Signal Integrity: VNA Applications

Transcription

1 Signal Integrity: VNA Applications Joe Mallon Business Development Manager VNA Products DesignCon February 2017

2 Agenda Why use both BERTS and VNA s? Anritsu VNA product types SI Challenges and Channel effects Measurement Frequencies and Signal spectra Time or Frequency Domain measurements Time Domain Transformations S-Parameter Quality Metrics Case Study Two VNA Architectures Summary and questions 2

3 BERT- Time Domain Measurements Tells you if you have issues: Jitter analysis Eye diagram opening Bit Error Rates Inter Symbol Interference (ISI) VNA and BERT strengths VNA Frequency Domain Measurements Shows you what is causing the issues: Dispersion Group Delay High Frequency losses Channel imperfections Trouble shooting Time domain analysis Impulse response Fixture effects and removal De-embedding Channel modeling issues S-Parameter Quality 3

4 Anritsu VNA Products 4 Copyright ANRISTU 4

5 Anritsu VNA Product Positioning High Performance VectorStar (MS464xB) Performance/Functionality Performance Economy ShockLine (MS46122A) ShockLine (MS46322A) ShockLine (MS46500B) ShockLine (MS46121A) Handhelds Price Note: Figure not to scale

6 ShockLine VNA Family MS46522B 2-port Performance VNA 300 khz to 8.5, 20 & 43.5 GHz N(f) or K(f) connectors MS46121A 1-port USB VNA 40 MHz to 4 GHz, 150 khz to 6 GHz N(m) connector MS46122A 2-port USB VNA 1 MHz to 8/20/43.5 GHz N(F) and K(M) connectors MS46522B 4-port Performance VNA 300 khz to 8.5, 20 & 43.5 GHz N(F) or K(f) connectors MS46322A 2-port Economy VNA 1 MHz to 4/8/14/20/30/43.5 GHz N(F) and K(M) connectors

7 VectorStar VNA Family Upper Frequency 20 GHz 40 GHz 70 GHz 110/145 GHz 1100 GHz Lower Frequency 10 MHz MS4642B MS4644B MS4647B (Option70 khz) ME7838 Systems VDI/OML All systems are upgradeable and 4 Port Microwave VNAs 2 and 4 Port Broadband, mm-wave and THzregion VNA Systems

8 ME7838 Broadband Systems ME7838A 2 Port, 70 khz to 110 GHz ME7838D 2 Port, 70 khz to 145 GHz ME7838A4 4 Port, 70 khz to 110 GHz Best physical package solution Modules 1/50 volume of other solutions Best broadband dynamic range 107 db to 110 GHz Best calibration stability 0.1 db and 0.5 o for S 21 over 24 hours Unique real-time closed loop ALC giving >55 db linear power sweep range E & W-band WG versions (WG Adaptor removable for use in coaxial applications) Receiver module for noise figure measurements to 125 GHz 2 and 4 port on-wafer configurations available 8 Copyright ANRITSU

9 VectorStar VNA Mainframe Architecture 9 Copyright ANRISTU 9

10 VectorStar Architecture: Two VNA s > 2.5 GHz High Band MS4640B Block Diagram (Fully Loaded Configuration) optional < 2.5 GHz Low Band a 1 a 1 a 2 a 2 Bridge Based Reflectometers Mixer Based Receivers b 1 b 2 b 1 Coupler Based Reflectometers b 2 Sampler Based Receivers Bias 1 Bias 2 Port 1 Port 2 Avoids Directional coupler roll off below 500 MHz and DR degradation Low Band swap over occurs well above GHz Each Technology is used in it s optimum frequency range Allows frequency coverage down to 70 khz Allows high DR measurements down to 70 khz 10

