1、JEDEC SOLID STATE TECHNOLOGY ASSOCIATIONJEP152MAY 2007JEDECPUBLICATIONDDR2 DIMM Clock SkewMeasurement Procedure Using AClock Reference BoardNOTICE JEDEC standards and publications contain material that has been prepared, reviewed, and approvedthrough the JEDEC Board of Directors level and subsequent
2、ly reviewed and approved by the JEDECLegal Counsel. JEDEC standards and publications are designed to serve the public interest through eliminatingmisunderstandings between manufacturers and purchasers, facilitating interchangeability andimprovement of products, and assisting the purchaser in selecti
3、ng and obtaining with minimum delaythe proper product for use by those other than JEDEC members, whether the standard is to be usedeither domestically or internationally. JEDEC standards and publications are adopted without regard to whether or not their adoption mayinvolve patents or articles, mate
4、rials, or processes. By such action JEDEC does not assume anyliability to any patent owner, nor does it assume any obligation whatever to parties adopting theJEDEC standards or publications. The information included in JEDEC standards and publications represents a sound approach to product specifica
5、tion and application, principally from the solid state device manufacturer viewpoint. No claims to be in conformance with this standard may be made unless all requirements stated in thestandard are met. Inquiries, comments, and suggestions relative to the content of this JEDEC standard or publicatio
6、nshould be addressed to JEDEC Solid State Technology Association, 2500 Wilson Boulevard,Arlington, VA 22201-3834, (703)907-7559 or www.jedec.org. Published by JEDEC Solid State Technology Association 2007 2500 Wilson Boulevard Arlington, VA 22201-3834 This document may be downloaded free of charge,
7、however JEDEC retains the copyright on this material. By downloading this file the individual agrees not to charge for or resell the resulting material. Price: Please refer to the current Catalog of JEDEC Engineering Standards and Publications At www.jedec.org Printed in the U.S.A. All rights reserv
8、ed PLEASE! DONT VIOLATE THE LAW! This document is copyrighted by the JEDEC Solid State Technology Association and may not be reproduced without permission. Organizations may obtain permission to reproduce a limited number of copies through entering into a license agreement. For information, contact:
9、 JEDEC Solid State Technology Association 2500 Wilson Boulevard Arlington, Virginia 22201-3834 or call (703) 907-7559 JEDEC Publication No. 152-i-DDR2 DIMM Clock Skew Measurement Procedure Using a Clock Reference BoardContents1 Scope.12 Differential vs. Single-ended probing.13 Measurement of skew, j
10、itter and slew rate using differential probes23.1 Test Equipment and setup .23.2 Preparation .33.3 Data aquisition 43.4 Skew .43.5 Jitter 63.6 Slew rate.74 Measurement of skew, jitter, slew rate and crosspoint voltage using single-ended probes94.1 Test equipment and setup.94.2 Preparation .104.3 Dat
11、a aquisition 114.4 Skew .124.5 Jitter 134.6 Slew rate.134.7 Crosspoint voltage 13Annex A Measuring Equipment Jitter14Annex B Overscaling 16Annex C Correlation bewteen equipment/instrument jitter, observed jitter and true jitter 17Annex D Correlating Clock Reference Boards and Normalized Skew19Annex
12、E Differences betwee revisions .21JEDEC Publication No. 152-ii-JEDEC Publication No. 152Page 1DDR2 DIMM Clock Skew Measurement Procedure Using a Clock Reference Board(From JEDEC Board Ballot, JCB-04-84A, Formulated under the cognizance of theJC-45.1 Subcommittee on RDIMM.)1ScopeThis document is the
13、work product of the JC-45.1 “DDR2 DIMM Clock Skew Measurement” task group.The purpose of this document is to define procedures to measure clock parameters on registered DIMMs using the DDR2 Clock Reference Board. It is NOT the intent of this document to set specification values or validation require
14、ments. 2 Differential vs. Single-ended probingThis document describes measurements with both single-ended and differential probes. There are several pros and cons of each method: hard to calibrate and keep calibrated four probes at once. more chances for human error with four probes. roughly half th
15、e effort with four single-ended probes if Vix measurement is required differential probes are perceived to be more accurate especially when close-by GND contact for the single ended probes is not availible. single ended jitter measurements can be more accurate than those with differential probes, es
16、pecially if the bandwidth of the single ended probes is higher. JEDEC Publication No. 152Page 23 Measurement of skew, jitter and slew rate using differential probes3.1 Test equipment and setup Oscilloscope with minimum bandwidth of 3 GHz and 8 Gigasamples/s, Two differential probes with minimum band
