Table of Contents
5 Experimental Verification ... 2
5.1 Introduction ... 2
5.2 Experiment Setup ... 3
5.3 Overview of the Instruments used in the experiment ... 4
5.3.1 Multi-‐channel Tunable Lasers ECL-‐210 ... 4
5.3.2 Agilent ParBert 81250 Overview ... 6
5.3.3 EOSpace Mach-‐Zehnder Modulator ... 11
5.3.4 SHF 11100 A Error Analyzer ... 15
5.4 BER vs OSNR ... 17
5.5 Conclusions ... 21
Chapter 5
5 Experimental Verification
5.1 Introduction
In this chapter we report some experimental results obtained in the laboratory. The goal
of this experimental study was to verify the correctness of the simulation results and
conclusions presented in the previous chapters. In particular we verified the
effectiveness and proper functioning of the monitoring technique in the case of a LS
modulated with OOK modulation format at 10 Gb/s and a supervisory channel with
modulation data rate ~ 20 Mb/s. We also verified the need of applying DC-balanced
line-coding (i.e. 8B/10B or 9B/10B) to properly receive the monitor signal, which
otherwise would be totally corrupted by the interference caused by the low-frequency
components of the line signal (LS).
5.2 Experiment Setup
Figure 5.1
depicts the experimental arrangement we used to demonstrate the Monitoring
Technique discussed in the previous chapters. .
Figure 5.1 Experimental Arrangement
By using the instrument Agilent E8401 A, we generated a 8B/10B coded sequence and a 2
7-1 PRBS sequence: the first one was applied at the input of the first Mach-Zehnder Modulator to generate the Line Signal at 10 Gb/s; the second one was used for the Monitor Signal. With the Multi-channel Tunable Laser ECL-200/210, we generated the optical carrier to be applied at the input of the first modulator. By sending the Line Signal to the input of the second modulator, driven by the PRBS sequence, the Overmodulated Signal is then generated and available at the optical output of the second Mach-Zehnder.
The Overmodulated Signal was then amplified by an EDFA and sent to the optical receiver composed by an optical filter (centered at the laser wavelength) and an O/E converter with 10 GHz of electrical bandwidth. The BER of the Line Signal was measured by connecting the output of the O/E converter to an Error Analyzer SHF 11100 A; the BER of the Monitor Signal was instead measured by the same instrument used for generating the binary sequences: in this case, an electrical filter with 3-dB
Francesco Giannone 6/5/14 5:03 PM Commenta [1]: Ce la fai a fare una foto decente!??! Io ne ho fatte un po ma ero di fretta e fanno tutte schifo.
Roberto 6/5/14 6:30 PM
Commenta [2]: La foto va bene – sei te che l’hai messa in 3D e fa venire il mal di testa – cmq provero’ a farne altre
Roberto 6/5/14 6:35 PM
Commenta [3]: Devo riscriverlo tutto…
bandwidth of ~ 20MHz was placed at the output of the O/E converter to filter out the unwanted components at higher frequencies belonging to the LS, so that the MS could be received properly.
In the following sections of this chapter we first explain the instrument used during the experiment and then the results of the experiment will be presented.
5.3 Overview of the Instruments used in the experiment
5.3.1 Multi-‐channel Tunable Lasers ECL-‐210
The ECL tunable laser series provides a rack-mounted, multi-channel solution that caters specifically to the needs of both DWDM production and research engineers.
Figure 5.2 The ECL-210
The ECL-210 (
Figure 5.2 The ECL-210Figure 5.2) offers excellent long-term wavelength and
power stability in conjunction with notably high output power. Each unit has a wide
tuning range of more than 100 nm which is selectable in the ranges 1270 ~ 1350 nm and
1430 ~ 1630 nm. In addition, Santec's unique tuning mechanism enables any wavelength
within the specified tuning range to be easily and accurately selected. This laser share
many standard features that include Automatic Power Control, fine-tuning wavelength control, coherence control, and a GPIB/RS- 232C interface with drivers for LabVIEWTM and Visual BasicTM. The ECL-210 offers complete electronic wavelength tuning (coarse and fine) which is fully controllable using remote communications. In addition, ECL-210 builts in an optical variable attenuator (OVA) as standard. The principle of operation is shown in
Figure 5.3.
Figure 5.3 Principle of operation
Table 5-1
shows the laser specifications.
