optical communication simulation

Bo-Ning Hu, LianHua, Rui-Mei Zhao and Hua-Wei Pang, "Computer simulation of optical fiber communication system," 2009 International Conference on Machine Learning and Cybernetics, Baoding, China, 2009, pp. 2509-2512

• Simulation domains:
• time domain
• frequency domain
• EMF: electro-magnetic field
• Nice 4 page paper with an overview of how to simulate optical fiber systems.  Primarily concerned with long-haul type fiber impairments.

Y. Hwang, D. Choi, H. An, S. Shin and C. G. Lee, "Development of Python-MATLAB Interface Program for Optical Communication System Simulation," 2019 International Conference on Green and Human Information Technology (ICGHIT), Kuala Lumpur, Malaysia, 2019, pp. 46-48

• Free space optics.

Modeling nonlinear CTLE

 A. Zargaran-Yazd and S. Sudhakaran, "Using deep neural networks to model nonlinear circuit blocks in wireline links," 2017 IEEE 26th Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS), San Jose, CA, USA, 2017, pp. 1-3

Observations:

• 7 cites in other papers.  Would be interesting to investigate.
• DNN vs Volterra-series
• Interesting but I wonder how well the DNN would work when varied across data rate and sample interval.
• This approach could have merit as part of a knowledge-based-NN where the NN cleans up after traditional modeling approach of pole/zero + memoryless-nonlinearity is done.
• Might be an interesting intern project to replicate this work.

Coherent Data Center Links

 J. K. Perin, A. Shastri and J. M. Kahn, "Coherent Data Center Links," in Journal of Lightwave Technology, vol. 39, no. 3, pp. 730-741, 1 Feb.1, 2021, doi: 10.1109/JLT.2020.3043951.

This was a great paper that opened my eyes to the use of coherent optical technology for data centers. What follows is the paper outline with selected sentences.

1. Introduction
• "Conventional coherent detection with polarization multiplexing and a strong local oscillator (LO) recovers phase and magnitude in each polarization, thus utilizing all four degrees of freedom of the optical channel, while maximizing optical power efficiency.
2. Optical Detection Methods
• In coherent detection, the decision variables are the electric field quadratures in each polarization, which can be measured after homodyne (or heterodyne) downconversion or estimated from an intensity measurement, in the case of KK detection.
• KK = Kramers-Kronig
1. LO-Free vs LO-Based Downconversion
• Traditional coherent detection employs homodyne LO-based downconversion, whereby the received signal is mixed with a LO laser whose frequency is approximately equal to the transmitter laser frequency. Note that heterodyne downconversion, whereby the LO and transmitter laser frequencies differ by an intermediate frequency, achieves the same performance as homodyne downconversion in most practical scenarios [7], but requires higher receiver bandwidth, since the downconverted signal is centered around an intermediate frequency.
• Although an LO increases the receiver complexity, LO-based downconversion offers two key advantages compared to LO-free downconversion. First, the LO amplification gain can boost receiver sensitivity by up to 20 dB in unamplified links [12]. This gain is not critical in amplified links, where the receiver is limited by amplified spontaneous emission (ASE) noise rather than shot noise. Second, the LO provides an absolute phase reference, allowing for full recovery of amplitude and phase components of the electric field. 
2. Impact of Co-Packaged Optics
3. Typical Embodiment
1. Direct Detection
• Noncoherent detection of OOK has been traditionally used in data center links, given its low cost and high receiver sensitivity.
• Level spacing optimization improves the receiver sensitivity by roughly 1 dB for APD-based receivers Fig. 1a, while in amplified systems Fig. 1b, it results in ∼3-dB OSNR improvement.
• APD= avalanche photo diodes
• RJA note: I believe that this means that the symbol voltage levels are not symmetric to take advantage of the different noise characteristics of the different symbol levels.
• CD mitigation through linear equalization is only effective when CD is small.
• CD=chromatic dispersion
2. Differentially Coherent Detection
3. DSP-Based LO-Based Coherent Detection
• Coherent detection based on high-speed DSP is a mature technology in long-haul systems, but it may be currently unsuitable for data center links, where cost and power consumption are paramount.
• The most power-hungry operations are CD equalization and polarization demultiplexing with PMD compensation, which together amount to roughly 55% of the receiver power consumption.
• PMD = polarization mode dispersion
• Short-reach links in data centers are subject to lower CD and negligible PMD, hence both CD and PMD filters can be simplified.
4. DSP-Free LO-Based Coherent Detection
• A coherent receiver must perform three basic operations: polarization demultiplexing, carrier recovery, and timing recovery. In traditional long-haul coherent receivers all these operations have been performed digitally.
• Three cascaded phase shifter pairs can perform any arbitrary polarization rotation.
• The performance shown in Fig. 3 for DSP-based coherent receivers is independent of dispersion, as CD is compensated effectively by digital equalization.
• The performance gap between DSP-free and DSP-based receivers can be reduced by leveraging analog equalization as proposed in [4]. 
• Although all-analog equalization does not offer the same flexibility or near-optimal performance of digital equalization, it requires substantially lower power since all the operations are still performed in the analog domain, thus avoiding high-speed ADCs and DSP.
5. Kramers-Kronig Coherent Detection
• The KK receiver performs simple direct detection and digitally recovers the electric field quadratures from a signal intensity measurement
• note: KK utilizes the Hilbert transform to recover the phase from the complex envelope.
• the phase ϕ(t) can only be uniquely recovered as long as the signal E(t) satisfies a minimum-phase condition
• The KK receiver enables coherent transmission without requiring a LO, significantly reducing the complexity of receiver optics. However, the additional optical power of transmitting a carrier signal along with data significantly reduces reach. Higher sampling rates, along with strict optical filter requirements, further increase the complexity of the receiver.
4. Complexity and Power Consumption Consideration
• DSP-free coherent solutions have similar, or more, optical complexity, but may offer lower power consumption by removing power-hungry ADCs and DSP.
• it is likely that that for the next few years, low-power direct detection of OOK and 4-PAM will continue to dominate the shortest links (<2 km) within data centers. 
• At the other extreme, links beyond 40 km will continue to rely on DSP-based coherent detection, as it offers the highest receiver sensitivity, highest throughput, and it is more robust to transmission impairments. 
• The intermediate distances from 2 km to 40 km still require a low-power alternative that enables more degrees of freedom than direct detection, while improving receiver sensitivity. So far, DSP-free coherent detection is the detection method that best meets these requirements. DSP-based coherent detection might become attractive over those distances is reduced by innovative designs and CMOS technology improvements.
5. Conclusion

