressive number of projects have beem conducted, mostly finamced by STU and then NUTEK. It is only possible to indicate the main lines, along which the research program developed. The start waS made with three subjects: coding for the Gaussiam channel which had beem the subject of my PhD thesis, active RC-filters which remewe8 the subject of circuit theory from my predecessor and thirdly amalysis of electrocardiograms, ECG. Our assistent KAre Mossberg waS im charge of the circuit project amd haS been so with this and similar projects later duringthewhole period. The ECG project waS new for me and for the department, and already in 1966 amalysis of electroencephalo- grams, EEG, waS added with far- reaching consequences. Work om coding for the Gaussian chanmel led to several PhD and tech. lic. degrees of which one concerned group codes for this chamnel. In parallel, work on data transmission over the telephone line began with studies on error statistics, on modu- latiom and systems amalysis but also research on adaptive equalizers and the Viterbi algorithm. During the coming years we notice a project on delta modulation with delayed deci- sion and one on broad band communications that later waS follo- wed by several projects on fiber optic communication. The latter projects were initiated by Lars-Erik Eriksson. In 1981 my former student Jam Uddemfeldt became adjoint professor which stromgly influemced our research activity. Re started a series of projects related to mobile radio ofwhich several comcerned adaptive equalizers but also one initial study of codimg against interferemce.This workwaS dependent on knowledge of system identificatiom amd adaptive filtering, subjects being taught in our graduate courses for many years. The primary motivation waSthem our research om EEG. In the laSt few years, coding for the Gaussian chamnel waS taken up once more, now with transmission baSed on coded modulatiom. Analysis of medical data, im parti- cular of EEG, soon grew into the major project amd engaged half of the personnel, including those in our electronic laboratory when it waS at the peek. Work was done in close cooperatiom with people at the Karo- limska Rospital. The project involved the development of programs for amalysis amd building of several cor- relators for preprocessing of data. It also led to theoretical studies om models for the generation of EEG activity and of epileptic spikes. This gave rise to studies on detectiom of isolated spikes in colored noise, or event detection, a subject later studied aS a gemeral topic. The second extemsive area in signal processing concerned coding of speech and movimg images. Speech waS a good replacement for EEG when funds ceased from STU amd students declared that they preferred studying communications rather than become neurophysiologists. The image codes were imtended for video conferences. Work started by investigations of transform coding and motiom com- pemsation but later, schemes baSed on identification of contours amd seg- mentation ofimages were investigated. An excentric project treated channel equalization by meams of meural nets. The surprising success led to a general study of classification using such networks. Active RC-networks was, aS men- tiomed, the first project on circuits. It wemt on for several years and led to implememtation in integrated form with imdustrial applicatioms. Later also the resistance was removed leading to studies of switched capacitor met- works, a technique that facilitates inte- gration. In addition to these projects, much work was done on developing analysis programs for both analogue and digital networks. The relation between synthesis ofthese two forms of networks was investigated. Finally, the area digital technique was inspired by our experimental work and development of additional units to our computer imstallation. It also led to the development of a microcom- puter, MICROFOX, implemented on circuit boards and with facilities developed for programming. Appli- cation ofspeech coding was studied. Scanning through the yearly reports, 1 find the work done by my research students and coworkers to be impressive. We have graduated 30 of them with PhD exam and 16 with tech. lic. exam. T~ree of my PhD's and four of my tech. lic. graduated ~er my retirement 1990. Much can be said about the social life in our group, let me take up one thread. We ~ad a prominent person in our group, Lars F. Nozkberg, both real and fictitious. He was the author of several course notes on circuit and filter theory.The name came from the four teachers that, during several years, contributed to the motes. The first name Lars was obvious since three of us had that name. Two had the family names ending with -berg and finally we took a few other letters to make Nozkberg. The letter F came by itself Gradually more and more items were added to make our goast author real. For some years he appeared in the telephone directory ofKTH. A place was created where his main items were collected. Our librarian and artist, Gunilla lohnson, added a sculptured head to the holy place, bearing marked resemblance with the figure Mad. One could see young students pay tribute to dr Nozkberg. His most important contribution was to introduce the unit "fnosk" used in dealing with sampled systems and discrete Fourier transforms. It stands for "Frekvens i NOrmerad SKala', fT, with T the sampling interval.