How time-stretch analog-to-digital converter works

Optical frontend for a time-stretch analog-to-digital converter is shown. The original analog signal is modulated over a chirped optical pulse (obtained by dispersing a ultra-short supercontinuum pulse). Second dispersive medium stretches the optical pulse further. At the photodetector (PD) output, stretched replica of original signal is obtained.
The basic operating principle of the TS-ADC is shown in Fig. 1. The time-stretch processor, which is generally an optical frontend, stretches the signal in time. It also divides the signal into multiple segments using a filter, for example a wavelength division multiplexing (WDM) filter, to ensure that the stretched replica of the original analog signal segments do not overlap each other in time after stretching. The time-stretched and slowed down signal segments are then converted into digital samples by slow electronic ADCs. Finally, these samples are collected by a digital signal processor (DSP) and rearranged in a manner such that output data is the digital representation of the original analog signal. Any distortion added to the signal by the time-stretch preprocessor is also removed by the DSP.

An optical frontend is commonly used to accomplish this process of time-stretch, as shown in Fig. 2. An ultrashort optical pulse (typically 100 to 200 femtoseconds long), also called a supercontinuum pulse, which has a broad optical bandwidth, is time-stretched by dispersing it in a highly dispersive medium (such as a dispersion compensating fiber). This process results in (an almost) linear time-to-wavelength mapping in the stretched pulse, because different wavelengths travel at different speeds in the dispersive medium. The obtained pulse is called a chirped pulse as its frequency is changing with time, and it is typically a few nanoseconds long. The analog signal is modulated onto this chirped pulse using an electro-optic intensity modulator. Subsequently, the modulated pulse is stretched further in the second dispersive medium which has much higher dispersion value. Finally, this obtained optical pulse is converted to electrical domain by a photodetector, giving the stretched replica of the original analog signal.

For continuous operation, a train of supercontinuum pulses is used. The chirped pulses arriving at the electro-optic modulator should be wide enough (in time) such that the trailing edge of one pulse overlaps the leading edge of the next pulse. For segmentation, optical filters separate the signal into multiple wavelength channels at the output of the second dispersive medium. For each channel, a separate photodetector and backend electronic ADC is used. Finally the output of these ADCs are passed on to the DSP which generates the desired digital output.