OIF 5.1 JTOL notes

 I love how OIF makes their specs open to the public and really quite useful.

https://www.oiforum.com/wp-content/uploads/OIF-CEI-5.1.pdf

• Section 2: Jitter and Interoperability Methodology
• Method A uses a CID patter, The other methods (B, C, D and E) use a free running PRBS31.
• Patterns with CID, (Consecutive Identical Digits) is a straightforward way to test system capability
• 
• Total and Relative Wander Masks: 
• 
  
  
  
  
• Jitter Transfer:
• 
  
  
  
  
• Interesting how jitter transfer is calculated.
• CEI 56G MR PAM4 Medium Reach Interface
• 
  
  
  
  
  
  
  
• For this interface the buad rate is in the range of [18 , 29] Gsym/s (from table 17.6)
• CRU = clock recovery unit
• The spec says that you need to test compliance at 5 points indicated by the blue x's above.

JTOL Resources

 JTOL Application Notes

• Microchip
• https://ww1.microchip.com/downloads/en/Appnotes/en567015.pdf
• USB3
• Calibration of jitter sources is critical
• Anritsu
• MP1900A
• https://dl.cdn-anritsu.com/en-en/test-measurement/files/Application-Notes/Application-Note/mp1900a-jitter-tolerance-ef1100.pdf
• (1) jitter tolerance testing isn't just SJ
• (2.1)  Sinusoidal Jitter
• "When the input Data signal jitter frequency and amount are within the Loop Bandwidth, an error does not occur due to a phase mismatch because recovered clock tracks input Data signal in terms of jitter"
• "jitter tolerance test for PCI Express, specifies applying two SJs"
• More than one sj's at a time!
• (2.2) Random Jitter
• "For RJ, the CEI 3.0 jitter tolerance test standardizes use of a High Pass Filter (HPF) to remove RJ components within the PLL band as shown below and applies only PLL out-of-band components as the load"
• The MP1900A scope can apply different amounts of RJ in different frequency bands!
• (2.3) BUJ - Bounded Uncorrelated Jitter
• crosstalk!
• (2.4) Half Period Jitter (F/2 Jitter) or HPJ
• due to half rate clock
• (3) Tips for Jitter Tolerance Measurement
• (3.2) "At jitter tolerance measurement, the amount of jitter is changed until the designated error condition is detected by searching. There are two search methods: Downwards from large to small values, and Upwards from small to large values. The search steps for both the Downwards and Upwards searches can be specified as either Linear using a fixed value, or as Logarithmic using a value changing at a fixed ratio."
• Interesting that both up and down searches are available
• Interesting that both linear and Logarithmic steps are also available.  
• Tek
• https://download.tek.com/manual/PCIe_Gen4_Base_Rx_Help_EN-US_077-1773-02_077177302.pdf
• 
  
