Gigabit transmission Fast Track

For future in-car infotainment systems, high-speed communications networks are necessary. Applications like parallel transmission of HD-video content, side and front view cameras, software update of connected components and USB-connected consumer devices will be the main factor for higher bandwidth demands. MOST150 is the latest automotive bus designed for that purpose. However, with increasing information data, even higher data rates in the gigabit per second range are of interest.

In this paper, several solutions for meeting this challenge are described. The research is based on Matlab simulation and takes into account the MOST150 physical layer. This proposal ensures the backward compatibility with today’s already used POF wire harness. 

Enhancing the MOST150 electronic transceivers by introducing 8-level pulse amplitude modulation (8PAM), Reed Solomon (RS) encoding, pre-filtering and adaptive post equalization, or in another approach the zero forcing Tomlinson-Harashima precoding (ZF-THP), gigabit per second transmission over a plastic optical fiber (POF) is achieved.

The cost sensitive automotive market is the reason to investigate the possibility of 2 to 3 Gbit/s transmission based on the already used and proven POF physical layer. Unlike in conventional gigabit optical transmissions, instead of laser diodes (LD) cost-effective LEDs from MOST150 are used as transmitter. The integrated electronic modules, which perform actions like peaking and clamping, are removed from the transceivers in order to be replaced by advanced ones.

Modelling the optical channel

In the MOST physical layer basic specification[1], POF is characterized by a Gaussian low-pass filter. Correspondingly, a 10 to 15 m POF has a 3 dB bandwidth between 95 MHz and 135 MHz. As the LED and the LD have larger bandwidths, the overall optical transmission link can be modelled by a Gaussian low-pass filter with the following electrical transfer function:

«math xmlns=¨http://www.w3.org/1998/Math/MathML¨»«mo»(«/mo»«mn»01«/mn»«mo»)«/mo»«mo»§nbsp;«/mo»«msub»«mi»H«/mi»«mn»0«/mn»«/msub»«mo»=«/mo»«mi»A«/mi»«mo»§#183;«/mo»«msup»«mi»e«/mi»«mrow»«mo»-«/mo»«mn»2«/mn»«mo»(«/mo»«mi»§#960;§#963;f«/mi»«msup»«mo»)«/mo»«mn»2«/mn»«/msup»«/mrow»«/msup»«mo»§#183;«/mo»«msup»«mi»e«/mi»«mrow»«mo»-«/mo»«mi»jx2§#960;f§#964;«/mi»«mo»§#183;«/mo»«msub»«mi»L«/mi»«mi»POF«/mi»«/msub»«/mrow»«/msup»«/math»

And the impulse response:

«math xmlns=¨http://www.w3.org/1998/Math/MathML¨»«mo»(«/mo»«mn»02«/mn»«mo»)«/mo»«mo»§nbsp;«/mo»«msub»«mi»h«/mi»«mn»0«/mn»«/msub»«mfenced»«mi»t«/mi»«/mfenced»«mo»=«/mo»«mfrac»«mi»A«/mi»«msqrt»«mrow»«mn»2«/mn»«mi»§#960;§#963;«/mi»«/mrow»«/msqrt»«/mfrac»«mo»§#183;«/mo»«msup»«mi»e«/mi»«mfrac»«mrow»«mo»-«/mo»«mfenced»«mrow»«mi»t«/mi»«mo»-«/mo»«msub»«mi»§#964;L«/mi»«mi»POF«/mi»«/msub»«/mrow»«/mfenced»«/mrow»«mrow»«mn»2«/mn»«msup»«mi»§#963;«/mi»«mn»2«/mn»«/msup»«/mrow»«/mfrac»«/msup»«/math»
With A the linear fiber loss, σ = 0.132/B the standard deviation, B = 1009 × LPOF - 0.8747 MHz the 3 dB bandwidth, τ = 4.97 × 10-9 s/m and LPOF the fiber length in meter.

Prefiltering and adaptive post equalization

The big gap between the one hundred MHz POF bandwidth and the gigabit transmission speed makes the inter symbol interferences (ISI) in the received signal rather critical. To cope with it, 8PAM is used to reduce the sym-- bol rate by a factor of 3, RS (255, 223) coding is applied to further decrease the bit error rate (BER) from 10-3 to 10-9 , and post adaptive equalizers are used to follow the minor changes in the optical channel. For a bit rate of 3 Gbit/s, a fifth order fractionally spaced high-pass filter is inserted as a pre-filter between the modulator output and the LED.

