Eliminate Pitfalls of DDR Memory Testing
- 1. Eliminate Pitfalls of DDR Memory Testing November 12th, 2014
- 2. Eliminate Pitfalls of DDR Memory Testing • Example DDR Test Configurations • Best Practices for DDR Probing • DDR Connectivity With An Interposer • Removing Resistors When Using Interposers • Connectivity Examples Not Recommended • Generating DDR Traffic • DDR Waveform Checklist Before Compliance • DDR Eye Pattern Formation • Multiple-Scenario DDR Eye Patterns and Jitter • Real-World DDR Debug: Missing Clock Cycles • Real-World DDR Debug: Problem With Read Burst • Real-World DDR Debug: Logic, Soldering, Power Supply • DDR Measurement Highlighter • DDR Virtual Probing • DDR Compliance Testing • DDR Specialized Connectivity • DDR Probe Deskew • Determining DDR Probe / Scope Bandwidth • Questions?
- 3. Example DDR Test Configurations
- 4. Example DDR test configuration for a desktop computer platform Memtest utility generating read-write test output (random number sequence) Probes soldered to Clock, Strobe, and Data lines Probe tips soldered and hot glued to vias on DIMM
- 5. Solder-In Probes Tips Connected to Clock, Strobe, and Data Lines of the DDR SODIMM on an ASUS Netbook Example DDR test configuration for a netbook computer Differential Clock Line – Probe leads soldered to two vias directly opposite the differential termination on the other side of the board. Single-ended Data Line – Probe leads soldered to the memory "chip side" of the series resistor array and ground Differential Strobe Line – Probe leads soldered to the memory "chip side" of the series resistor array
- 6. Best Practices for DDR Probing
- 7. Recommended DDR probing method 1: Hands-free probe holder mounted in reverse positionStrain relief for the probing connection can be provided by utilizing this counter-weight configuration The typical use model for a hands-free probe holder is designed to place weight at the tip of the probe This reverse mount acts as a counterweight, removing force from the probe tip, and providing strain relief for the probing connection during DDR testing Not recommended probing configuration for DDR Recommended probing configuration for DDR
- 8. Recommended DDR Probing Method 2: Gooseneck strain relief Mount adhesive base on nearby chip, strain relief is provided to the solder tips Adhesive base mount Tips soldered to chip pins Probing discrete components A probe with flat geometry and rubberized flex circuit lead can be easily secured in place
- 9. Recommended DDR Probing Method 3: Chip clip secures probe to board or chassisA chip clip prevents movement of the probe platform cable assembly when mounted on a board edge or chassis Chip clip mounted on corner of DDR board Chip clip mounted on edge of computer chassis
- 10. Best practices example: using Kapton tape and signal labels on the board Signal labels on the board are recommended Kapton tape is recommended on the probe interconnect lead Kapton tape is recommended on the probe interconnect lead (but not the probe flex circuit tip or damping resistor leads)
- 11. Tips are soldered to DDR test points Gooseneck strain relief protects solder tips while probing strobe and data lines Best practices example: Gooseneck strain relief is used to prevent movement of probe head
- 12. Best practices example: using a chip clip, two probe holders, and two gooseneck strain reliefsShown below is a combination use of a chip clip, two reverse-mounted hands-free probe holders, and two gooseneck strain reliefs while probing DDR Clock, Strobe, and Data Reverse-mounted hands-free probe holders Chip clip Gooseneck strain reliefs
- 13. Probe leads are taped to the board at two different locations to provide secure strain relief Although it would be ideal to not move the device under test --if it must be moved --using a cart is recommended. The cart provides support for the three probe platform cable assemblies. This configuration could also be improved by anchoring the probe platform cable assembly to the cart with plastic ties, clips, or tape) An ESD bag is placed under the DDR DUT for static protection. The use of an ESD bag is always recommended, especially when the device is contained on a a metal cart. Best practices example: Multiple Taped Locations, Rolling Cart, Signal Labels, ESD Protection for the DUTShown below is a combination use of multiple taping of the probe heads, a rolling cart to support the probes and DUT, and ESDprotection Adhesive signal labels are placed on probe tips. Since the probe amplifiers will be disconnected during transport, the labels allow for rapid reconnection by the operator. Labeling could be further improved by adding adhesive labels to the probing points also. Probe tips are labeled
