Pixelated PET systems produce higher count rates as they integrate several detecting channels per detector module. An increased data flow from the detectors posses higher needs on the bandwidth requirements. We aim to optimize the bandwidth usage efficiency by filtering on the fly the detected events with non valid energies. PET systems with a SiPM-ASIC readout scheme are being extensively used to get enhanced images on Time-Of-Flight PET scanners. These kind of digital readout systems are specially interesting for the application of on-line processing techniques given the ease of access to each detected event digital information. This study purses the analysis of on-line processing techniques on the DAQ front-end level (on-detector electronics) for pixelated PET systems with SiPM-ASIC readout. In particular, we worked with a tunable on-line energy discriminating stage. For the optimization of its hardwired internal limits we analyzed the system energy space. We explored different solutions dependent or not on the system's energy calibration. Results obtained through the different filter versions confirm the minimal resources consumption of such processing techniques implemented at DAQ front-end level. Our experiences showed how the filtering process reduces the bandwidth needs excluding from the data stream all non valid energy events and thus improving the system sensitivity under saturation conditions. Additionally, these experiments highlight how setting proper energy limits we ensure the preservation of the system performance, which maintains its original energy and time resolution. Under the light of these findings, we see a great potential on the application of on-line processing techniques for Time-Of-Flight PET at the DAQ font-end level (on-detector electronics) and so we envisaged more complex processing methods.