11 VNA Technology ShockLine NLTL MMIC ShockLine VNA Harmonic Samplers & Mixers Anritsu VNA s use patented NLTL harmonic samplers and Mixers for better dynamic range NLTL Samplers create sampling pulses with very fast transitions Step Recovery Diode transfer function tends to drop off by 50 GHz Non Linear Transmission Line (NLTL) technology offers a higher sampler frequency response with less drop off in high frequency performance Results in excellent dynamic range up to 70 GHz in the MS4647B VNA and up to 145GHz in Anritsu s mmw Broadband modules 11

12 Signal Integrity Challenges 12 Copyright ANRISTU

13 Typical SERDES Parallel Data 8B/10B Encoder Serializer Equalizer Driver High Speed Serial Data Channel VNA s measure Channel Issues Clock/PLL Parallel Data Data Recovery Equalizer De- Serializer 10B/8B Decoder BERT s measure Jitter, BER & Eye openings CR/PLL

14 Types of Channels Connectors Cables Backplanes Daughter boards Board to board IF PCB s Packages Package IF

15 VNA s Help Understand Structures and Imperfections Analyzing real world channel defects: Tolerances on PCB artwork, plating and dielectric thickness variations Connector mounting and performance and connector construction Signal vias on multilayer PCB stack ups & imperfect vias Imperfect vias Lack of Ground Planes De-embedding features and fixtures can help analyze issues

16 Channel Impulse Response Impulse response shows every characteristic of the channel Every impedance variation All the reflections and losses ISI root causes Dirac delta function We can calculate the impulse response and transfer functions from S-parameters Most accurate way to characterize a channel VNA s measure in the Frequency domain we can transform into the Time Domain to determine the impulse response 16 Copyright ANRISTU

17 S-Parameters can calculate Impulse response The Dirac function is impossible to generate (perfect impulse = infinite frequencies) VNA s can simulate this with a series of sine wave measurements in the frequency domain These are narrowband sequential measurements with very high S/N ratios We can measure all reflection and transmission characteristics of the channel at each frequency The result is a high dynamic range S-parameter matrix which captures all of the channel performance These S-parameter matrices allow us to do a Time domain transformation which will show the channel Impulse response Impulse response VNA Stimulus VNA TD Transformation V(f) Z(t) Dirac delta function DC Frequency On Measured S-Parameters Time 17 Copyright ANRISTU

18 Production compliance masks make use of S-Parameters Insertion loss Return loss Differential-common mode conversion 18 Copyright ANRISTU

19 Channel Effects High Frequency losses 19 Copyright ANRISTU

20 High Frequency losses 0 ] B [ d s t ic -20 r is t e c r a a h -40 C n io s is m-60 s n r a T Backplane Loss vs Frequency Frequency [GHz] High Frequency losses can close the eye Fixes: - Use lower loss materials - Use shorter path lengths - Use an eye opener with high frequency equalization - Use Emphasis

21 Adding Emphasis Emphasis adds additional high frequency energy to the transmit waveform by accentuating the rise/fall times which is where all of the high frequency content resides. This increased high frequency content in the data helps to offset the high frequency losses in the channel. 4-Tap Emphasis Block Diagram Emphasis Data Waveform (2Post/1Pre-cursor)

22 Setting Ideal Emphasis for the Channel Challenge: Difficult to find the ideal settings from the many possibilities. It s an iterative process at best. Problem: One method is to search for the ideal settings while verifying the output waveform, but this method takes an extremely long time and it is hard to come up with an explanation of why those settings are ideal.

