time stretch analog-to-digital converter
A device for digitizing analog signals of very high bandwidth, effectively slowing down the original signal in time (or compressing its bandwidth) before it can be digitized by a slow electronic ADC.
time stretch ADC: sampling signals faster than your digitizer can handle
A time stretch analog-to-digital converter is a signal processing technique that optically or electronically stretches a fast analog waveform across a longer time window before digitization. The incoming signal, typically with bandwidth in the gigahertz range, is processed through a dispersive medium or time-lens system that maps its frequency content onto time. A conventional ADC then samples this stretched version at rates it can actually handle, usually in the megahertz range. After digitization, the signal is mathematically compressed back to reveal the original high-speed waveform. This allows measurement of transient events and fast signals using standard digitizers that would otherwise be unable to capture them.
The optical implementation uses fiber dispersion or a wavelength-dispersive component to stretch the signal. A short, broadband pulse entering the system emerges as a longer pulse with its frequency information encoded as position along the time axis. A photodiode and conventional sampler then digitizes this stretched optical pulse. Electronic time stretch converters use similar principles with electrical components, often employing a phase modulator and low-pass filtering to achieve the bandwidth compression. Both approaches require precise calibration, as the stretching factor and the mapping between original and stretched frequencies must be known exactly for signal reconstruction.
Where time stretch matters
Time stretch ADCs excel at capturing single-shot, non-repeating events that are too fast for real-time digitization. Telecommunications testing, radar pulse analysis, ultrafast photonics research, and high-energy physics detectors all rely on these converters. A 100 gigahertz signal can be stretched by a factor of 1000, dropping its effective bandwidth to 100 megahertz, which a standard ADC can record at 200 megasample/second. The trade-off is that only a limited window of the original signal fits into the stretched domain; you capture milliseconds of stretched time from nanoseconds of real time.
Practical constraints include the need for precise synchronization and knowledge of when the event arrives. Time stretch depth is limited by the available dispersion, the noise floor of the photodetector or amplifier, and the linearity of the stretching medium. Reconstruction algorithms must account for the specific transfer function of the dispersive element. Temperature drift and component aging degrade the stretching linearity, requiring periodic recalibration. The technique adds latency and cannot be used for real-time control or feedback systems that demand immediate response.
The name reflects the core mechanism: the converter does not directly digitize at high speed. Instead, it stretches time, making fast events appear slow enough for conventional sampling. This is distinct from traditional undersampling or equivalent-time sampling techniques, which rely on repetitive waveforms and phase shifting to reconstruct periodic signals. Time stretch is the tool for one-off transients that occur too quickly and at unpredictable moments for conventional methods.