- 14. Best practices example: tape has completely secured the probing area for shipping After soldering, the probe tips are completely immobilized in tape, protecting the solder connections from all outside impact. This amount of taping is recommended if the setup needs to be shipped from one location to another, to ensure that the solder-in leads are firmly secured and the electrical connections remain intact. Kapton tape is recommended on the probe interconnect leads (but not the probe flex circuit tips or damping resistor leads) Kapton tape is recommended on the probe interconnect leads
- 15. Example DDR4 Test Configuration DDR4 DIMM mounted in slot in computer chassis Probes leads pre-taped to DDR4 DIMM
- 16. Probe leads soldered and taped to DDR4 DIMM before inserting into slot
- 17. DDR probe heads hot glued to the back of a single-sided DIMM Probe tips hot glued to the back side of the BGA ball out of this single-sided DIMM. This allows for secure connections during card insertion into the DDR slot. Note: always hot glue the top, not the bottom of the probes
- 18. Testing tip: perform test evaluations using single-sided DIMMs when available Single-sided DIMMs provide fairly easy access to probing on the back side of the BGA. Probing the back side of a DIMM is preferred, as reflections are minimized. When given a choice, a single-sided DIMM is always better than dual-sided, because access to the BGA ball out provides the best signal integrity for the probing connection. For evaluation and testing purposes, single-sided DIMMs provide the easiest path to signal integrity. However, since most DIMMs are dual-sided with DRAM chips on both sides, the access to vias on the back side of the chip is not always available and an interposer is often needed. Probe tips hot glued to the back side of the BGA ball out of this single-sided DIMM.
- 19. DDR Connectivity with An Interposer
- 20. Interposer Connectivity: Chip interposers are designed for probing extremely close to memory components, and are essential when probing points are otherwise inaccessible. Side View Top View The interposer is installed by soldering the bottom side of the interposer to the DDR BGA footprint on the target where the memory component would normally be soldered. The memory component is then soldered to the top side of the interposer. Chip with interposer Chip Without interposer Interposer Without Chip An interposer can be most useful in embedded applications and applications where there are chips on both sides of the DIMM
- 21. Another example of SI probes soldered to Clk, DQS, and DQ using a DDR chip interposer A DDR chip interposer mounts between the chip and target, or between the chip and DIMM.
- 22. View of entire board containing the same interposer
- 23. Testing tip: when using an interposer on a DIMM, target a central chip location if possible When installing an interposer, the side clips which secure the DIMM to the main PCB may not latch if an interposer is installed on the end chips. For this reason, it is recommended to install interposers on the six central chip locations highlighted above, rather than the end locations. DIMM with heat sync DIMM with heat sync removed Recommended to install interposer on central chip locations
- 24. Example of SI probes soldered to Clk, DQS, and DQ using a DDR Interposer Clock, Strobe, and Data probed using a DDR chip interposer Two interposers mounted, one not currently being used
- 25. Example of using a DDR chip interposer on a device with many serial interfaces including DDR3, HDMI, Ethernet, USB, and others DDR chip interposer DDR chip used on a motherboard with other serial technologies
- 26. Example of SI probes soldered to Clk, DQS, and DQ on a DDR Interposer mounted near the edge of a board DDR chip interposer (showing 3D perspective) Tweezers used during the soldering process Additional interposer (not being used at time of photo)
- 27. Best practices: use tweezers to manipulate SI probe tips when soldering Tweezers are more nimble and provide better placement and accuracy during the soldering process
- 28. Probing example using Nexus socketed interposer Clock Strobe Data
- 29. Probing example using socketed interposer for SODIMM Socketed interposer Taped gooseneck strain relief SODIMM will be connected to DDR slot
- 30. Probes soldered to Clock, Strobe, and 7 different Data lines Best practices: solder in advance, extra SI tips to test multiple data lines 9 probe tips are soldered directly to the interposer. Any of the data lines with a probe tip attached can be connected to the amplifier during testing.