23 Use VNA S-Parameters to set emphasis Transmission Path s2p or s4p file From VNA S21 Insertion Loss 's31a.txt' 'pre_ampd.txt' Measure Transmission path S-Parameters Import this file into the BERT to set the ideal emphasis

24 Channel Effects - ISI 24 Copyright ANRISTU

25 Inter-Symbol Interference (ISI) Square waves contain odd harmonics in their spectrum such as 1 st, 3rd and 5th harmonics. Dispersion: Different frequencies can travel at different velocities for certain media types

26 BERT: VNA Group delay measurement shows ISI ISI can also close the eye because low frequency components travel faster than high frequency components. This causes bunching of the data stream and a closed eye. VNA Group Delay vs Frequency: Group delay vs Frequency Non-Dispersive Channel This channel attribute can be measured on a VNA (group delay). Non-constant group delay over frequency means ISI may be present. Dispersive Channel BERT tells you that there is a problem VNA tells you why. 26

27 Measurement frequencies: High speed digital signal frequency content 27 Copyright ANRISTU

28 Square Wave Harmonic spectral content Data waveforms transmitted are supposed to be square waves. Synthesizing a square wave: An ideal square wave has only odd-integer harmonic content. Harmonics go from fundamental to infinity. How high in frequency do I need to measure to characterize the channel? "SquareWave" by Peretuset - Own work. Licensed under CC BY 3.0 via Commons -

29 Channel Measurement Frequencies Ideally, we would like to measure the channel with as many odd harmonics as possible but these high frequency systems can be impractical for real world applications 1 st harmonic Digital signal rise/fall times and can determine the measurement bandwidth. Since the transitions contain most of the higher frequency content. If your measurement system is capable of measuring the 5 th harmonic, it will have minimal impact on the measurement and adequately characterize the channel. 3 rd harmonic Signals with slower rise/fall times may only need measurement to the 3 rd harmonic to be properly characterized. 5 th harmonic

30 Channel Issues Frequency - 28 Gb/s NRZ & 56 Gb/s PAM4 28 Gb/s NRZ & 56 Gb/s PAM4 Data and Clock Spectrum NRZ: Data Spectrum Envelope Clock Odd Harmonics PAM4: DC 1st 3rd 5th 7th 14 GHz 28 GHz 42 GHz 56 GHz Frequency 70 GHz 84 GHz 98 GHz 28 Gb/s NRZ spectrum has significant content above 28 GHz 56Gb/s PAM4 Encoding gets 2x the data rate of by using more encoding levels in same BW Sin (x)/x shaped - Sinc function spectral envelope for both Higher frequency content depends on Data rise/fall times Recommend measuring S-Parameters to 3 rd (42 GHz) or 5 th (70 GHz) Clock Harmonic depending on signal Rise/Fall time 30

31 Channel Issues - Frequency 56 Gb/s NRZ 56 Gb/s NRZ Data and Clock Spectrum Data Spectrum Envelope Clock Odd Harmonics DC 1st 3rd 5th 7th 28 GHz 56 GHz 84 GHz 112 GHz Frequency 140 GHz 168 GHz 196 GHz Some of our customers are working on both PAM4 and NRZ Recommend measuring S-Parameters to 3 rd (84 GHz) or 5 th Clock Harmonic (140 GHz) These systems need a Broadband 110 GHz or 145 GHz Broadband VNA 31

32 Time vs Frequency domains 32 Copyright ANRISTU

33 Measurement in the Time or Frequency domains? SI measurements SI engineers typically work in the time domain using fast scopes and TDRs These can provide quick time domain results TDR s can determine S-Parameters with FFTs but have some limitations due to their architecture TDR s are wideband - TDR wideband noise can limit dynamic range (DR~40 db) VNAs have lower noise floors because they are narrowband systems and can have IFBW as low as 10s of Hz (DR>100dB) VNAs use vector error correction Corrects for imperfect system components Engineers need high resolution TD response to use as a tool for locating channel defects GHz VNA has ~3X the TD resolution of best TDR available What are the real VNA effects on time domain performance and quality?

34 Time Domain transformations VNA realities 34 Copyright ANRISTU

35 VNA Transforms Into The Time Domain Objective find out how the Channel impacts the eye diagram Is this the true picture? 0 ] B [d s tic -20 ris te c a ra h -40 C n s io is -60 s m n ra T Frequency [GHz] Potential Measured S-Parameter Issues: Noise Effects Fixture Effects Measurement Uncertainties Low Frequency Data Quality IFT s and Complex Algorithms Or this? What s important in making the transition to Time Domain?

