We present a new experimental technique for the separation of dynamic chemical signals based on their frequency domain characteristics. Such a technique can be used to create filters that separate slow signals from fast signals from a common input flow stream. The propagation of time-varying chemical waves through networks of microfluidic channels is first examined. Mathematical models and a set of simple experiments are developed that demonstrate that short microfluidic channels behave as linear delay lines. The observed dispersive broadening and delay behavior can be explained in Fourier space in terms of corresponding phase delay, amplitude decay and characteristic transfer functions. Such delay components can be utilized to implement frequency dependent interference filters. An 8th order PDMS bandpass filter chip demonstrating these ideas was constructed. The filter chip has a central frequency of 0.17 Hz and a bandwith of 0.04 Hz at a flow rate of 4 microL h(-1).