Analog Optical Link Laboratory Tests (September 1999)

In autumn 1999, we received a prototype of the analog optical link with four channels. It consists of a laser driver optohybrid, which converts electrical to optical signals with a bias programmable over its $\rm I^2C$ interface. On the other end of approximately $100\,\rm m$ of a 4-way optical fiber, a receiver hybrid provides an electrical output.

The bias of the semiconductor laser was selected to achieve the best linearity in the specified input range while minimizing the power consumption. The optical link characteristics were tested by measuring the electrical output while applying a defined input.

Table: Target specifications of the analog optical link compared to HEPHY measurements.

Operational specifications	Min.	Typ.	Max.	Measured
Total length $\rm [m]$	60	100	120	97
Gain $\rm [V/V]$	0.25	0.8	2.5	2.08 (avg.)
Signal-to-noise ratio $\rm [dB]$		48		56.4 (avg.)
Integral linearity deviation $\rm [\%]$		2	4
Bandwidth $\rm [MHz]$	70			110
Settling time to $\pm 1\%$ $\rm [ns]$		18	20	not measured
Skew $\rm [ns]$			2	0.25
Jitter $\rm [ns]$			1	0.077
Crosstalk $\rm [dB]$		-48

As mentioned above, the bias was selected to get a linear input-output relation within the specified input range. The laser threshold can be visualized when exceeding this input range towards lower values. Fig.

shows the input-output characteristics of all four channels with a doubled input span. Towards the left edge, the laser threshold is exhibited, while at high input the receiver begins to saturate. Thus, nonlinearity occurs on either side of the nominal input window.

**Figure:** The input-output characteristics for all four channels of the analog optical link prototype. The nominal input span is in the linear range between $-400\ldots +400\,\rm mV$ .
$\begin{figure}\centerline{\epsfig{file=aol_ioc.eps,height=8cm}} \protect \protect\end{figure}$

Fig.

shows the integral linearity deviation of the four channels, which is defined as the full-scale-normalized error one makes when, for a given link output signal y, the link input signal is assumed to be the linearized value instead of the real value.

**Figure:** The integral linearity deviation of the analog optical link prototype. It remains below $0.5\%$ within the nominal input span, but increases dramatically outside especially below the laser threshold, as shown in the insert.
$\begin{figure}\centerline{\epsfig{file=aol_ild.eps,height=8cm}} \protect \protect\end{figure}$

The transfer function of the analog optical link was measured with an oscillator generating a sine wave of defined amplitude and frequency. The oscillator output was converted to a differential signal and sent into the optical link transmitter. Both input and output amplitudes were measured as a function of frequency, with their ratio defining the gain. From the transfer function (fig.

), a $-3\,\rm dB$ bandwidth of $110\,\rm MHz$ can be extracted. With this frequency response, the effect of transients on the APV output signal is negligible.

**Figure:** Transfer function of the analog optical link prototype. The bandwidth is the frequency where the amplification is $3\,\rm dB$ below the DC gain.
$\begin{figure}\centerline{\epsfig{file=aol_transfer.eps,height=8cm}} \protect \protect\end{figure}$

**Figure:** Noise of the analog optical link extrapolated to $500\,\rm MHz$ bandwidth.
$\begin{figure}\centerline{\epsfig{file=aol_noise.eps,height=9cm}} \protect \protect\end{figure}$

Apart from linearity, the noise figure, which is dominated by the laser contribution, plays an important role for the analog optical link. The output noise has been measured by the VME-ADC with an analog bandwidth $50\,\rm MHz$ and with a digital oscilloscope at several bandwidth limits. As the ADC measurement perfectly fits into the oscilloscope measurements, the ADC values have been scaled to represent the noise obtained at $500\,\rm MHz$ .

Fig.

shows the optical link noise vs. the input voltage, where the ADC noise is already subtracted. At the low edge, the optical output of the laser is zero, and thus the remaining noise (approximately $0.5\,\rm mV$ ) is attributed to the receiver part. When the laser is turned on, its noise slightly increases with rising input voltage. The signal-to-noise ratio given in tab.

has been obtained by dividing the full range input span ( $800\,\rm mV$ ) by the average noise referred to the input.

The noise of the analog optical link has been modelled and systematically studied with respect to device tolerances [70]. The measured noise behavior could be well reproduced by the model.

Apart from timing and crosstalk characteristics, the electrical power consumption of transmitter and receiver sections and the noise contribution in an APV6 system were measured. With the APV6, no difference in noise between copper cable and the optical link prototype could be observed, since the APV6 noise is quite high and the noise of the optical link prototype is lower than that of the final one due to the gain reduction. In the target system, using the APV25 and the final optical link, the latter will contribute approximately $600\,\rm e$ of noise referred to the APV input.