When examining overload recovery in AC-coupled amplifiers, or for testing the attack/release times in audio compressors, it is useful to be able to feed in a tone burst of a defined duration and amplitude. This simple circuit accepts a steady tone from an ordinary audio oscillator at its input, and produces repeating bursts at its output. The duration of the burst, and the repetition rate, are variable from less than a millisecond to several seconds. The amplitude and switching point are also adjustable.
 
The input signal divides between two paths; one goes straight into buffer U1a, the other goes via a level pot into buffer U1b. The level pot P1 allows you to adjust the low-level signal amplitude during the ‘off time’. The low-level output from U1b goes straight to the output buffer U1c via a 10k series resistor, while the output of U1a goes via JFET Q1 which serves as an analog switch. You need a JFET with the smallest possible Vgs(off) rating, like -1.5V or smaller. The J201 or J113 should work, and many J112s will too. When the gate of the FET is pulled down to ground it is ‘off’ and no audio can pass through it, so the output of the circuit is only the low-level audio. When the gate is pulled up by R3 the FET is ‘on’, so the high-level signal can pass through and overrides the other smaller signal. The 100pF capacitor helps to suppress switching transients. The rest of the circuit is the side-chain that controls when to switch the FET on and off.

The high-level signal from U1a is fed to U1d which works as a comparator, producing a square-wave output every time the input signal pass through the zero-crossing. The trim pot P4 allows this trigger point to be finely adjusted to remove any discontinuity from the start of the burst.

U2 is a dual monostable, a bit like a pair of 555 timers each dedicated to monostable operation, i.e. they each produce a pulse of a specific duration when triggered by a rising-edge input. When the output of U1d first goes high it triggers U2a which produces a short pulse that switches the FET on. In other words, the length of this pulse is the burst period or ‘on time’. This can be adjusted from about 450µs to 45ms or from 4.5ms to 450ms when the 1u timing capacitor C8 is switched into the circuit. It is also sent to the trigger output which you plug into your oscilloscope trigger input; this gives the oscilloscope a clean signal to lock onto so the audio signal doesn’t flicker on screen.

The pulse from U2a is also used to trigger U2b which produces a longer pulse. This long pulse is immediately fed back to U2a which prevents U2a from triggering again until the long pulse has ended. In other words, the long pulse determines the repetition period of the burst generator. It can be adjusted from about 4.5ms to 450ms or from 45ms to 4.5s when the larger 10u timing capacitor C9 is switched into the circuit. The timing capacitors should ideally be tanatalum for best stability, although the circuit functions well enough even with ordinary electrolytics. When adjusting the controls, the start of the burst always jumps between zero crossings (use the trim pot to remove discontinuity) but the end of the burst is smoothly variable and can be set anywhere to eliminate discontinuity.

With a 9V supply the circuit will accept input signals up to 5Vpp before clipping, or the supply can be increased up to 15V for more headroom. Operation is reliable up to 20kHz with input signals as small as 400mVpp. With smaller signals the maximum frequency reduces somewhat. Current consumption is <4mA.
I chose to arrange the power supply to accept a centre-negative DC jack –a standard adopted from guitar pedals, but this is of course optional.