Category Parameter Unit ECL-210
Wavelength Characteristics Tuning Range nm 1530 to 1630
100
Resolution nm <0.024
Accuracy nm < ± 0.1
Repeatability nm < ± 0.1
Stability nm < ± 0.1
Fine Tuning Range GHz 200
Power
Output Power mW >8(Peak)
Accuracy % 5
Repeatability dB < ± 0.1
Stability dB < ±0.1
APC Flatness dB < ±0.2
Built in Attenuator Option dB 0 to 20
Environmental Conditions
Operating Temp. Range °C 15 to 35
Operating Humidity Range % <80
Storage Temp. Range °C 10 to 40
Storage Humidity Range % <80
Recommendation Calibration Period Year 1 Spectrum
Spectrum Line Width (Coh. OFF)
MHz <0.2 Spectrum Line Width (Coh. ON) 0.2 to 200
SSR dB >50
RIN dB >145
Interface
Optical Connector - FC or SC
Optical Fiber - SMF or PMF
Connector Polish - SPC or APC
GP-IB - None
RS-232-C - Yes
Modulation LF Modulation KHz DC-10
(RF Modulation Option) MHz 1-100
Dimensions Width x Height x Depth Mm 35x85x350
Weight kg 0.5
Table 5-1 ECL-210 Specifications
5.3.2 Agilent ParBert 81250 Overview
The modular ParBERT 81250 (
Figure 5.4) can be tailored to individual
to 132 synchronous channels. Different modules are available for the ParBERT 81250
System that cover data generation and analysis from 333 kb/s up to 13.5 Gb/s.
Figure 5.4 ParBERT 81250
The ParBERT 81250 consists of the user SW and the ParBERT modules. These can be categorized by their maximum data rate (675Mb/s, 3.35Gb/s and 7/13.5Gb/s) and their functionality (clock or data). A minimum system consists of one clock module and one data module forming a so-called clock group which is installed in the ParBERT VXI- mainframe. The mainframe can hold multiple clock groups; each is operated through its own graphical user interface (GUI).
5.3.2.1 ParBERT 81250 Generation Software
A common test pattern for serial (communication) links is a Pseudo Random Binary Sequence PRBS. To simplify Mux/DeMux testing ParBERT provides the so-called Pseudo-Random-Word Sequences (PRWS) for generation and as expected data for analysis. A PRWS consists of a PRBS per lane with a lane-to-lane delay chosen such that, when muxed together, the same PRBS-polynomial as on each lane of the parallel side is generated on the serial side.
The appropriate lane-to-lane delay is determined by the number of lanes (see figure 3).
In addition, ParBERT’s pattern memory enables user defined test patterns as well.
Physical layer testing includes the establishment or termination of a connection or the set-up of a DUT into a specific test mode, e.g. a loop-back of received data.
Furthermore, specific device parameters often correlate with suitable test patterns. To perform such complex tests in one shot without interruption to download different test patterns, the ParBERT 81250 is equipped with a powerful pattern sequencer allowing up to five nested loops and branching on internal and external events or upon SW command. Set-up is made easy through the powerful ParBERT 81250 graphical sequence editor, which also aids you maneuvering around HW restrictions when setting up user patterns not directly matching selected block lengths. Debugging
of the test sequence is supported e.g. by highlighting the currently executed block within the sequence (see figure 5). Combining sequencing and precise timing with internal or external generator channel-add allows realization of e.g. user defined multi-level or de- emphasized signals. The latency between DUT input and output is often not exactly known or not even deterministic. Having to set ParBERT’s sample point for analysis manually would be tedious. Therefore ParBERT 81250 analyzers provide three modes to automatically synchronize and align upon expected data with BER below a user defined threshold as a criteria:
• AutoBitSync: Finds the proper bit position in a data stream. For memory based patterns a unique detect word is required. Automated phase alignment is optional.
• Fast Bit Sync: Possible for PRBS/PRWS only. Especially useful for burst-mode applications, e.g. optical re-circulating loops.
• AutoDelayAlign: Only the sample point timing of the analyzers is adjusted. The latency between input and output must be within the delay range of the analyzers used.
For applications w/o expected data with ParBERT running in data acquisition mode, e.g.
for A/D test, proper sample point adjustment can automatically be achieved with the
CDR / lane mode (7 and 13Gb/s analyzer modules only). The analyzer sample timing
can be adjusted ±1 period while the instrument keeps running w/o interrupting the measurement. For the 13.5 Gb/s, 7 Gb/s and 3.35 Gb/s the delay of the generator modules can be adjusted ±1 period while running as well.
The architecture of classes in one clock group. Choosing the right combination of system data rate and binary frequency multipliers (..., 1/16, 1/8, 1/4, 1/2, 1 , 2 ,4 ,8 , 16, ...) enables each module to operate within its valid data rate range. Furthermore even channels within one module can operate at different data rates of binary ratio. ParBERT can be configured with one or more clock groups each controlled via an independent instance of the GUI.
