AND9017 ON Semiconductor P1P8160A: Spread Spectrum Clock Solution to Address Peak EMI in AMD Graphics Platforms with GDDR Memory http://onsemi.com APPLICATION NOTE Prepared by: Jay Hsu ON Semiconductor Abstract Active Spread Spectrum (SS) technology modulates the frequency to suppress peak EMI and also ensures signal integrity is not degraded in the GDDR clock signal by preserving the rise/fall times. All signals derived from this SS clock will be modulated and peak EMI levels are suppressed. SSCG technology is an optimal solution to solve this peak EMI problem. P1P8160A is the ideal SSCG timing solution for AMD’s Northern Islands and Evergreen series graphic cards with GDDR3 and GDDR5 memory support. The P1P8160A provides two clocks: a 27 MHz reference clock to GPU core, and a 100 MHz SS clock to the GDDR memory. Graphics Processing Units (GPU) running with Graphics Double Data Rate memories, Version 3 and 5 (GDDR3/GDDR5), could potentially cause peak Electro−Magnetic interference (EMI) failure at compliance certification. This article is intended to address and solve this peak EMI failure using ON Semiconductor’s new Spread Spectrum Clock Generator (SSCG) device*P1P8160A, specifically designed for use with AMD Northern Islands/ Evergreen series GPUs. Introduction GDDR3 and GDDR5 memories are commonly used in high performance graphics cards and run at very high frequencies. If memory capacity goes up to 1 GB GDDR3 and GDDR5, it requires eight memory modules with eight pairs of differential memory clocks from the GPU to the GDDR3 and GDDR5 memory modules. These clocks are a major source of EMI problems. The EMI certification tests come very late in the product cycle and can cause unexpected delays if the system fails EMI compliance. © Semiconductor Components Industries, LLC, 2011 July, 2011 − Rev. 1 Using P1P8160A for AMD Graphic Chips Platforms P1P8160A uses a 27 MHz fundamental crystal or an external reference clock to generate a 27 MHz (no SS) and a 100 MHz (SS) clock output. As shown in Figure 1, the 100 MHz SS clock will supply the reference clock to the internal GDDR controller. 1 Publication Order Number: AND9017/D AND9017 Figure 1. P1P8160A for AMD Graphic Chips Application P1P8160A: Key Functional Characteristics allows to user to customize the modulation depth such that P1P8160A provides the maximum EMI benefit without violating system timing specifications. The deviation settings of the P1P8160A are shown in Table 1. P1P8160A provides eight different user selectable SS deviation options to control the extent of frequency modulation by setting Pin 3 (SS1%) and Pin 7 (SS2%) to one of three levels (High, Medium, or Low). This flexibility Table 1. P1P8160A: FREQUENCY DEVIATION SELECTION SS2% (Pin 3) SS1% (Pin 7) Deviation at 100 MHz (%) (Pin 5) L L SSOFF L M −0.5 L H −2.5 M L −0.25 M M −0.75 M H −1 H L −1.5 H M −2 H H −3 P1P8160A: AC Electrical Characteristics The P1P8160A AC Electrical Characteristics for the 27 MHz and 100 MHz output clocks are shown in Table 2. These clocks meet AMD’s GPU clock specifications. Table 2. P1P8160A OUTPUT CLOCKS AC ELECTRICAL CHARACTERISTICS Symbol fOUT Parameter Min Typ ModOUT Clock frequency (SS1% & SS2% = 0) (Tolerance: ±30 ppm) 100 RefOUT Clock frequency (Tolerance: ±30 ppm) 27 Max tLH, tHL RefOUT Rise and Fall time, CL = 15 pF (Measured between 20% to 80%) 1.25 2 tLH, tHL ModOUT Rise and Fall time, CL = 15 pF (Measured between 20% to 80%) 1.25 1.75 TDCOUT Output Clock Duty Cycle 50 55 45 http://onsemi.com 2 Unit MHz ns % AND9017 Table 2. P1P8160A OUTPUT CLOCKS AC ELECTRICAL CHARACTERISTICS Symbol Typ Max TJC Cycle−Cycle Jitter (For ModOUT, RefOUT) 125 200 TJL Long Term Jitter (10k cycles), 27 MHz, RefOUT 150 300 Long Term Jitter (10k cycles), 100 MHz ModOUT (SSOFF) 350 600 32 33 MF Parameter Min Modulation Frequency 31 100 MHz Clock Output EMI Comparison Unit ps kHz without SS. As seen in Table 3, the EMI reduction in the 30 MHz− 1.3 GHz frequency range is 2 − 10 dB better with −0.5% down spreading, when compared to SS turned off. Device settings are: VDD = 3.3 V, Deviation = −0.5% and Modulation rate = 32 kHz. Application testing was done using Anritsu MS2711D Spectrum Analyzer. The plots in Figure 2 display the results of EMI bench testing using the P1P8160A. This test is performed on a characterization board and the EMI is measured on the 100 MHz output clock using a spectrum analyzer. The results show the difference in peak EMI levels, with and Figure 2. 