I’m still working steadily on the update to SID and have now got to about the middle of the letter S. This means I have been dealing with entries about speech synthesis, spectrographs and the like. Coupled with my mention of the phonetic typewriter in my last post, this has set me off on a reminisce about equipment from the 1970s and earlier. A few months after I was appointed associate research assistant in the Department of Phonetics (as it then was) at UCL, a decision was taken to convert an old laboratory into a room for research assistants. Up until that time I and others had been lurking in corners. I was “persuaded” to help with the removal of some of the old stuff that was there.
In the corner was a speech synthesiser. It was about six feet high (nearly two meters) and about three feet across. The controls consisted of an array of dials. The columns represented time-slices and the rows were frequency bands. The setting on a particular dial controlled the intensity of the signal at that particular time and frequency. If you were very skilled, you could set all the dials and press a button and the machine would produce something which, with a little imagination and goodwill, sounded like a vowel.
There was also a sound spectrograph which sat on its own trolley. Its electronics contained valves! It was extremely slow, but it made the most beautifully clear spectrograms. Well, it did if the operators knew what they were doing. The drum revolved at a leisurely rate and the whole process of making a spectrogram of a maximum 2.4 seconds of speech took about 20 minutes. Then of course one had to measure the result to estimate durations and formant frequencies. Yep. Measure — with a ruler. And then one had to calibrate the machine to allow the measurements to be translated into frequency and time values.
The picture at the top of the post is a wide-band spectrogram of the word reminiscing which I produced sitting at my desk here about 15 minutes ago. It took me approximately 30 seconds to do it.