In this paper we analyze the effectiveness of a serial delta compression scheme for reducing data rate of neural signals in brain implanted systems. Due to their location neural implants have limited power budget in order not to damage the surrounding tissues. In an implant with a wireless communication link the data rate needs then to be reduced to meet the power requirements since the transmitter power consumption is often a dominating factor. In this paper we propose and analyze a digital delta compression scheme for multichannel implants which reduces the number of registers and offers more flexibility compared to the canonical scheme. Instead of subtracting from samples belonging to one channel we explore the possibility of using samples from different but correlated channels, this removes the need of storing previous samples for all channels. Synthesized and recorded data sets are used to investigate the effectiveness of the method together with a spike sorting program, to also evaluate the quality of the reconstructed signals.