100 MHz EMI Spectrum Analyzer performance Table 3. EMI PERFORMANCE TABLE Freq EMI P1P8160A SSOFF (dBm) EMI P1P8160A SSON = −0.5% (dBm) EMI Reduction (dB) 100 MHz −5.16 −7.46 2.3 300 MHz −10.58 −15.73 5.15 500 MHz −13.01 −19.51 6.5 700 MHz −16.17 −23.33 7.16 900 MHz −18.33 −26.58 8.25 1100 MHz −23.48 −32.55 9.07 1300 MHz −30.71 −41.63 10.92 Validation in Barts (HD6870) System The P1P8160A meets the AMD GPU clock specifications. Measurements were performed using a Tektronix TDS6604B Digital Oscilloscope, P7260 Active Probes and TDSJITV3 Jitter Analysis Software. This case study is based on an AMD Radeon HD6870 (DDR5 GPU) with a P1P8160A. The 27 MHz output clock AC electrical characteristics are shown in Table 4 and the output waveforms are shown in Figures 3a, 3b and 3c. Table 4. APPLICATION TEST FOR P1P8160A: 27 MHZ OUTPUT CLOCK P1P8160A Rise Time Fall Time 1.25 ns 1.34 ns Duty Cycle 48% (min) Cycle−Cycle Jitter 51% (max) −183 ps (min) http://onsemi.com 3 139 ps (max) Long Term Jitter 200 ps AND9017 a. Rise/Fall Time, Duty Cycle b. Cycle−Cycle Jitter c. Long Term Jitter Figure 3. Table 5 and Figures 4a, 4b and 4c show the 100 MHz output clock Rise/ Fall time, Duty Cycle, Cycle−Cycle jitter and Long term jitter measured values and waveforms. Long term jitter is with SS turned OFF and the other parameters are measured with SS ON (−0.5% down spreading). The P1P8160A meets the AMD GPU clock specifications. Table 5. APPLICATION TEST FOR P1P8160A: 100 MHz OUTPUT CLOCK P1P8160A Rise Time Fall Time 0.785 ns 0.838 ns a. Rise/Fall Time, Duty Cycle Duty Cycle 48% (min) Cycle−Cycle Jitter 52% (max) b. Cycle−Cycle Jitter −97.4 ps (min) 85.8 ps (max) Long Term Jitter (SSOFF) 300 ps c. Long Term Jitter Figure 4. lowered by 9.5 dB when −0.5% (down spreading) SS is used at 1.225 GHz. Testing was performed using Anritsu MS2711D Spectrum Analyzer. The EMI Spectrum Analyzer performance comparison with the original non−SS DDR5 clock and a DDR5 SS Clock is shown in Figure 5. The results show that the EMI peak is http://onsemi.com 4 AND9017 Reduction 9.5 dB Figure 5. DDR Clock EMI Performance 2. On the 100MHz output clock, the measured Cycle*Cycle Jitter values are *97.4 ps(min) to 85.8 ps (max) and long term jitter is 300 ps. The peak EMI comparison on the GDDR5clock shows that the EMI is reduced by 9.5 dB at 1.225 GHz when −0.5% down spread is used. The P1P8160A device is an efficient and cost effective way to reduce the EMI peaks in systems that use GDDR3 and GDDR5 memories. Summary Based on this case study, the Peak EMI benefits that can be expected with −0.5%(down spreading) SS when using the P1P8160A are: 1. 2 -3 dB, In 30−200 MHz frequency range 2. 5-9 dB, In 200−900 MHz frequency range 3. 9-10dB, In 1100−1300 GHz frequency range The P1P8160A was designed to meet AMD clock specifications and this case study highlights the benefits of using the P1P8160A in AMD GPUs. The key jitter specs are as follows: 1. On the 27 MHz output clock, the measured Cycle*Cycle Jitter values are −183 ps(min) to 139 ps(max) and long term jitter is 200 ps. ON Semiconductor and are registered trademarks of Semiconductor Components Industries, LLC (SCILLC). SCILLC reserves the right to make changes without further notice to any products herein. SCILLC makes no warranty, representation or guarantee regarding the suitability of its products for any particular purpose, nor does SCILLC assume any liability arising out of the application or use of any product or circuit, and specifically disclaims any and all liability, including without limitation special, consequential or incidental damages. “Typical” parameters which may be provided in SCILLC data sheets and/or specifications can and do vary in different applications and actual performance may vary over time. All operating parameters, including “Typicals” must be validated for each customer application by customer’s technical experts. SCILLC does not convey any license under its patent rights nor the rights of others. SCILLC products are not designed, intended, or authorized for use as components in systems intended for surgical implant into the body, or other applications intended to support or sustain life, or for any other application in which the failure of the SCILLC product could create a situation where personal injury or death may occur. Should Buyer purchase or use SCILLC products for any such unintended or unauthorized application, Buyer shall indemnify and hold SCILLC and its officers, employees, subsidiaries, affiliates, and distributors harmless against all claims, costs, damages, and expenses, and reasonable attorney fees arising out of, directly or indirectly, any claim of personal injury or death associated with such unintended or unauthorized use, even if such claim alleges that SCILLC was negligent regarding the design or manufacture of the part. SCILLC is an Equal Opportunity/Affirmative Action Employer. This literature is subject to all applicable copyright laws and is not for resale in any manner. PUBLICATION ORDERING INFORMATION LITERATURE FULFILLMENT: Literature Distribution Center for ON Semiconductor P.O. Box 5163, Denver, Colorado 80217 USA Phone: 303−675−2175 or 800−344−3860 Toll Free USA/Canada Fax: 303−675−2176 or 800−344−3867 Toll Free USA/Canada Email: [email protected] N. American Technical Support: 800−282−9855 Toll Free USA/Canada Europe, Middle East and Africa Technical Support: Phone: 421 33 790 2910 Japan Customer Focus Center Phone: 81−3−5773−3850 http://onsemi.com 5 ON Semiconductor Website: www.onsemi.com Order Literature: http://www.onsemi.com/orderlit For additional information, please contact your local Sales Representative AND9017/D