36 Time Domain Transformation - Getting There - Although the mechanics differ, in the end, getting to time domain looks like: X N k n ( ) ( ) j 2 π t x f e f t n k = 1 - When the frequency (f k ) is small, the TD Response changes slowly with time: These low frequency S-parameter data points really affect the flat-tops of TD responses (like eye diagrams). - An extrapolated DC term also ends up being used for an integration reference. Get this wrong and step responses can have a slope. k

37 Time Domain Quality - Accurate Low Frequency Data DC Point Extrapolation Ripple Longer electrical length DUT s have more ripple in the lower frequencies. This ripple can make extrapolation of the DC point difficult. In general, the closer you can measure to DC the better the DC point can be determined. Noise and Instability f 2 f 1 Noise, instability and uncertainty in low frequency S-Parameter data also impacts the TD response and can tend to add noise to the flat tops of TD step responses Having high quality low frequency data enhances the TD transformation process and quality

38 Time Domain Quality Locating Defects - Resolution Two Mismatches Separated by 2mm (air) Time resolution proportional to 1/BW Rule of thumb (air): Distance (resolution) = 150mm/BW (GHz) Non-air distance : Divide by BW Resolution 1) 40 GHz 3.75mm 2) 50 GHz 3.0mm 3) 70 GHz 2.14mm BW Air Resolution 110 GHz 1.36 mm 145 GHz 1.03 mm BW (Frequency Span) Drives TD Resolution

39 Time Domain Quality Locating Defects - TD Modes There are different VNA TD Modes - Low Pass time domain mode has the highest resolution Provides TDR like step responses Can show impedance changes in channel Requires a quasi-harmonically related set of frequencies Start frequency (f start ) = Step size (f step ) The smaller the f step the more points can be used Smaller f start closer you are to DC for better DC point extraction LP Harmonic Calibration: TDR Like display: f start = f step Z - Frequency DC + Frequency Distance (or time)

40 Time Domain Quality Alias Free Range Sampling in the frequency domain and alias free range TD Transform is circular and repeats at T max (alias free range) Max alias free range vs Step size: T max = 1/(2f step ) These repeating TD results can obscure TD data that occurs after T max Important when measuring long structures or structures with high D k f start = f step Longer Alias Free Range - Frequency DC + Frequency Example: Alias free range on a PCB with a D k of 4: f step =40 MHz 75 inches alias free range f step =4 MHz 750 inches alias free range

41 S-Parameter Quality Metrics 41 Copyright ANRISTU

42 S-Parameter Quality Metrics Quality is important Reciprocity Forward and reverse transmission are equal in both Magnitude and Phase VNA s have excellent Reciprocity because of the architecture More of an issue for TDR s and scopes than VNA because of timebase jitter Passivity Issues occur when a passive structure appears to have gain Calibration, De-embedding and contact repeatability can effect passivity Verify with high quality low loss thru - airline (not cal thru!) Causality All VNA s have causality issues. Incomplete DC to Daylight data will cause S-Parameters to be non-causal. This shows up as output energy occurring before the input stimulus in the time domain Verify with S-Parameter CW rotation on a polar chart in the frequency domain or look for energy in the time domain for t<0

43 Causality - Bandwidth & Window Choices affect Causality Basic causality is always impacted by available finite bandwidth. The windows used on the data can also have a major effect. Note the signal levels for t<0. 20 GHz BW t=0 40 GHz BW with two different windows t=0 40 GHz BW No Weighting Weighting BW (Frequency Span) effects causality

