Clarification to Fangas marked ///\\\: ///I was trying to present some applications that would support the theoretical link that is explaining the burst modulation in mathematical terms.
One of those duplicate links is incorrect. I had there something else (I beg your pardon).
The paper indicates:\\f. Pulse Burst Modulation (PBM): the activity value is a function of the number of pulses contained in a relatively short burst:
Ni = Kn x delta(ALPHAi); (7)
where Kn is a proportionality factor which gives the maximum number of pulses in each burst. Within bursts, fB is the peak bit rate.
Quite often, Kn is a low number, therefore PBM heavily discretizes activity values. The e ects of this discretization can often be reduced by using a non-
linear (e.g. logarithmic) PBM [35]. Neither the pulse
rate fB, nor the repetition frequency are relevant.
This modulation (especially in its di erential form, see
Sect. II-B) has recently been used in several \percep-
tive" applications, such as retinas [35], in conjunction
with Event Driven Multiplexing (see Sect. II-C).
In practice, PBM is very similar to PRM and several authors inappropriately call it PRM. The main di erence is that, while the former is more a time-continuous method (in the sense that the average frequency can be latter is clearly an intrinsically time-sampled method.
Bursts of pulses occur only when the input activity is sampled. Yet, the long-term average pulse rate of the two is identical, except for an immaterial multiplicative factor ( fB/fmax). The maximum sampling frequency
of the system is fB/Kn, while Kn is also the maximum
number of pulses in each packet.
///The paper shows one example where the fB is 1MHz. This would be a relatively high frequency packet for the motor control/propulsion.\\\
The Reference [35] is:
Lazzaro, et al., \Silicon Auditory Processors as Computer Peripherals", in IEEE Trans. on Neural Networks", Vol. 4, no. 3, May 1993, pp. 523-528.