The clock groups can either run completely independent from each other or can be locked to each other in a frequency ratio of m/n, with m, n=1...256 and m x n<1024.
Receiver margining is a measurement that is very demanding in terms of signal conditioning capabilities of the stimulating pattern generator, because it shall emulate worst case conditions as they may appear in mission mode of the DUT or as they are specified in relevant standards. For this purpose not only voltage levels and cross point of differential signals shall be adjustable. Jitter shall be injected as well. For this purpose the higher data rate modules of ParBERT, i.e. the 3.35 Gb/s, 7Gb/s and 13.5 Gb/s generator modules are equipped with a delay control input, that allows phase modulation of the data output signals equivalent to the applied voltage signal.
Modern computer standards use spread spectrum clocking (SSC) techniques to decrease the power density of radiated emissions per frequency. The 13.5 GHz clock module allows direct feed-through of such a multi-UI low frequency modulated external clock enabling the data modules to generate and analyze such data patterns with SSC.
Agilent's signal generators can be used to generate such a modulated clock.
5.3.2.2 ParBERT 81250 Measurement Software
ParBERT Measurement software consists of six different measurements graphical and numerical results allowing in- depth characterization for use in R&D and Device Verification (DV). Fast pass/fail tests against user definable limits for manufacturing purposes are provided as well. ParBERT, as any other BERT, physically can only do one measurement: it digitizes the input signal with respect to a predefined threshold voltage and, at a predefined portion of the bit width (the sample point), compares this to its expected data and counts an error if it doesn’t match the expected binary value. ParBERT Measurement Software automates the variation of these mentioned parameters (sample point timing and threshold voltage) and the repetition of this measurement for a number of subsequent bits and by this allows performing a variety of measurements:
• BER measurement: During the bit error ratio (BER) measurement sample point timing and threshold voltage are kept constant, usually at the optimum values in the middle of the TX’s output eye. Compared to the BER-measurement of the regular the BER, as actual and accumulated values (for last measurement timeframe and since start of measurement), the BER measurement of the MUI delivers additional information and measurement control: Errors for expected ones and zeroes are counted and reported separately. Repetitive and single shot measurements can be set-up along with error counting, run- mode options and stop criteria such as logging, automatic resynchronization or stop after a specified number of errors, bits or seconds. These features enable R&D usage for root cause failure analysis (e. g. BER log during a temp cycle), as well as manufacturing with minimum sample sizes and short measurement times.
• DUT output level measurement;
• Eye opening;
• DUT output timing measurement;
• Spectral decomposition of jitter;
• Fast eye mask measurement;
5.3.2.3 Linecards used in our experiment
Figure 5.5 The two
modules used in theexperiment.
In our experiment we use two different versions of the Agilent ParBert 81250 Front-Ends (
Figure 5.5).
For the Line Signal we use the Agilent N4872A ParBERT Generator Front-End for generate the 8B/10B-coded sequence. The Line Signal BER measurement was done with the SHF 11100 A Error Analyzer (see SHF 11100 A Error Analyzer ).
The generation of the PRBS, through which we implemented the Monitor Signal, was done using the Agilent E4838A ParBERT Generator Front-End; the BER measurement of the MS was made using the Agilent E4835A ParBERT Analyzer Front-End.
5.3.3 EOSpace Mach-‐Zehnder Modulator
Figure 5.6 EOSpaceMach-Zehnder Modulator
A low-loss and wideband modulator (
Figure 5.6) for chirp control or coherent optical applications.
The Specifications are shown below (
Table 5-2).
Ultra-Low V
πExtra BW Ultra-Low Loss Insertion Loss <4 dB
(<3 dB opt)
<4 dB (<3 dB opt)
<2 dB (<3 dB opt) 3 dB Bandwidth >10, >20 GHz >30 GHz >10, >20 GHz Modulation Port V
π(@1 GHz) < 3 volts <4 volts <5 volts
S11 <-10 dB
Optical Return Loss >50 dB
Package Dimensions 3.48”x0.35”x0.35”
RF Connector Female K (<20 GHz), female V (20-40 GHz) Operating Wavelength 1.55 micron (C- and L-band)
Input Fiber PM
Output Fiber SM or PM
Optical Connectors FC/UPC standard
Options
Ultra-low insertion loss Custom ultra-low V
πIntegrated Polarizer Integrated Attenuator (VOA)
Polarization Scrambler
Table 5-2 Mach-Zehnder Modulator Specifications
The right way to use and to connection this modulator is shown in
Figure 5.7.
Figure 5.7 Modulator Connections.