17、width of 3 GHz, resulting in a probe tip bandwidth of no less than 2.5 GHz PC2-3200/4300 JEDEC CRB Clock Reference Board (=CRB) by CST, DC Power Supply for 3.3 V to CRBFigure 1 Measurement setup with differential probesScopeChDiff. ProbesCLK SynthesizerCLK BufferSwitchDRAM Reg.PLLDIMMDRAMDRAMDRAMDRA
18、MDRAMDRAMDRAMDRAMCLK Ref. NetsVddDRAM Reg.DRAMDRAMDRAMDRAMDRAMDRAMDRAMDRAMJEDEC Publication No. 152Page 33 Measurement of skew, jitter and slew rate using differential probes (contd)3.2 Preparation1) Allow instruments to warm up ( 20 min),2) Disconnect all probes from the scope,3) Perform the scope
19、calibration,4) Attach differential probes 1 a flight time of 50 ps per 12 mm is a reasonable estimate for the required correction. The deltas need to be determined from the net topology and may be obtained from the DIMM design specification of the DIMM under test.tD-tX-CH2VTH()tX-CH1VTH()=tsk(LH)1N-
20、tD+N=tsk(HL)1N-tD-N=JEDEC Publication No. 152Page 63 Measurement of skew, jitter and slew rate using differential probes (contd)3.5 Jitter1) Obtain by non-linear interpolation, the positive going time of threshold (Vth = 0 V) crossing for the recorded CH2 signal. Do this by first finding the pair of
21、 points on an upward slope straddling Vth = 0 V, and then interpolating the time at which the line is estimated to intersect the threshold (same as step 1 in 3.4). Record each of these values as T+n.2) In a fashion nearly identical to step 1, obtain by nonlinear interpolation, the next negative goin
22、g time of threshold crossing the recorded CH2 signal. Thus a single value for T-nis given by the time of downward crossing of CH2Vth for the nth time (same as step 3 in 3.4).3) Repeat steps 1 and 2, for the remaining threshold crossings in the recorded signals, thus obtaining two arrays of values T+
23、nand T-n(for n = 1 to N) where N is the total number of clock cycles in acquired CH2 recorded signal, and is about 10,000.4) Two other arrays of half-period values are obtained from the T+nand Tn-values by forming the difference between adjacent T+nand T-n, thus HP+n=T+n T-nand HPn-=T-n T+n+1. See F
24、igure 2.Figure 2 Graphic relationship between T+-n, HP+-nand P+-nTn+Tn+1+Tn-Tn+1-Tn+2+Pn+HPn+ HPn+1+HPn-HPn+1-Pn+1+Pn-VthJEDEC Publication No. 152Page 73 Measurement of skew, jitter and slew rate using differential probes (contd)3.5 Jitter (contd)5) Yet one more array of period values is obtained fr
25、om the transition time values T+nby forming the difference between adjacent T+nentries; thus P+n= (T+n) (T+n+1). See Figure 2.6) Now create a new array as follows: using the array of P+values, calculate: for each non-overlapping window of 1,000 cycles in the array. For an array with 10,000 cycles, t
26、here will be 10 (m = 0 . 9) non-overlapping windows with 1,000 cycles each.7) Compute the average value of the Tj(P+)m array. Report this number as tjit(per)1. 8) Repeat steps 6 and 7 for each array HP+and HP-to obtain Tj(HP+) and Tj(HP-).3.6 Slew rate1) Obtain by non-linear interpolation, the posit
27、ive going time tX+ of threshold (VTH1= -250 mV) crossing and the positive going time tX+ of threshold (VTH2= 500 mV) crossing for the recorded CH2 waveform. Do this for each threshold by first finding the pair of points on an upward slope straddling the threshold, and then by interpolating the time
28、at which the line or curve2passing through both point intersects the threshold. Thus a single estimate for sr+ is given by the ratio:1.the calculation of jitter used here differs slightly from the definition of jitter in the PLL specifications e.g. JESD65. In the PLL specification one would calculat
29、e not half of the delta between min and max, one would calculate the max and min deviation from the average of all measurements. It is believed that the dif-ference is negligible. The advantage of using this formula is that this function is readily availible on most scopes.2.It is a standard procedu
30、re to up-sample between the two straddling points (using the nearest neighboring points on either side) to obtain a new set of data in the region of the crossing. The two new straddling points from the new set are determined, and a final linear interpolation used to estimate the time of thresh-old c
31、rossing.Tj P+()mmax P+(1000m).P+(1000m+999)min P+(1000m).P+(1000m+999)2-=tjit(per)110-Tj P+()mm 0=9=sr+VTH2VTH1tX+CH2VTH2()tX+CH2VTH1()-=JEDEC Publication No. 152Page 83 Measurement of skew, jitter and slew rate using differential probes (contd)3.6 Slew rate (contd)2) As for step 1, obtain by non-li
32、near interpolation, the negative going time tX- of threshold crossing for each of the thresholds (VTH3= 500 mV and VTH4= -250 mV). Thus a single estimate for sr- is given by the ratio:3) Repeat steps 1 and 2, for the remaining events (zero crossings in the up-sampled difference waveform) forming an