44 S-Parameter Metrics Passivity Passivity problems occur when it appears a passive device has gain Receiver saturation issues can cause passivity issues during calibration or with the measurement De-embedding is often the problem. Small errors can cause channels to appear to have gain. This is most prevalent in fixtures with high IL fixtures and low loss DUT s Having a wide range of extraction methods for de-embedding fixtures and other devices available can be an advantage

45 Anritsu VNA Network Extraction and De-Embedding 45 Copyright ANRISTU

46 VNA based Network Extraction & De-embedding Fixture or Connector Fixture or Connector Extracting the S-Parameters of one or more connectors or fixtures on one or both test ports. Also used to extract part of a circuit on a Channel Network Extraction generates one or more S-parameter files that will be used to embed or de-embed measurements. De-embedding removes those features from the measurement. Can also do time domain gating and test port extensions. Add Slide Title, Date or URL Copyright ANRITSU

47 Anritsu VNA De-embedding Options All done on the VNA No external SW required New De-embedding options available soon

48 Case Study SI Channel measurements 48 Copyright ANRISTU

49 VNA Performance - Case Study Series of measurements made using 2 VNA s on same 40 inch line. Low pass TD data will be examined. Anritsu Lightning VNA 40 MHz to 40 GHz Step size 40 MHz (1,000 points) Reflectometers - Couplers for entire band LF Roll of starts at 500 MHz Anritsu VectorStar VNA 4 MHz to 40 GHz Step size 4 MHz (10,000 points) Reflectometers - Hybrid of bridges and couplers Flat response down to 70 khz

50 VNA Case Study - DC Extrapolation DC extrapolation depends on quality of S-parameter measurement and how close to DC the start frequency is 40 MHz 4 MHz Lightning VNA limited low frequency performance - Coupler based - 92 db DR at 40 MHz Lightning f start = 40 MHz VectorStar f start = 4 MHz VectorStar VNA with better low frequency performance - Bridges at LF db DR at 4 MHz

51 VNA Case Study - Step Response Good low start frequency data improves DC extrapolation Low start frequency sets frequency domain sampling for low pass step response and hence alias free range Slope due to poor DC extrapolation Aliasing due to 40 MHz sampling Lightning: 40 MHz to 40 GHz data fs=40 MHz, 1,000 points VectorStar: 4 MHz to 40 GHz data fs=4 MHz, 10,000 points

52 Eye at Port 2 (V) Eye Diagrams at 10 Gb/s 0.3 Veye of Bit Stream 0.5 Veye of Bit Stream Voltage (V) Voltage (V) Eye at Port 2 (V) Time (ns) Time (ns) Simulated Eye Lightning VNA - 40 MHz to 40 GHz Data Points - Couplers entire band Simulated Eye VectorStar VNA - 4 MHz to 40 GHz - 10,000 Data points - Couplers above 2.5 GHz - Bridges below 2.5 GHz Measured Eye Please stop by Booth 633 for a preview of on-vna Eye Diagram simulations from measured S-Parameters

53 New VNA SI Features 53 Copyright ANRISTU 53

54 Signal Integrity Software for ShockLine (Option 022) Advanced Time Domain Option: Based on ADK software developed by AtaiTec Corp Available for MS46500B Series Signal Integrity measurement tool Many SI apps in one place SI Functions Passivity and Causality Combining.SnP files Plotting Simulated Eye Diagrams Crosstalk (NEXT and FEXT) Skew Compliance testing Signal Integrity Software Menu Post Processed SnP files Allows SI engineers to characterize and test passive devices High Speed Cables Interconnects PCB Backplane Example Eye Diagram 54 Copyright ANRITSU

55 New VectorStar Eye Diagram Simulation (Option 47) Real Time trace based Eye diagram simulation tool. Sweep by Sweep indication of the data transmission characteristics. 55 Copyright ANRITSU

56 Defining VectorStar Simulated Eye Diagram Parameters Data Stream Configuration System Jitter System Noise 56 Copyright ANRITSU

57 Thank You. Questions? 57 Copyright ANRISTU

58