PDN IR Drop - Terminal for calculation?

[AkaBlur]

I've been working with a commercial PDK and want to use OpenROAD for my design. Recently I also looked into parasitic extraction using OpenRCX, which works quite well.

However, when analyzing the powergrid of my chip, I noticed that the PDN get's connected on some predefined way when I analyze the IR drop. I have prepared an example with the nangate45/jpeg design from OpenROAD and just added a manual PDN configuration for the jpeg example:

####################################
# global connections
####################################
add_global_connection -net {VDD} -inst_pattern {.*} -pin_pattern {^VDD$} -power
add_global_connection -net {VDD} -inst_pattern {.*} -pin_pattern {^VDDPE$}
add_global_connection -net {VDD} -inst_pattern {.*} -pin_pattern {^VDDCE$}
add_global_connection -net {VSS} -inst_pattern {.*} -pin_pattern {^VSS$} -ground
add_global_connection -net {VSS} -inst_pattern {.*} -pin_pattern {^VSSE$}
global_connect
####################################
# voltage domains
####################################
set_voltage_domain -name {CORE} -power {VDD} -ground {VSS}
####################################
# standard cell grid
####################################
define_pdn_grid -name {grid} -voltage_domains {CORE}
add_pdn_ring -grid {grid} -layer {metal7 metal8} \
    -spacings {2.8} -widths {5.6} -core_offset {1.4}
add_pdn_stripe -grid {grid} -layer {metal1} \
    -width {0.17} -pitch {2.4} -offset {0} -followpins -extend_to_core_ring
add_pdn_stripe -grid {grid} -layer {metal4} \
    -width {0.48} -pitch {28.0} -offset {2}
add_pdn_stripe -grid {grid} -layer {metal7} \
    -width {1.40} -pitch {15.0} -offset {2} -extend_to_core_ring

add_pdn_connect -grid {grid} -layers {metal1 metal4}
add_pdn_connect -grid {grid} -layers {metal4 metal7}
add_pdn_connect -grid {grid} -layers {metal7 metal8}

It is basically the same PDN as in the GCD example with an added periphery ring around the core.

---

Now when analyzing IR drop I notice that the lower left corner get's almost no IR drop and the upper right has the highest. This is already the case in the ring itself. It seems like the IR drop get's calculated from the lower left corner as terminal (actually the 0.0 V position is slighty above the corner).

Upper Corner

Lower Corner

---

Is there a way to put the complete ring at a fixed potential, or at least define the positions where my supply get's connected? I already looked at the OpenRCX docu and pdnsim but couldn't find anything regarding this.

[maliberty]

Most likely you are lacking pin on your pdn grid. In the platform we use -pins {metal7} to define those (see

OpenROAD-flow-scripts/flow/platforms/nangate45/grid_strategy-M1-M4-M7.tcl

Line 17 in 99312fa

define_pdn_grid -name {grid} -voltage_domains {CORE} -pins {metal7}

). Without pins you get a fairly arbitrary bump assumption.

[AkaBlur]

Thanks @maliberty,
that was exactly what I missed.

Regarding this configuration, the whole metal layer will be put at the given potential, right?
When trying to connect only a specific strip of the PDN I would need a dummy metal strip acting as pin on a completely different metal.

[maliberty]

All shapes on that layer afaik - @gadfort any more color?