- 31. Probe LED indicates channel (Yellow = CK, Pink = Strobe, Blue = Data) Method to keep signals organized
- 32. Removing Resistors When Using Interposers
- 33. Should damping resistors be removed from solder-in probes when using an interposer? Eye pattern with damping resistors removed Eye pattern with damping resistors attached If the damping resistors are removed from the probe when using an interposer --the probes terminate into the interposer without damping resistors --this results in slightly better signal fidelity when connected to the interposer but this also brings disadvantages. Below are the pros and cons associated with removing the resistors. Pros: + AC response accuracy improvement (3%) + DC gain accuracy improvement (4%) Cons: -Need to modify the probe: old leads and resistors removed, replaced with new leads -The probes cannot be used for other applications or for general purpose use until the resistors and leads are reattached to the probe tip PCB Notes: + Most users do not remove the resistors. + If the resistors are removed, the leads should be replaced with 34 gauge wire extending 3mm the probe tip PCB edge
- 34. Connectivity Examples Not Recommended
- 35. Caution: This configuration has non-standard leads added to probes which are not matched, and add inductance When damping resistors are removed for connection to the interposer, the leads should be replaced with 3mm 34 gauge wire
- 36. Fragile: In this example there is no hot glue, tape, probe holders or clips, or other strain relief provided to the delicate solder joints. Even a small impact to the board could dislodge the connections.
- 37. Caution: Using square pin or other adapters which reduce rated bandwidth can reduce measurement accuracy (unless testing low speed grade DDR signals) Square pin connectors are typically rated at 3 GHz bandwidth Solder-in tip rated at 13 GHz bandwidth
- 39. Example of good data transition density for testing Example of bad data transition density for testing CLK DQS CLK DQS DQ DQ Read/Write Burst Traffic -Goal Is To Generate Good DDR Transition Density Programs which communicate with DIMMs, such as Memtest86, are widely available. The recommended output is Test 7 in Memtest86 which continuously outputs R/W bursts for the duration of the test. A read or write burst should occur at least once per 10 us for compliance testing, but much higher density is recommended. The higher the burst density, the more statistical results can be computed during a DDR compliance test. Memtest86 generates DDR read/write burst pattern (random number sequence)
- 40. Additional DDR Transition Density Examples Excellent Bad OK OK
- 41. DDR Waveform Checklist Before Compliance
- 42. Waveform Characteristics to Verify Before Test Clock Strobe Data Signal check list: • Check that CK, DQS, DQ, etc. are on expected channels • Verify correct signal amplitudes • Validate that the CK signal is clean and at correct frequency • Verify presence of R and W bursts • Observe the relative amplitudes of R and W bursts • Validate proper idle levels Clock is a continuous waveform. Strobe appears as a bursted clock Address / control toggles slowly, either high or low Data is also bursted, but without as much activity as DQS Address / Control
- 43. Maximize Acquisition Dynamic Range For best results, signals amplitudes should occupy between 80 -90% of the grid vertical area Channel variable gain can be adjusted from compliance test menu
- 44. Verification of DDR Clock Example of a good DDR Clock signal Non-monotonic Example of a bad DDR Clock signal 1. Verify that the frequency is as expected (2 x Freq = Transfer Rate) 2. CK should not have any non-monotonic edges Non-monotonic edges can create significant DDR measurement timing problems
- 45. Visual Difference Between Read and Write Bursts Read bursts should be in phase with DQS If probing at the memory, Read bursts will have a larger amplitude Write bursts should be a quarter out of phase with DQS If probing at the memory, Write bursts will have a smaller amplitude W W R R R R R R Write burst
- 46. Check Idle Levels DQS:~0 mv DQ: ~750 mv (For DDR3: 1.5V/2) Offset levels can be adjusted during compliance
- 47. DDR Eye Pattern Formation
- 48. How a DDR Eye Pattern is Formed: Separation of Reads and Writes Write burst Read burst Skew between DQ and DQS is used to determine whether each burst is either a read burst or a write burst
- 49. How a DDR Eye Pattern is Formed: Unit Intervals Overlaid Read/Write Burst is sliced into unit intervals and overlaid to form the eye pattern Data read eye Strobe read eye
- 50. How a DDR Eye Pattern is Formed: Separated R/W Bits Overlaid Write burst Read burst Separated Read bursts form the Read Burst Eye Pattern, and separated Write bursts form the Write Burst Eye Pattern Write burst eye pattern Read burst eye pattern
- 51. Eye Pattern Provides Both Debugging and Validation of Signal Good data write eye Bad data write eye Shown below is a write eye pattern showing good signal integrity (left) and a write eye pattern with significant signal integrity problems (right). DDR eye patterns allow for a quick way to detect both voltage and timing problems simultaneously. The types of problems which can be identified with an eye pattern include jitter, reflection, glitches, runts, intersymbol interference, crosstalk, non-monotonic edges, slow rise or fall times, overshoot, noise, and more. Good strobe write eye Bad strobe write eye
- 52. Debug Eye Diagram Automatically Counts Number of Acquired Read / Write Bursts tDQDQS (P2) measures the skew between DQ and DQS once per burst. Therefore, the number of bursts is equal to the number of measurement instances displayed in the table of statistics. The number of Read bursts is counted
- 53. Real-World Problem: Data Probe Was Soldered To Wrong Node
- 54. QPHY-DDR Eye Diagram Debug It is recommended that this test is run prior to compliance testing to ensure the signals meet the requirements for compliance. Running this test will allow the user to verify: R/W separation Number of read and write bursts detected Rough check of signal skew Signal amplitude Reference for eye patterns can be either clock or strobe. 11/12/2013 54
- 55. Multiple-Scenario DDR Eye Patterns and Jitter
- 56. DDR Eye Mask Testing Mask testing using a DDR standard or custom mask can show mask failuresdue to jitter, reflection, glitches, runts, intersymbol interference, non-monotonic edges, crosstalk, slow rise or fall times, overshoot, noise, and more.
