John Rettig Principal Engineer Tektronix, Inc. Beaverton, OR 97077 (503)-627-3232

Measurement Accuracy Considerations for Characterizing Low Impedance Stripline and Microstrip PC Board Traces John Rettig Principal EngineerTektronix, Inc.Beaverton, OR 97077(503)-627-3232 john.b.rettig@tek.com

Agenda • Launch resistance • Launch inductance • Device coupling error • Multiple reflections • Line loss • Strobe placement error • Introduction • Sources of TDR error • Standard impedance error • Inadequate resolution • Stimulus amplitude error • Sampler nonlinearity error • Rho zero error • Stimulus / Sampler aberrations • Recommended TDR measurement technique

Interconnect Issues • At high-speeds, interconnect limits system performance. • It is desirable to characterize and model interconnect to predict the performance early in the design phase. ImpedanceChange Incident Energy Transmitted Energy Reflected Energy

Interconnect Issues • Whenever energy transmitted through any medium encounters a change in impedance, some of the energy is reflected back toward the source • The amount of energy reflected is a function of the transmitted energy and the magnitude of the impedance change • The time between the transmitted energy and the reflection return is a function of the distance and velocity of propagation

TDR Overview • Time Domain Reflectometry - a measure of reflection in an unknown device, relative to the reflection in a standard impedance. • Compares reflected energy to incident energy on a single-line transmission system • Known stimulus applied to the standard impedance is propagated toward the unknown device • Reflections from the unknown device are returned toward the source • Known standard impedance may or may not be present simultaneously with the device or system under test

Incident Step Sampler Reflections 50 t = 0 Step Generator TDR Overview - Typical System

TDR Overview Elements • High speed stimulus, usually a step generator • High speed sampling oscilloscope • Back (internal) termination • Interconnect system • Device launch • Known standard impedance

+ 1  Open (Z =) Reflected 0  (Z = 50) Incident - 1  Short (Z = 0) Time t0 t1 TDR Overview TDR Waveforms - Open, Short and 50 terminations Amplitude

Reflection Coefficient and Impedance Where Z represents the test impedance Z0 is the reference impedance r is measured by the oscilloscope

Measuring Impedance  Cursor 1 at Cursor 2 at 50 Reflection Reference from Load Line Resistor  div  200ps/div

Measuring Impedance

Nonlinear Impedance / r Mapping • Everything else equal, lower impedance line measurements can tolerate more r error for a given impedance tolerance • Assumed conditions • 250 mV step • 50  Reference Line • 1 mV or 4 mr error equates to: • 0.40  for a 50 W test line • 0.24  for a 28 test line • 0.79  for a 90 test line • 1.23  for a 125 test line

Measuring Impedance

TDR Overview - Lumped Discontinuities

TDR Overview - Microstrip Discontinuities Open Circuit Volts or  Connector InductiveDiscontinuity CapacitiveDiscontinuity Incident Step Round Trip Time

TDR Accuracy Issues - First Order, no Explicit Z0 Standard • Controlled-impedance environment • Back termination impedance • Interconnect system • Vertical • Step amplitude accuracy • Sampler vertical accuracy • Horizontal • Step position accuracy • Sampler position accuracy

Employing a Standard Impedance • Inline with launch • Exchange with device under test EXCHANGE INTERCONNECT INLINE 1 TDR UNIT 2 DUT

Standard Impedances • Absolute Standards • Usually air lines built with extreme care employed for control of physical dimensions • Easily disassembled for checking dimensions • Standards Labs can characterize and periodically check • Based on physical equations like coaxial expression

Standard Impedances • Golden Standards • Absolute accuracy less important • Can be anything stable, repeatable, portable • Often need several sets depending upon supply chain • Characterization needs “round robin” or similar redundant checking

Microstrip Standards • Microstrip standards depend on many parameters with dispersed controls • Physical dimension (width w, thickness t, height h) • Dielectric media (glass and epoxy characteristics, glass/epoxy ratio, homogeneity) • PCB processes

Inline Versus Exchange Standard Inline Exchange Always in place - remainsprimary standard in system One connector change Two connector aberrations Measurement zone afterstandard location Loss in standard can affectaccuracy Additional multiplereflection compounding Recommending exchanged standard in this presentation Substituted for DUT - interconnectbecomes secondary standard One connector change One connector aberration Measurement zone atstandard location No affect from loss in standard No additional multiplereflection compounding

Z1 Z2 Standard Impedances • For reasons that will become apparent, want to use standard impedance close to device under test • Error on device measurement is primarily additive • 50W is the most commonly available and least expensive line • Can Parallel Two Lines for 25W, Using Tee(s)

