Introduction here. This template is just to give you a starting point. Feel free to add more sections, remove sections, or change it in any way that makes sense to you.
This section will summarize all a block builder needs to know about how to interpret the additional delta-T report. Further background information can be found later on in the "BACKGROUND" section.
Encounter signal_em violations can be waived if (1) Irms < 1.5 * Ilimit AND (2) DeltaT < 5C
The avo/signal_em/ff.rc_wc/rpts/deltaT.net_specific.rpt.gz report is broken into five different sections. Each is described below:
This section lists all of the signal-em violation that fail requirement (1) above (i.e. Irms < 1.5 * Ilimit). The TD group was not comfortable waiving anything that had an rms current so much greater than the limit. Any nets listed in this section MUST have their EM issues fixed.
This section lists any signal-em violating nets that overlap IP. Unfortunately there is currently no way to determine the power-contributing elements of the IP abstract and, thus, we don't feel we can safely waiver nets that cross over IP. Any nets listed in this section MUST have their EM issues fixed.
This section lists "via-areas" (via-area definition can be found in the BACKGROUND section) of the design that fail requirement (2) above (i.e. had delta-T greater than 5C). If there are no violations listed in this report and none of the previous sections identified potential risks, there are no signal-EM violations that need to be fixed.
This section details the delta-T for each of the signal-em violating nets that were extracted from the violations report. The worst delta-T is reported regardless of failure status. If there is an asterisk (*) next to the MaxDT(C) value, that means the net in question crosses over IP and, thus, falls into the "Specified nets that overlap IP" category. Any nets that have a delta-T greater than 5C MUST have their EM issues fixed.
This section simply reports toggle factors, max deltaT and min/max power observed for the via-areas involved with the specified signal-em violating nets. There is nothing actionable in this section.
To review deltaT violations, review the deltaT.net_specific.power_components.rpt.gz report for source(s) of primary joule heating (delta-T). All via-areas that failed delta-T have a section in the power_components report. Below is an example entry:
Via-area: 92.4,1037.96 Delta-T: 5.142 C Total Power: 23.008 W/mm^2 Net: L4_clk_725__LEQ1_T00P1S1R2I20B0, power: 21.520 W/mm^2 (93.531 %) Net: xl_rb_class_369__agPORTISO, power: 0.178 W/mm^2 (0.775 %) Net: FE_OFN187426_wireB_sw_bm2_fwd_105__agPORTISO, power: 0.123 W/mm^2 (0.535 %) Net: FE_OFN158528_mtc_xl_core_inst0_q_q_hash_tbl0_rd_data_0_53, power: 0.104 W/mm^2 (0.454 %) Net: xl_rb_class_304__agPORTISO, power: 0.090 W/mm^2 (0.389 %) Net: FE_OFN122931_wireB_sw_nq2_rwinst_154__agPORTISO, power: 0.086 W/mm^2 (0.372 %) ... Inst: ip_flop_inst14_d_out_reg_277_, power: 0.000 W/mm^2 (0.000 %) Inst: ip_flop_inst14_d_out_reg_304_, power: 0.000 W/mm^2 (0.000 %)
Note that the L4_clk* net contributes to over 93% of the overall power for this violating "via-area". Often times if the primary contributer(s) can be addressed, the rest of the signals in the via-area will not yield a delta-T failure.
John to add new information
In order to ensure metal and, thus, EM reliability, it is important that wire temperature not exceed the worst case specification by more than 5.0-degrees C. An increase in temperature in the wire is due the the power (RI^2) dissipation in the wire. An increased temperature will decrease EM reliability in a particular area of the die (referred to as a 'via area' which is defined as the area in which there is an average of 1 via23 power grid via-array, for 40nm, this area is 3.08u x 1.54u whose reported location is in the center of this via-area) , specifically in the power distribution. In other words, the lower the temperature the higher the metal reliability. The 'agReportViaAreaDeltaT' command checks that the total change in temperature per "via area" does not exceed the 5.0-degrees C limit. The script also assumes that the Iavg (AC lines) is effectively zero. If either of these is not the case, further analysis will need to be performed outside this script.
Unfortunately, the existing signal-em analysis tools (Encounter) must make some pessimistic assumptions so as to check signal-em on a net-by-net basis. Instead of checking Joule heating directly, it checks for current that exceeds a max current value. The added delta-T analysis was developed so as to be able to waive some of the overly pessimistic assumptions that the signal-em checking tool has to make.
Special thanks to Kevin Woolf for contributing the majority of this document as an email to his fellow block builders.