  
  
• Interesting how different frequency ranges uses a different step size.
• Search Algorithms options, {Binary,  Downwards Linear,  Downwards Log,  Upwards Log,  Upwards Linear, Binary + Linear}
• 

  
  
  
• It would be helpful to plot the JTOL performance as the data comes in.

Silicon Photonics and Linear Drive (Non-Retimed) Electric-Optic-Electrical Links

Last updated on 4/19/23

This post is really a listing of resources for Silicon Photonics and Linear Drive Links. 

Linear Drive Links

• Samtec + Alphawave demonstration at DesignCon 2022
• https://blog.samtec.com/post/linear-direct-drive-proof-of-concept-can-support-100-g-signaling/
• This is a demonstration of a linear drive system
• Serialink Systems webinar to IPEC
• Great presentation.
• https://www.ipec-std.org/11243.html

Silicon Photonics:

• Silicon photonics design, from devices to systems, book by Lukas Chrostowski

Dan-Xia Xu, Jens H Schmid, Graham T Reed, Goran Z Mashanovich, David J Thomson, Milos Nedeljkovic, Xia Chen, Dries Van Thourhout, Shahram Keyvaninia, Shankar K Selvaraja, "Silicon photonic integration platform-have we found the sweet spot?", Journal of Selected Topics in Quantum Electronics, 2014

https://eprints.soton.ac.uk/367864/1/__userfiles.soton.ac.uk_Users_nl2_mydesktop_Slabakova_Chemistry_articles_Silicon%2520photonic%2520integration%2520platform_revised_final_2.pdf

• Why 220 nm silicon thickness?
• MPW = multi project wafers, i.e. shared wafers
• "Is there a ‘sweet spot’ in the silicon layer thickness that enables optimizing the performance of devices, minimizing fabrication complexities, and maximizing tolerances to the inevitable variations in fabrication?"
• "monolithic silicon carrier-depletion modulators"  

Clock and Data Recovery

 Lee, Hae-Chang. An estimation approach to clock and data recovery. Diss. Stanford University, 2007.

• Section 2.1.1 Linear Phase Detector, gives a great overview of the Hogge phase detector, how it works (Creates two pulses that cancel each other out when zero phase difference), why it was popular (linear PD are easy to analyze), it's downsides (if the CDR is digital you need to convert output from analog to digital).

• KPD (loop gain) is define as the slope of the transfer function at 0.

• Section 2.2, Benefits of digital loop filter
• "Additional benefits of this architecture deriving from the digital loop filter are reduced pattern dependent jitter caused by leakage currents in the loop filter in the presence of CID, reduced phase offset error caused by charge pump current mismatch, reduced sensitivity to supply noise, and finally loop dynamics that are not affected by process, voltage and temperature variations."
• Section 2.2.1 Bang-Bang Phase Detector
• "KPD (loop gain) is difficult to define as the slope of the curve through the zero crossing is infinite (Figure 2.8 (b)). A method to approximate KPD exists and will be explored in a later section.
• Section 4.3.2 Jitter Tolerance
• 
  
  
  
• 
  
  
  
• 
  
  
  
• Also swept reference offset as part of the JTOL testing of the CDR
• "The first order CDR can only take corrective action after the error has occurred. For this reason, a time lag exists between the TX data transitions and the recovered clock in steady state which in turn offsets the data sampling point from its optimum.
• "The second order CDR learns the phase offset ramp rate (frequency offset) from past bits and takes predictive correction on the deterministic phase offset trajectory. This allows the second order CDR to drive the average steady state phase estimation error to zero.
•