33、average value for each of the measurements, until an average of at least N = 100 values is obtained. These average values constitute the final values for the measurement of SR+ and SR- for this node. Record these values.4) By choosing different levels for VTHa more accurate value for “slew rate” cou
34、ld be obtained; especially for high slew rates. However the purpose of the procedure is to produce values that can directly be compared to the DDR2 SDRAM input requirements. This is the reason why VTHvalues consistent with the DDR2 data sheet were chosen. Only if the measurement result is in the ran
35、ge of 1 V/ns to a little bit above 2 V/ns (differential), the accurancy is relevant. If the measured value is higher than 2 V/ns then this fact alone is enough to judge the DIMM performance with respect to system timing budgets.sr-VTH4VTH3tX+CH2VTH4()tX+CH2VTH3()-=SR+1N-sr+N=SR-1N-sr-N=JEDEC Publica
36、tion No. 152Page 94 Measurement of skew, jitter, slew rate and crosspoint voltage using single-ended probes4.1 Test equipment and setup Oscilloscope with min bandwidth of 3 GHz and 8 Gigasamples/s, Four single-ended probes with minimum bandwidth of 3 GHz resulting in a probe tip bandwidth of no less
37、 than 2.5 GHz PC2-3200/4300 JEDEC CRB Clock Reference Board (=CRB) by CST, DC Power Supply for 3.3 V to CRBFigure 3 Measurement setup with single-ended probesScopeChSingle-ended ProbesCLK SynthesizerCLK BufferSwitchDRAM Reg.PLLDIMMDRAMDRAMDRAMDRAMDRAMDRAMDRAMDRAMCLK Ref. NetsVddDRAM Reg.DRAMDRAMDRAM
38、DRAMDRAMDRAMDRAMDRAMJEDEC Publication No. 152Page 104 Measurement of skew, jitter, slew rate and crosspoint voltage using single-ended probes (contd)4.2 Preparation1) Allow instruments to warm up ( 20 min),2) Disconnect all probes from the scope,3) Perform the scope calibration,4) Attach probes 1 th
39、rough 4 to CH1 through CH4 of the scope,5) Perform the probes calibration1using the calibration fixture delivered with the scope,6) Determine equipment jitter as decribed in Annex A7) Attach wires of equal length a flight time of 50 ps per 12 mm is a reasonable estimate for the required correction.
40、The deltas need to be determined from the net topology and may be obtained from the DIMM design specification of the DIMM under test.tD+tX+WD2VTH()tX+WD1VTH()=tD-tX-WD2VTH()tX-WD1VTH()=tsk(LH)1N-tD+N=tsk(HL)1N-tD-N=JEDEC Publication No. 152Page 134 Measurement of skew, jitter, slew rate and crosspoi
41、nt voltage using single-ended probes (contd)4.5 JitterMeasure the jitter of WD2 using the same procedure as is used with a differential probe. Begin with step 3 in section 3.5 of the differential procedure and use WD2 instead of CH2.4.6 Slew rateMeasure the slewrate of WD2 using the same procedure a
42、s is used with a differential probe. Begin with step 1 in section 3.6 of the differential procedure and use WD2 instead of CH2.4.7 Crosspoint voltage1) Obtain by linear interpolation, the positive going time of zero-crossing for this difference signal Wdiff2 (by first finding the pair of points stra
43、ddling the 0V level, and then interpolating the time at which a line through them would have passed through zero).NOTE This is the same as step 1 in 4.4.2) With this time, using either (CK or CK) of the up-sampled channel waveforms, determine the voltage associated with this time. (a good consistenc
44、y check is to assure both upsampled channel waveforms yield the same result). This voltage is a single estimate for Vix+. 3) Repeat step 2, but for a negative going zero-crossing. The value obtained is a single estimate for Vix-.4) Repeat steps 1, 2 and 3, for the remaining events (zero crossings in
45、 the up-sampled difference waveform) forming an average value for each of the measurements, until an average of at least 100 values is obtained. These average values constitute the final values for the measurement of Vix+ and Vix- for this differential clock termination resistors on the DIMM. Record
46、 this value.JEDEC Publication No. 152Page 14Annex A Measuring equipment jitterThe basic idea is simple, and it is worth stating outside the definition of the precise steps comprising the method. It is to measure the jitter between two channels while presenting the same (analog) clock or data signal
47、to both of the input channels.The beauty of this concept is that it is simple and unpretentious. Everyone can easily grasp that at the actual point(s) probed, there can be no jitter between the point and itself. The only difference between the two channels is therefore the probes, the scope channel
48、amplifiers and the two different recording devices (presumably ADCs). As such, in a perfect world there should be no jitter, and therefore in the real world the jitter that is observed is jitter due to the measurement (apparatus and method) itself. Now, beware, that does not mean that the measured jitter “is” exactly the jitter introduced by the instrument. It is indeed closely related, but it s not exactly equal to this, as will be shown. Jitter m