- 57. DDR Debug: View of 10 Eyes Simultaneously Data and strobe bursts are sorted and arranged to display all acquired Read and Write UIs as separate scenarios. Overlaying eyes allows valuable skew and timing information. Clock and Strobe can each be the timing reference.
- 58. Parameter Measurements on the DDR Eye
- 59. DDR Jitter Analysis with Tj / Rj / Dj jitter breakdown, DCD, Pkpk. RMS, TIE Track and Histogram, BTub Curve, CDF, etc can quantify and identify sources of jitter
- 60. Debugging with DDR Specific Parameters with Statistics
- 61. Debugging DDR With Analysis on Isolated Portions of Burst Data and Strobe on Read eye showing only the first bit after the preamble Data and Strobe on Write eye showing ignoring the first bit after the preamble
- 62. Real-World DDR Debug: MissingClock Cycles
- 63. Debug Steps: Eye Pattern Identifies two modes in the DQS and DQ relative to CK Problem identified in the Strobe eye pattern, with Clock used as the timing reference
- 64. Debug Steps: Triggering on the Strobe shows unstable (bimodal) clock timing
- 65. Debug Steps: Parameter Track of Skew Identifies Anomalies
- 66. Debug Steps: Parameter Track Locates Missing Clock Cycles
- 67. Debug Steps: Centering the DQS reference point on a DDR eye Strobe and Data eye patterns with Strobe used as the timing reference Strobe and Data eye patterns with Clock used as the timing reference
- 68. Real-World DDR Debug: Problem With Read Burst
- 69. Clock signal is correct Strobe has a problem during Reads Strobe is correct during Writes Debug Steps: Strobe waveshape is correct during Writes, but incorrect during Reads
- 70. Strobe not driven correctly during Read bursts Read Write The strobe voltage levels during a Read burst are centered around the Strobe idle voltage level. Notice that Read bursts are preceded by correct preample voltage drops, but are centered around electrical idle with clock crosstalk coupled into the signal. This seems to indicate that the Strobe is being driven incorrectly by the chip during Read bursts. Below is an example of a good Strobe signal from a different DUT. Notice that the idle voltage should only occur between Strobe bursts.
- 71. Debug Steps: Strobe not driven correctly during Read bursts Read burst Write burst Noise at the clock rate occurs during Read bursts. The Strobe appears to be driven to electrical idle during Read bursts. This problem does not appear during write bursts, indicating that this is not traditional interconnect crosstalk (or else it would occur whenever the clock was driving).
