EMS Electronic Music Studios EMS Vocoder 2000 Mixer
EMS Vocoder 2000
Signal Processing Device


The Vocoder 2000 is a signal processing device, which takes two input signals and produces one output. These two signals are known as the 'speech' and the 'excitation'. The output from the Vocoder is a signal which contains a proportion of the harmonic structure of the excitation and the formant structure (the articulation) of the speech. Thus it is possible to make normally inarticulate sounds speak. For instance, if you were to take speech and an Organ, it would be possible to make a 'talking organ'. The heart of the Vocoder is the analysing and synthesizing filter bank„ The speech is analysed into 16 frequency bands which cover the audio spectrum. The time varying energy levels in each channel is extracted by an envelope follower. This is in fact a real time spectrum analysis of the speech. Another signal, the excitation is introduced into the Vocoder. This is the signal that we will make talk. The excitation signal is also analysed into 16 frequency bands throughout the audio spectrum. However, the signal that is presented to each band is multiplied by a control voltage, which is the envelope from the speech channels. Thus the time varying spectrum of the speech is imposed upon the excitation signal; that is the excitation is filtered in a way entirely prescribed by the speech signal. The speech and the excitation signals are connected via mono jack sockets to the input amplifiers. The signal level is adjusted and given a pre-emphasis, a top lift. A switch connects a VU meter to either the speech or the excitation signal. A device called a voiced/unvoiced detector looks at the speech signal and decides whether or not th^ speech is unvoiced ('s' sounds) or voiced (sounds derived from vocal chord vibrations). These voiced/unvoiced decisions can be used to turn on or off the machine's internal oscillator and noise source. Also, the unvoiced decision can be used to alter the external excitation signal so that it becomes more 'S' like. This is done by operating the 'S' generator switch. The Vocoder 2000 has two internal excitation sources, the noise source and the oscillator. The noise source is generated by zener diode and can be used continuously, or gated on by the unvoiced decisions or it can be switched off so that it has no effect. The oscillator produces a pulse waveform which can also be used continuously or gated ON by the voiced decision or it can be switched off. Also the pitch of the oscillator can be manually set as well as the option of being varied by a control voltage generated by the filter bank analysis. The slew freeze section controls the synthesis section of the filter bank. It is possible to freeze or to slew limit the control voltages inside the filter and this will of course have an effect on the Vocoder's output. When frozen, the formant structure remains fixed. When heavily slewed, the output sounds very reverberent. This is because information in the filter bank is time smeared. There are two possible types of time smearing that are available, symmetrical and asymmetrical . The symmetrical option takes as long for a sound to build up as it does to decay, but the asymmetrical gets loaded with a sound structure immediately and then it slowly decays. When you are making a piece of orchestral music talk, the amplitude dynamics of the speech will be imposed upon the output. Thus there will be lots of big gaps in the Vocoder output which might not be very desirable. To overcome this problem we use a device called a pause stuffer. This fills in the silences by fading up the music signal, immediately fading it down when the Vocoder output reappears. Speech or excitation can be used to 'stuff' the pauses and there is a choice of fade in times. The output Mixer is used to select either the Vocoder output or the speech or external excitation signal. It is also possible to control any of these signals with an external swell pedal.


11.80 Mbytes (12370217 Bytes)
1977 05 01
Gescanntes Dokument, teilweise schwer oder teilweise nicht zu lesen.
Upload Datum:
2017 08 16