A two-dimensional signal space for bandlimited optical intensity channels

Zhang, Dingchen; Hranilovic, Steve

doi:10.1109/iccw.2015.7247375

Cited by 2 publications

(3 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [34], a Gaussian approximation method to numerically compute P (X a < 0) is proposed. The idea is to approximate the distribution of residual sum R using an Gaussian distribution.…”

Section: Gaussian Approximation Methodsmentioning

confidence: 99%

“…A two-dimensional signal space is presented based on amplitude and derivative modulation for optical IM/DD systems in which all constructed signals are strictly bandlimited [34]. A novel feature of this model is that the strict non-negativity and peak amplitude constraints are relaxed.…”

Section: Contributions Of the Thesismentioning

confidence: 99%

“…In this part, the severity of clipping is quantified by the likelihood that x(t) < 0 and x(t) > pφ 1 under the continuous approximation, where x(t) is the output electrical current signal from a sequence of inputs. In [34], an upper bound on P(x(t) < 0) using Hoeffding's inequality [54] is obtained and a Gaussian approximation is applied for numerical computation of the likelihood of negative amplitude excursion under average optical power constraint. In this thesis, the comparison amongst upper bound on P(x(t) < 0) by applying onesided Cantelli's inequality, Bhattacharyya's inequality, and Hoeffding's inequality are presented.…”

Section: Likelihood Of Clippingmentioning

confidence: 99%

See 2 more Smart Citations

Bandlimited Optical Intensity Modulation Under Average and Peak Power Constraints

Zhang

Hranilovic

2016

IEEE Trans. Commun.

Self Cite

View full text Add to dashboard Cite

Bandlimited optical intensity channels, arising in applications such as indoor infrared communications and visible light communications (VLC), require that all signals satisfy a bandwidth constraint as well as average, peak and non-negative amplitude constraints. However, the signaling designed for conventional radio frequency (RF) electrical channels cannot be applied directly, since they take energy constraints into consideration instead of amplitude constraints. In addition, conventional transmission techniques optimized for broad-band optical channels such as fiber optics, terrestrial/satellite-to-satellite free-space optical (FSO) communications are typically not bandwidth efficient.In this thesis, a two-dimensional signal space for bandlimited optical intensity channels is presented. A novel feature of this model is that the non-negativity and peak amplitude constraints are relaxed. The signal space parameterizes the likelihood of a negative or peak amplitude excursions in the output. Although the intensity channel only supports non-negative amplitudes, the impact of clipping on system performance is shown to be negligible if the likelihood of negative amplitude excursion is small enough. For a given signal space, a tractable approximation approach using a finite series is applied to accurately compute the likelihood of clipping under average and peak optical power constraints. The uncoded asymptotic optical power and iv spectral efficiencies using two-dimensional lattice constellations are computed. The Monte-Carlo (MC) simulation results show that for a given average or peak optical power, schemes designed in the presented signal space haver higher spectral efficiency than M -ary pulse amplitude modulation (PAM) using previously established tech- g 1 (t) first branch receive filterg 2 (t) second branch receive filterh(t) channel response I(t) instantaneous optical intensity n(t) channel noise n 1,i output sample noise after filter g 1 (t) at i-th symbol interval n 2,i output sample noise after filter g 2 (t) at i-th symbol interval n c,1,i clipping distortion after filter g 1 (t) at i-th symbol interval n c,2,i clipping distortion after filter g 2 (t) at i-th symbol interval q variable defined as the ratio of t to symbol period T Q(·) tail probability of the standard normal distribution R residual sum r 1 (t) output waveform of filter g 1 (t)

show abstract

“…In [34], a Gaussian approximation method to numerically compute P (X a < 0) is proposed. The idea is to approximate the distribution of residual sum R using an Gaussian distribution.…”

Section: Gaussian Approximation Methodsmentioning

confidence: 99%

Section: Contributions Of the Thesismentioning

confidence: 99%