- 72. Debug Steps: Identifying Why Only The Read Burst Has Coupled Clock Noise Read The noise on DQS is related to the clock but is not traditional interconnect crosstalk. If it was traditional crosstalk it would be present anytime the clock is driving but is observed only when the memory DQS is driving. This indicates the noise is originating from the memory DQS driver and not coupled onto the trace later. It appears to be crosstalk inside of the DRAM chip onto VDD. Inspecting the ballout of the chip we find CK# is adjacent to VDD. Since VDD powers DQS, while VDDQ powers DQ, this is why the issue appears only on DQS and not DQ. Ballout of DRAM Write
- 73. Real-World DDR Debug: Logic Problems, Soldering Problems and Power Supply Noise
- 74. Real-World Debug: Logic Levels on DQ inconsistent Data Strobe
- 75. Real-World Debug: High speed and low speed noise coupled 20 ns/div (10,000:1 zoom) High speed noise visible between data bursts 20 ms/div (256 Mpts) Low speed noise visible between data bursts
- 76. Real-World Debug: Power Supply Noise coupled into DQ Data Strobe
- 77. Real-World Debug: Intermittent anomaly on only one side of Strobe diff pair Strobe + inv (Strobe -) Diff Strobe
- 78. Strobe distorted Real-World Debug: Data shape is ok, Strobe distorted
- 79. Real-World Debug: DDR4 Data Probe soldered to wrong node Data probe leads had been soldered between DQ1 and DQ5 instead of between DQ0 and ground. Strobe
- 80. Data probe positive lead soldered to DQ0, with negative lead floating instead of soldered to GND. Strobe Real-World Debug: DDR4 Data Probe negative lead floating
- 81. Non-monotonic waveshape Real-World Debug: Non-monotonic Strobe and Data
- 82. Read and Write Bursts Too Close Together
- 83. Data mask shape, no data transitions Real-World Debug: Pattern Incorrect -No Data Transitions
- 84. Data mask shape, no data transitions Real-World Debug: Pattern Incorrect -No Data Transitions
- 85. Since data mimics the strobe (every transition is identical), this may not allow the ability to detect ISI (not the most stressful test) Real-World Debug: Pattern Incorrect -Data Mimics Strobe
- 86. Data is too sparse Real-World Debug: Pattern Incorrect -Data Is Too Sparse
- 87. Also: Data pattern has no transitions or is very sparse most of the time The data pattern has only zeroes Sparse data transitions A different output pattern needs to be programmed. Use a random number generator similar to Memtest86 Test 7 for best results. The current sparse transitions will not allow for good DDR3 testing.
- 88. Debug starting point: Simultaneously check CK, DQS, DQ, Zooms and Eye Patterns
- 89. Debug Tip: DDR Cycle and Edge Measurements
- 91. Measurement Highlighter Identifies All Reads In Waveform (separated from Writes) W R R
- 92. Measurement Highlighter Identifies All Skew Measurement Instances Between DQ and DQS
- 94. Virtual Probing In DDR systems, access to probe points is limited. Ideally, probe should occur at the balloutof the DRAM. In practice the probe is often required to be over an inch away.
- 95. DDR Virtual Probing Since the probe is not directly attached to the receiver we see the superposition of the incident signal and reflected signal The distance from the probing point to the receiver can be measured by using the signal edge like a TDR. It will be ½ the measured time delay.
- 96. DDR Virtual Probing This LP-DDR2 signal shows significant reflections impacting the signal shape. It is not possible to make accurate measurements on a signal like this. The reflections are due to non-ideal probe placement. The reflections must be removed before proceeding to characterize the signal
- 97. DDR Virtual Probing Virtual Probe at Receiver • Quickly compensates for reflections due to imperfect receiver termination • Does not require S- parameters (S-parameters can be used with Eye DrII or VirtualProbe) • Uses familiar termination model • Effective work-around for probe access challenges on DDR systems
- 98. Virtually probed signal: Reflections removed Original raw signal: Reflections present Virtual probe at receiver settings DDR Virtual Probing
- 99. Testing Strategy –Virtual Probe Virtually move probe location to BGA Remove any effects of the interposer or channel through de-embedding
- 100. Testing Strategy –Virtual Probe at Receiver (VP@Rcvr) Often s-parameters files are not easily obtained for entire channel Model the circuit to reduce reflections in the signals under test
- 101. Virtual Probe and VP@rcv Setup Examples VP@rcv setup example Virtual Probe setup example
- 102. Testing Strategy –VP@RcvrExample DQS Before DQS After DQ Before DQ After Reflections at Vrefhave been removed
- 103. Testing Strategy –VP@RcvrExample Before After Virtually Probing at the receiver dramatically improved the eye DQ Eyes DQS Eyes