Figure 5.8
shows the eye diagram at the output of the first Mach-Zehnder Modulator (Line
Signal), while
Figure 5.9shows the eye diagram at the output of the second modulator
when the Line Signal is not active (Monitor Signal). Finally,
Figure 5.10shows the eye
diagram at output of the second modulator when the Line Signal is active (Overmodulated Signal).
Figure 5.8 Line Signal Eye-Diagram. Time scale on horizonatal axis is 20 ps/div.
Figure 5.9 Monitor Signal Eye-Diagram. Time scale on horizonatal axis is 8 ns/div.
Figure 5.10 Overmodulated Signal Eye-Diagram. Time scale on horizonatal axis is 20 ps/div.
5.3.4 SHF 11100 A Error Analyzer
The SHF 11100 A (
Figure 5.11) is an error analyzer plug-in which can be fitted into any the SHF 10000 Series mainframes. It has broadband operation from 1.5 to 56 Gbps and features high sensitivity and a wide clock phase margin.
Figure 5.11 SHF 11100 A
It allows the analysis of PRBS signals with pattern lengths of 2
7-1, 2
9-1, 2
11-1, 2
15-1,
2
20-1, 2
23-1, 2
31-1. User patterns can also be analyzed; in fact for the BER estimation of
the Line Signal we imported the 8B/10B coded sequence in the Error Analyzer and
connecting the data output and the clock output of the Agilent Pattern Generator at the
data input and clock input of the SHF 11100 A, the Error Analyzer self synchronized and measured the BER.
Figure 5.12 SHF 11100 A intuitive software interface
The specifications are shown in Errore. L'origine riferimento non è stata trovata..
Table 5-3. SHF 11100 A Specifications
Parameter Unit Min. Typ. Max.
Data Input
Bit Rate Gbps 6 50
S11 dB -10
Sensitivity mV 25 50
Clock Phase Margin ° 200
Threshold adjustment mV -300 300
5.4 BER vs OSNR
It is common practice to measure the BER of a system as function of the OSNR at the RX. To do so, it is necessary to perform the noise loading operation in order to vary the OSNR. To do that, we used an Optical Amplifier with any signal applied to its input port. In this way, the amplifier generates a broadband optical noise (ASE). By coupling the amplifier output with the overmodulated line signal and changing the amplifier pump laser current, it is possible to adjust the power of the optical noise and therefore the OSNR at the receiver input.
Figure 5.13,
Figure 5.14and
Figure 5.15show different OSNR values measured by an optical spectrum analyzer.
Clock Input
Frequency GHz 3 25
Input level dBm 0 4
Phase Adjustment Ps 0 160
Clock outputs
Frequency GHz
GHZ
3 3
50 (full clock) 25 (half clock)
Output level mV 300 600
S11 dB -10
System
Data Patterns 2
7-1, 2
9-1, 2
11-1, 2
15-1, 2
20-1, 2
23-1, 2
31-1;
User-programmable pattern Mbit 128
Back to Back Q-factor Linear 25 30
dB 28 30
Figure 5.13 Optical Signal-to-Noise Ratio equal to 26.326 dB
Figure 5.14 Optical Signal-to-Noise Ratio equal to 16.190 dB
Figure 5.15 Optical Signal-to-Noise Ratio equal to 5.152 dB
The BER of the Line Signal is shown in
Figure 5.16. The blue line shows the experimental
BER of the LS whit the Monitor Signal and the green line shows the experimental BER of the
LS without the MS. As we can see, the overmodulation introduce very limited power penalty
due to the low modulation index used for the 20 Mb/s OOK.
Figure 5.16 Experimental BER of Line Signal
the blue line shows the experimental BER of the LS with the Monitor Signal and the green line shows the experimental BER of the LS without the MS.
Shows the experimental BER of the Monitor Signal.
5.5 Conclusions
We demonstrated the proper functioning of the proposed technique by conducting a real experiment in the laboratory for a line signal OOK modulated at 10Gb/s and a supervisory channel at 20 Mb/s. In particular, we demonstrated the critical aspect related to the need of DC-balanced line-coding techniques (in case of signals with intensity modulation) in order to preserve the information carried by the monitor signal, which would be otherwise completely destroyed by the low-frequency content of the line signal. Due to the limited amount of time, we could not verified all the other simulation results obtained for different modulation formats and signal impairments. However, the experimental verification carried out was enough to verify the working principle, giving us confidence that all the other simulation results are correct and could be reproduced in
Roberto 6/13/14 4:57 PM
Commenta [4]: Ir problema e’ che sta curva fa venire il mal di testa
Francesco Giannone 6/5/14 5:03 PM Commenta [5]: Se ce la fai qui ci andrebbe la ber del monitor.
