There is a good question online about ptpx peak power. With
set_power_analysis_options -waveform_interval 0.1
(100ps interval) peak power is 2.2w. With
set_power_analysis_options -waveform_interval 0.01
(10ps interval) peak power jumps to 3.2w.
Here is the script to run ptpx:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
set power_enable_analysis TRUE set power_analysis_mode time_based set link_library " * DBH_1830BD18BA_GESVT1P8V_ISO_TT_1P8V_25C.db DBH_1830BD18BA_BD18LVGPIO_ISO_INLINE_TT_5P0V_25C.db b2hd111am4bi_8192x8x16_TT_25_1.80_TYP.db db2hd111am4bi_256x8x16_TT_25_1.80_TYP.db db2vs110am4bi_512x8x32_TT_25_1.80_TYP.db kpds18bcd64_r08w01_300c3pl_1v0_typ_1_1_172.db" read_verilog ../tmax/test_top_1222_s.v current_design test_top link read_sdc ./MCU.sdc read_parasitics ../sim_post/test_top_1222_tt.spef update_timing check_timing report_timing read_vcd -strip_path test_top_test/dut ../tmax/test_top_1225_p.vcd set_power_analysis_options -waveform_format fsdb -waveform_output mac.fsdb -include all_with_leaf -sdpd_tracking enabled check_power update_power report_power -verbose |
Here is ptpx result:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 |
**************************************** Report : Time Based Power -verbose Design : test_top Version: K-2015.06 Date : Fri Dec 29 09:28:08 2017 **************************************** Sampling Interval: 0.001 ns Library(s) Used: DBH_1830BD18BA_GESVT1P8V_ISO_TT_1P8V_25C (File: /home/worklib/DBH_STD_1830BD18BA_GESVT1P8V_ISO_16Q3_V2.20.10/LIBERTY/DBH_1830BD18BA_GESVT1P8V_ISO_TT_1P8V_25C.db) DBH_1830BD18BA_BD18LVGPIO_ISO_INLINE_TT_5P0V_25C (File: /home/worklib/DBH_IO_1830BD18BA_BD18LVGPIO_ISO_16Q4_V2.21.10/LIBERTY/DBH_1830BD18BA_BD18LVGPIO_ISO_INLINE_TT_5P0V_25C.db) b2hd111am4bi_8192x8x16_TT_25_1.80_TYP (File: /home/worklib/b2hd111am4bi_8192x8x16/lib/b2hd111am4bi_8192x8x16_TT_25_1.80_TYP.db) db2hd111am4bi_256x8x16_TT_25_1.80_TYP (File: /home/worklib/db2hd111am4bi_256x8x16/lib/db2hd111am4bi_256x8x16_TT_25_1.80_TYP.db) db2vs110am4bi_512x8x32_TT_25_1.80_TYP (File: /home/worklib/db2vs110am4bi_512x8x32/lib/db2vs110am4bi_512x8x32_TT_25_1.80_TYP.db) typ (File: /home/worklib/KPDS18BCD64-R08W01-300C3PL-1V0_031617_4977/KPDS18BCD64-R08W01-300C3PL-1V0_031617_4977/models/kpds18bcd64_r08w01_300c3pl_1v0_typ_1_1_172.db) Operating Conditions: TT25 Library: DBH_1830BD18BA_GESVT1P8V_ISO_TT_1P8V_25C Wire Load Model Mode: top <no wire load model is set> Power-specific unit information : Voltage Units = 1 V Capacitance Units = 1 pf Time Units = 1 ns Dynamic Power Units = 1 W Leakage Power Units = 1 W Attributes ---------- i - Including register clock pin internal power u - User defined power group Internal Switching Leakage Total Power Group Power Power Power Power ( %) Attrs -------------------------------------------------------------------------------- clock_network 9.546e-04 2.653e-04 5.850e-08 1.220e-03 (19.10%) i register 1.118e-03 3.168e-04 6.348e-07 1.435e-03 (22.47%) combinational 7.329e-04 6.901e-04 4.905e-07 1.423e-03 (22.28%) sequential 0.0000 0.0000 0.0000 0.0000 ( 0.00%) memory 0.0000 0.0000 0.0000 0.0000 ( 0.00%) io_pad 0.0000 0.0000 0.0000 0.0000 ( 0.00%) black_box 2.283e-03 1.332e-05 1.308e-05 2.310e-03 (36.16%) Net Switching Power = 1.285e-03 (20.12%) Cell Internal Power = 5.089e-03 (79.66%) Cell Leakage Power = 1.426e-05 ( 0.22%) --------- Total Power = 6.388e-03 (100.00%) X Transition Power = 3.776e-05 Glitching Power = 2.273e-08 Peak Power = 3.2201 Peak Time = 100247.114 |
It should be expected that lots of times with smaller interval peak power is higher. Indeed peak power is the average power across analysis interval. So with larger interval, peak power is averaged out over a longer period.
I saw this in lab measurement too. We use current meter to check current profile. Peak power will be higher if we set interval to be smaller. Same reason: peak power is the average power in interval.
But is it a real issue? It depends your chip’s decoupling cap design. Although peak power is high, it only last for a very short period of time. So many times on-chip caps can provide that peak current. Obviously out-of-chip decoupling cap is needed for current lasting a long time.
In the report, clock network consumes 19% pwr. Is it typical or a little bit high?