TDR Resolution • Insufficient TDR resolution • Results from closely spaced discontinuities being smoothed together • Can miss details of device under test • May lead to inaccurate impedance readings

TDR Resolution • SMA through F-F barrel(8.6 mm dielectric) • Each end individually loosened 2.5 turns(1.8 mm) • Both ends loosened 2.5 turns; risetime limits • Top -- Full (~28 ps) • Mid -- 50 ps • Bottom -- 75 ps

TDR Resolution • TDR system risetime is related to resolution • Reflections last as long as the incident step and display as long as the system risetime • First discontinuity reflection is witnessed • Twice the propagation delay between discontinuities elapses • Second discontinuity reflection is witnessed • Ideally, leading corner of reflection from second discontinuity arrives back at first discontinuity no earlier than lagging corner of reflection from first discontinuity, thus

TDR Resolution • Previous rule assumes 0-100% ramp model; real world specifies 10-90% quadratic-type responses • System rise time is characterized by fall time of reflected edge from ideal short at test point • Other second order factors enter picture • System rise time approximated by:

TDR Resolution • Resolution Factors • Rise Time • Settling Time • Foot and Preshoot Settling Time Rise Time Foot Preshoot

Stimulus Amplitude Error • Percentage Z error roughly doubles percentage r error near Z0 • Amplitude shift with DC termination • 6-8 mr typical 0 - 50W • 6-8 mr typical 50W - open • Can be entirely compensated by directly measuring incident amplitude Amplitude impacts here

Stimulus Amplitude Compensation • Measure incident amplitude during calibration cycle • Keep same DC loading on TDR system during device test • Calibration cycle can be same as standard impedance check • Note that use of standard impedance close to device minimizes any residual errors

Sampler Nonlinearity Error • Samplers have limited dynamic range relative to general purpose oscilloscopes • Extremes of dynamic range may have small nonlinear behavior • Use of limited range stimulus minimizes this • Lower than Z0 impedance measurement environment is less vulnerable because of inverted reflection • Usually very small effect

+ 1  Open (Z =) Reflected 0  (Z = 50) Incident - 1  Short (Z = 0) Time t0 t1 Sampler Nonlinearity Error • Lower than Z0 impedance measurement environment is less vulnerable because of inverted reflection

Sampler Nonlinearity - Recommendations • Preserve identical gain and possibly offset settings for all portions of a given standard impedance calibration followed by DUT measurements

Rho Zero Error • Start of incident edge may not register as zero r • Ohms calibration usually assumes -1r = 0W

Rho Zero Error • Physical hardware may not use zero volt baseline - baseline instead restored in software • Typical 2 mhysteresis total (1 m) in software algorithm

Rho Zero Error • Can be entirely compensated - use amplitude measurements for incident and reflected amplitudes • Disable software baseline restoration if present • In absence of other accuracy improvement techniques, use of a set zero rho feature along with standard impedance close to device, provides remarkable improvement

Stimulus / Sampler Aberrations • Aberrations - foot, preshoot, ringing, longer term effects • Can amount to several percent • Short or long term • All reflections are replicas of the original step, filtered by reflecting device • Foot distorts response before the reflection, affecting establishment of baseline • With linear devices, it doesn’t matter how aberrations are apportioned between stimulus and sampler

Aberration Replication Preshoot replica Settling replica Reflected Incident

Stimulus / Sampler Aberrations • Typical specifications • 10 ns to 20 ps before step: ±3% or less • <300 ps after step: +10%, -5% or less • 300 ps to 5 ns after step: ±3% or less • Elsewhere: ±1% or less

Stimulus / Sampler Aberrations • 100 ps/div • 1 ns/div • 10 ns/div • 100 ns/div

Stimulus / Sampler Aberrations • Measurement over a zone to establish r levels and amplitude includes: • Device reflections • Stimulus / Sampler aberrations incident • Stimulus / Sampler aberrations reflected • Recommended measurement technique will involve: • Use of exchanged impedance standard • Measurement over identical zones for both standard impedance and device

Incident Step Sampler Reflections 50 t = 0 Rcontact Step Generator Launch resistance • Indiscernible from device impedance • Effect is additive, always positive, and similar to an equal amount of standard impedance error

Launch resistance • May not necessarily be repeatable • Most often is reasonable bounded (tens of mW) • Can be measured with 4-wire measurement - replicate several in series if necessary • May also be present to some degree in ground contact • Use of exchanged standard impedance will cancel launch resistance provided it is constant • Otherwise, make estimate of mean value and include for every measurement

John Rettig Principal Engineer Tektronix, Inc. Beaverton, OR 97077 (503)-627-3232