Zero Forcing Tomlinson Harashima Precoding (ZF-THP)

Using post equalizers can reduce ISI, however it is at the expense of noise enhancement or error propagation. A way out is to move the equalizers to the transmitter side as a pre-filter. However, in this case the dynamic range of the time domain signal is increased, which puts additional burden on the DAC. This drawback can be overcome by proper pre-coding. 

The Tomlinson-Harashima precoding (THP) was investigated, which is a linear pre-coding method to cope with ISI and is capable of stabilizing the inverse channel filter without increasing the dynamic range of transmitted signals[2]. In several aspects it fits well for the applied scheme: first, the POF channel is a slowly time-variant channel, since the in-car environment like temperature or humidity, changes slowly within several consecutive data blocks. Second, THP is particularly suitable for MPAM transmission. Third, the transmitted signals are bound by THP with respect to the dynamic range of LED. The transmission scheme with THP and ZF-FFE is sometimes called ZF-THP.

Signal-to-noise ratios

In this section, the signal-to-noise ratio (SNR) margins right after the photo diode are estimated. This is essential for analytical studies as well as for simulations. In theory, every optical receiver can be modelled as a combination of a photodiode, an input resistor and an amplifier. They contribute noise to the system in the form of quantum noise, diode dark current noise and thermal noise. So the SNR can be estimated as[3]:

«math xmlns=¨http://www.w3.org/1998/Math/MathML¨»«mo»(«/mo»«mn»03«/mn»«mo»)«/mo»«mo»§nbsp;«/mo»«mi»SNR«/mi»«mo»=«/mo»«mfrac»«mrow»«mo»§lt;«/mo»«msub»«msup»«mi»I«/mi»«mn»2«/mn»«/msup»«mi»p«/mi»«/msub»«mo»§gt;«/mo»«/mrow»«mrow»«mo»§lt;«/mo»«msub»«msup»«mi»I«/mi»«mn»2«/mn»«/msup»«mi»shot«/mi»«/msub»«mo»§gt;«/mo»«mo»+«/mo»«mo»§lt;«/mo»«msub»«msup»«mi»I«/mi»«mn»2«/mn»«/msup»«mi»thermal«/mi»«/msub»«mo»§gt;«/mo»«/mrow»«/mfrac»«mo»=«/mo»«mfrac»«msup»«mfenced»«mrow»«msub»«mi»R«/mi»«mn»0«/mn»«/msub»«mo»§#183;«/mo»«mi»P«/mi»«/mrow»«/mfenced»«mn»2«/mn»«/msup»«mrow»«mn»2«/mn»«mi»e«/mi»«mfenced»«mrow»«msub»«mi»R«/mi»«mn»0«/mn»«/msub»«mo»§#183;«/mo»«mi»P«/mi»«mo»+«/mo»«msub»«mn»1«/mn»«mi»dark«/mi»«/msub»«/mrow»«/mfenced»«mo»§#183;«/mo»«mi»B«/mi»«mo»+«/mo»«mfrac»«mrow»«mn»4«/mn»«mi»kTB«/mi»«/mrow»«mi»R«/mi»«/mfrac»«/mrow»«/mfrac»«/math»

With Ip the photodiode current, Idark the dark noise current (300 nA in this case), Ithermal the thermal noise current, R0 the spectral sensitivity (0.35 A/W), P the received optical power, e the elementary charge, B the receiver bandwidth, b the Boltzmann constant, T the absolute temperature and R the resistance.

As the minimum and maximum received optical power at SP3 according to[1] is -2 dBm and -22 dBm respectively, the SNR range can be estimated. As a result, the SNR ranges from 20.4 ~ 59.5 dB without a pre-filter (B = 135 MHz), and from 18.4 ~ 57.6 dB with a pre-filter (B = 210 MHz), with R = 50 Ω and T = 300 K. For the simulation, AWGN is generated with the power defined by the receiver SNR, which is bound by 20.4 and 59.5 dB without a pre-filter (B = 135 MHz), and by 18.4 and 57.6 dB with a pre-filter (B = 210 MHz). After equalization, the data sequence is decoded by the RS decoder. The result is compared with the transmitted bit sequence to get the BER. The target BER is 4 × 10-3, so RS(255,223) coding is able to decrease the BER to 10-9.