- 105. Automated DDR Compliance Testing Provides Highest Confidence In Less Time • Performs each measurement in accordance with the JEDEC standard • Compares each measured value to the specified limits • Easily document worst-case results and all statistical information • Connection diagrams guide user through each test setup • Stop on test for easy debugging
- 106. Instant Accumulation of results, worst-case scenario identified and reported
- 107. Automatically Generated Compliance Reports Report contains: Test values Specified test limits Screen captures Can be created as: HTML PDF XML
- 108. Wizard Compliance Feedback Example Occurs when less then 10 R/W burst acquired Warning depends on which tests are selected
- 109. DDR Compliance Example Measurements Clock TeststCK(avg) –Average Clock PeriodtCH(avg) –Average High Pulse WidthtCL(avg) –Average Low Pulse WidthtCK (abs) –Absolute Clock PeriodtCH(abs) –Absolute High Pulse WidthtCL(abs) –Absolute Low Pulse WidthtJIT(duty) –Half Period JittertJIT(per) –Clock Period JittertJIT(cc) –Cycle to Cycle Period JittertERR(n per) –Cumulative error Electrical TestsSlewR –Input Rising Edge Slew RateSlewF –Input Falling Edge Slew RateVIH(ac) –AC Input Logic HighVIH(dc) –DC Input Logic HighVIL(ac) –AC Input Logic LowVIL(dc) –DC Input Logic LowVSWING –Input Signal MaximumPeak to Peak SwingSoutR –Output Slew Rate RiseSoutF –Output Slew Rate FalltSLMR –Output Slew Rate Matching RatioAC Overshoot Peak AmplitudeAC Overshoot Area Above VDDQAC Undershoot Peak AmplitudeAC Undershoot Area Below VSSQVID(ac) –AC Differential Input VoltageVIX(ac) –AC Differential Input Cross Point VoltageVOX(ac) –AC Differential Output Cross Point Voltage Timing TeststHZ(DQ) –DQ High Impedance Time from CK/CK# tLZ(DQ) –DQ Low Impedance Time from CK/CK# tLZ(DQS) –DQS Low Impedance Time from CK/CK# tHP –CK Half Pulse WidthtQHS –DQ Hold Skew FactortQH –DQ/DQS Output Hold Time from DQStDQSH –DQS Input High Pulse WidthtDQSL –DQS Input Low Pulse WidthtDSS –DQS Falling Edge to CK Setup TimetDSH –DQS Falling Edge Hold Time from CKtWPRE –Write PreambletWPST –Write PostambletRPRE –Read PreambletRPST –Read PostambletDQSQ –Skew between DQS and DQtDQSS –DQS Latching Transition to Clock EdgetDQSCK –DQS Output Access Time from CK/CK# tAC –DQ Output Access Time from CK/CK# tDS(base) –DQ and DM Input Setup TimetDH(base) –DQ and DM Input Hold TimetIS(base) –Address and Control Input Setup TimetIH(base) –Address and Control Input Hold TimetDS1(base) –DQ and DM Input Setup Time (Single-ended Strobe) tDH1(base) –DQ and DM Input Hold Time (Single-ended Strobe)
- 111. Difficult-to-access DDR test points can be reached, and/or measurements can be taken under extreme temperatures using a high temperature long lead probe High-temp probe: 90 cm leads, -40 deg C to +105 deg C, 5 GHz bandwidth High-temp probes inserted into temperature chamber
- 112. Multi-Channel Testing: Clock, Strobe, and dozens of Data lines can be simultaneously monitored using multi-channel scopes Example: 20 GHz x 16 channels (Enables monitoring and comparison of Clock, Strobe and 14 Data or Control lines) Example: 30 GHz x 8 channels (Enables monitoring and comparison of Clock, Strobe and 6 Data or Control lines) Example: 13 GHz x 64 channels (Enables monitoring and comparison of Clock, Strobe and 62 Data or Control lines)
- 113. Using a multi-channel scope allows all of the data lines to be viewed simultaneously (up to 80 channels)
- 114. Multi-channel high bandwidth DDR debug and failure analysis Debug work on address lines from the register buffer spec, testing returned DIMMs DDR Address line debug Multiple probes activated
- 115. DDR Probe Deskew
- 116. Equipment for DDR Probe Deskew Three complete probe systems SMA 50-ohm termination Spare solder-in tip Characterization Fixture Torque wrench Oscilloscope
- 117. Since the three (or four) probe tips will likely already be soldered to the device under test, a fourth (or fifth) probe tip will be used for deskewing. Three probe tips are already soldered to Clock, Strobe, and Data A spare probe tip, used for deskewing all (3 or 4) other probes Deskew Convenience: Reserve A Spare Probe Tip for DDR Probe Deskew
- 118. Probe tip can be disconnected from probe amplifier anytime during operation (ok to hot-swap) Probe platform/cable assembly can be disconnected from scope anytime while powered. Probe leads can be disconnected from DUT anytime during operation Safety Note About Connecting/Disconnecting Active Probe Components While Powered Probe amplifier should not be connected or disconnected from probe platform/cable assembly while powered (toggle power off or disconnect probe platform/cable assembly from scope first)
- 119. Using a Probe Characterization Fixture for Deskewing Using a probe characterization fixture attached to the fast edge output of an oscilloscope, the probed signal delay can be directly compared to a known reference while triggering internally on the same edge, allowing precise deskew measurements of all channels without requiring a power splitter. SMA 50-ohm termination torqued to PCF PCF torqued to fast edge Probe tip leads clamped to PCF.
