Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now


All of a flutter

This week Professor Rumsey looks into the various ways of dealing with high speed flutter.

Speed variations of one sort of another, even if small, are a characteristic feature of analogue recording systems. Any variation in speed manifests itself directly as a change in the pitch of the audio signal. Digital recording equipment, by and large, is capable of reclocking audio data so that this doesn’t happen.

Nearly everyone knows what extreme “wow” sounds like – the pitch lurches up and down in a way that creates a queasy effect. (There’s an American phone ringing sound on older movies that seems to do this every time! Or how about “Yes Minister”, where it plagues those classic Ronnie Hazlehurst horns.) While you might experience clear wow on some old recordings, or on LPs with a non-central hole, there can be much more subtle effects of high speed “flutter” that arise from capstan and pinch roller irregularities or the effects of other small rolling or scraping parts. Flutter is less noticeable as a pitch variation, and more as a “roughness” in the sound quality, which results from the creation of frequency modulation sidebands. It can be periodic or more random, in which case the result is more noise-like. A portable tape recorder can even drop gradually in speed as the batteries go flat. Wouldn’t it be great if you could remove these effects and “reclock” old analogue recordings?

Talking to Terry Finn of Affinity Audio, while enjoying Peter Filleul’s booze at the recent AES Convention in London, I remembered my early days as a trainee engineer with Angus McKenzie, when Terry demonstrated Nakamichi’s “Computing Turntable”. A robot arm would pop up out of the depths of the deck and drop a stylus on your LP record to measure the eccentricity of the disk. Then the turntable would shift around a bit to centre the disk, thereby minimising wow and flutter. Remarkable for the early 1980s, but now we have magic software to handle these things. Recently launched by CEDAR is Respeed, and by Celemony is Capstan, both of which aim to track the inherent speed variations in a recording and remove them by relatively simple varispeed replay.

Peter Neubäcker of Celemony suggests that Capstan works by evaluating musical content, so it isn’t suitable for speech or sound effects. (You can add remarkably large amounts of wow to speech, though, before anyone notices, whereas the smallest amount on a sustained horn note is a killer.) With both the CEDAR and Celemony products the amount of smoothing can be adjusted so that natural vibrato in musical instruments is retained, and not treated as wow, for example.

An American company called Plangent Processes, on the other hand, offers a service to “unflutter” recordings based on tracking the residual ultrasonic bias (or logic control) tone on analog tapes, claiming to be able to track and remove very fast flutters of both periodic and random natures. (It was developed with researchers from Cambridge University.) This requires a very high quality transfer so that the ultrasonic information is preserved, whereas products that analyse the audio signal can work on analog copies or on standard quality digital versions, as well as on recordings from disks. Respeed includes the option to give extra weight to a region of the spectrum containing a known reference tone in a recording, such as mains hum or television line frequency, which can act as a more reliable guide to the correct pitch. It stops short of looking for residual bias tones at higher frequencies, though.