Probing Techniques for Signal Performance Measurements in High Data Rate Testing

Probing Techniques for Signal Performance Measurements in High Data Rate Testing K. Helmreich, A. Lechner Advantest Test Engineering Solutions GmbH Contents: 1 Introduction: High Data Rate Testing 2 Signal Performance Measurements: Why? How? 3 Probe Requirements and Solutions 4 Measurement Results 5 Summary

Introduction Automated Test Equipment (ATE) for high clock rate / high data rate logic / memory ICs... parallel test: up to 128 DUTs thousands of interconnects signal transition time: down to 0.4 ns (20% / 80% @ 3 Vpp) clock speed / data rate: up to 1GHz / 1.066 Gbit/s per pin ATE / DUT interconnect: multi-channel device interface (DI) ATE / DI interconnect: multi-channel, non-coaxial connectors DI / DUT interconnect: test sockets, probe card 'fly-by wiring': DUT I/O pin in: high Z out: low Z 50 Ω driver comparator

Memory IC Test Fixture

Introduction Motivations for ATE Signal Performance Measurements ATE: flexible, multi-channel measurement equipment of specified, high and traceable precision but... ATE manufacturer specification only for defined termination appliction specific DIs qualified separately ATE user may want to... verify overall ATE & DI performance in target application use ATE 'beyond spec' compare ATE from different vendors: 'benchmark'

Introduction ATE Signal Performance Evaluation requires... in-situ measurements on test sockets or wafer probe tips high bandwidth oscilloscope and probes high impedance probes to emulate application load Challenges test sockets contact pitch: down to 0.5 mm wafer probe card tip pitch: down to 50 µm oscilloscope probes limit measurement bandwidth

Probe Bandwidth Requirements Assumptions: 1. Test signals show 'Gaussian' edges (i.e. time integral of Gaussian pulses) 2. Probe shows Gaussian filter characteristic: a(db) ~ -f 2 B 3dB = 0.22 / t t 20/80 (signal / intrinsic) 80% 20% t t 20/80 signal passing probe: t 2 t, signal, out = t2 t, signal, in + t 2 t, probe, intrinsic 0dB -3dB 0 B 3dB f tolerated transition time prolongation 1%: B 3dB, probe > 1.55 / t t 20/80, signal, in t t 20/80, signal, in 0.4ns B 3dB, probe > 3.8 GHz

Model and Reality Real Test Signals... may not show exactly Gaussian edges Real Probes... may not show Gaussian filter characteristic Fortunately... above equations are fairly accurate for a variety of egde shapes and monotonous low pass filter responses Unfortunately... real probe inductance leads to undulating magnitude response, frequency dependent phase (delay) response, overshoot, parasitic transition time reduction

Probe Selection Among commercially available probes... active (FET) probes do not show sufficient bandwidth passive 50 Ω and 500 Ω (1:10) coaxial probes are more appropriate and even show lower input capacitance however: ground lead provided with commercial probes limits bandwidth due to its inductance probes characterised in frequency & time domain improved ground lead developed: - maintains coaxial geometry - fits application topology

Probe Characterisation, Measurement Assembly Vector Network Analyser Digital Sampling Oscilloscope Micromanipulator Assembly Coaxial Probes

Probe Characterisation, Measurement Assembly Detail micromanipulator assembly allows µm-positioning and safe, reproducible contacting

500 Ω (1:10) Coaxial Probe: Input Impedance vs. Frequency vendor specification sheet analytical model calculation with vendor specified value for C and reasonable value for L 10 3 input impedance of 500 Ohm / 0.15 pf / 2 nh probe 10 0-10 mag(z) / Ohm 10 2-20 -30 phase(z) / o -40-50 10 1-60 10 7 10 8 10 9 10 10 frequency / Hz model of 500 Ω probe: R 500Ω C 0.15pF 10*U out U in Z = R/(1+iωRC) + iωl L 2nH

500 Ω (1:10) Coaxial Probe: Input Impedance vs. Frequency vendor specification sheet measurement assembly: reference plane, 500 Ω probe, termination, Z from s 11 Z Z phase phase (Z) (Z)! Specified and measured probe characteristics differ! Resonance frequency 7GHz (instead of 9GHz). Maximum phase error larger than 60 degrees. Vendor spec apparently based on model!

500 Ω Probe Insertion Loss and Group Delay expected attenuation: 20log 20log 10 { 10 {[Z/(Z+R S +R S +R P )] P )]//[Z/(Z+R S )] S )]}} = -14.8dB standard assembly: reference plane 50 Ω probe 500 Ω probe reference plane modified ~ R S = 50Ω R P =450Ω Z=50Ω standard modified modified 500 Ω probe still shows amplitude error of about 2dB (>25%) and group delay distortion of more than 50ps in frequency range up to 4GHz

50 Ω Probe Insertion Loss and Group Delay standard modified assembly: reference plane 50 Ω probe #1 50 Ω probe #2 reference plane ~ R S = 50Ω Z=50Ω modified standard modified 50 Ω probe shows amplitude error of less than 0.4dB (5%) and group delay distortion of less than 20ps in frequency range up to 8GHz

500 Ω Probe In-System Insertion Loss and Group Delay through expected attenuation: 20log 20log 10 {[ZR 10 {[ZR T /((R T /((R T +Z+R T +Z+R P )R P )R S +(Z+R S +(Z+R P )R P )R T )]/[Z/(Z+R T S )]} S )]} = -20.4dB assembly: reference plane PC3.5 terminated 50 Ω transmission line contacted by 500 Ω probe reference plane PC3.5 through probe probe probe probe ~ R S = 50Ω R T = 50Ω R P =450Ω Z=50Ω amplitude error of about 4dB (60%) and group delay distortion of more than 70ps in frequency range up to 4GHz

Application: ATE Benchmark Digital Sampling Oscilloscope Delay Line Modified Probes Micromanipulator Assembly RDRAM 2 32 FPBGA DI DI 1.066 GBit/s ATE

Detail: Probing on FPBGA Sockets (0.5mm pitch)

ATE Signal Performance Measurement Contact Travel Impact depressing socket contact by by 330 µm increases delay by by 12 12 ps ps due to to reduced inductance

ATE Signal Performance Measurement Probe Impact measurement equipment transition times TDT step (black): 18ps + cable (green): 31ps + 50 50 Ω probe (red): 40ps

ATE Signal Performance Measurement ATE Signal 500 MHz clock signal picked up up via modified 500 Ω probe

ATE Signal Performance Measurement Transition Time 500 Ω probe: t t t 20/80 20/80 = 337ps 50 50 Ω probe: t t t 20/80 20/80 = 387ps edge timing linearity measurement inhibited by by oscilloscope timebase inaccuracy of of about 10ps

ATE Signal Performance Measurement Eye Diagram eye diagram 1GBit/s random pattern eye opening >900ps

ATE Signal Performance Measurement clock signal: 12ps RMS Jitter random pattern: 9ps RMS

ATE Signal Performance Measurement Crosstalk crosstalk to to neighbour pin 70mV pp pp

Summary Probe amplitude and phase (or group delay) response need to be qualified in a frequency range of up to 4GHz, if high-end ATE signal transition times shall be measured with about 1% accuracy Precision measurements beyond a few 100MHz require 50Ω=technology If in-system measurements with high impedance probes have to be taken, results will suffer from significant amplitude error and group delay distortion Experience shows, that probe / oscilloscope vendor specifications are highly questionable or insufficient Therefore, a thorough characterisation of probes is absolutely mandatory Traceable equipment needs to be used for characterisation