Techniques for quickly and accurately determining whether a channel is being used for transmission of data using one of a plurality of communications protocols for low power signals using random data of a packet are disclosed. The techniques increase sensitivity and reduce the false alarm rate for a wide range of signal and noise levels. A noise detection technique uses an adaptive window size for fast noise detection that increases the rate of scanning channels during a signal identification period. In a BLE1M detection mode, detection of clusters of zero frequency deviation are used to reduce the false detection rate. Adaptive Zigbee symbol detection improves detection sensitivity beyond −97 dBm. The techniques use a chip-based differential to generate frequency deviation samples for Zigbee detection or data filtering frequency deviation samples generated using sample-based differentials based on an oversampled received signal to improve the signal-to-noise ratio.