- 120. Strain Relief Techniques for Probes During Deskew Process Additional strain relief for the probe tip can be provided while deskewing by reverse-mounting the platform cable assembly using the hands-free probe holder, or by placing it on a surface near the PCF-200 Probe mounted in hands-free probe holder during deskew Probe resting on extra probe accessory box during deskew
- 121. Deskew Summary Measure the fast edge delay through each probe system, and adjust the skew value in the channel menu. Deskew values will be retained during the compliance process.
- 122. Determining DDR Probe / Scope Bandwidth
- 123. How To Determining Scope Bandwidth Requirements for DDR Testing: Step 1. The JEDEC DDR3 Standard Specification JESD79-3F Table 35 lists the DDR3-1600 maximum slew rate as 10 V/ns: Equation 1: DDR3 Fastest Risetime = MaximumVoltageswingMaximumSlewrate This calculation assumes both worst case voltage swing and worst case slew rate.
- 124. Step 2. Convert from slew rate to risetime: The JEDEC DDR3 Standard Specification JESD79-3F Table 1 lists the DDR3 Vdd supply voltage as 1.5 V. For differential data, the maximum voltage swing is 2 x 1.5 V = 3.0 V: Equation 1: DDR3 Fastest Risetime = MaximumVoltageswingMaximumSlewrate To compute 20-80% Tr below, we need to convert the maximum voltage swing (3.0 V) to 20%-80% voltage swing (which is 60% of the total voltage swing): Maximum voltage swing (20% -80%) = (60% x 3.0 V) = 1.8 V. Combining the values from Tables 1 and 35: DDR3 Fastest Risetime 20-80% Tr =1.8V10V/ns= 180 ps
- 125. To relate measured risetime to the actual (device) risetime and system (oscilloscope + probe risetime: Equation 2: Measured risetime = (Systemrisetime)2+(Actualrisetime)2 The system risetime for each scope + probe combination is listed in the corresponding probe datasheet:
- 126. Equation 2: Measured risetime = (Systemrisetime)2+(Actualrisetime)2 Using Equation 2, and inserting values from the probe datasheets and the DDR3 JEDEC specification: Using a 6 GHz probe/scope system for DDR3-1600: Measured risetime 20-80%: (56푝푝푝푝)2+(180푝푝푝푝)2= 3136+32400= 188.5 ps Measurement error = Measured−ActualActual= 188.5−180180= 4.7 % error Using an 8 GHz probe/scope system for DDR3-1600: Measured risetime 20-80%: (37.5푝푝푝푝)2+(180푝푝푝푝)2= 1406.25+32400= 183.9 ps Measurement error = Measured−ActualActual= 183.9−180180= 2.2 % error
- 127. Questions? • Example DDR Test Configurations • Best Practices for DDR Probing • DDR Connectivity With An Interposer • Removing Resistors When Using Interposers • Connectivity Examples Not Recommended • Generating DDR Traffic • DDR Waveform Checklist Before Compliance • DDR Eye Pattern Formation • Multiple-Scenario DDR Eye Patterns and Jitter • Real-World DDR Debug: Missing Clock Cycles • Real-World DDR Debug: Problem With Read Burst • Real-World DDR Debug: Logic, Soldering, Power Supply • DDR Measurement Highlighter • DDR Virtual Probing • DDR Compliance Testing • DDR Specialized Connectivity • DDR Probe Deskew • Determining DDR Probe / Scope Bandwidth