Palace Width Sweep: Parallel Simulation¶
Sweep the microstrip width from 5 to 50 um (5 points), run all simulations in parallel on GDSFactory+ cloud using the non-blocking API, then compare S11 and S21 across widths.
Requirements:
- IHP PDK:
uv pip install ihp-gdsfactory - GDSFactory+ account for cloud simulation
Define the sweep¶
Build components and configure simulations¶
import gdsfactory as gf
from ihp import LAYER, PDK, cells
from gsim.common.stack import get_stack
from gsim.palace import DrivenSim
PDK.activate()
stack = get_stack() # auto-detects active PDK
sims = []
for w in widths:
# Build component
c = gf.Component()
r1 = c << cells.straight_metal(length=1000, width=w)
r = c.get_region(layer=LAYER.TopMetal2drawing)
r_sized = r.sized(+20000)
c.add_polygon(r_sized, layer=LAYER.Metal1drawing)
c.add_ports(r1.ports)
# Configure simulation
sim = DrivenSim()
sim.set_output_dir(f"./palace-sim-w{w:.1f}")
sim.set_geometry(c)
sim.set_stack(stack)
# Surround the design with air. Without this the absorbing boundary sits
# directly on the passivation, microns above the trace, and adds large
# spurious loss and reflection.
sim.set_airbox(margin_x=50, margin_y=50, z_above=100, z_below=100)
for port in c.ports:
sim.add_port(
port.name, from_layer="metal1", to_layer="topmetal2", geometry="interlayer"
)
sim.set_driven(fmin=1e9, fmax=100e9, num_points=80)
sim.mesh(preset="default")
sims.append(sim)
print(f"Configured {len(sims)} simulations")
pyvirtualdisplay not available; continuing without Xvfb
Small conductor feature detected (2.000 um) may be under-resolved by refined_mesh_size=5.000 um. Pass auto_size=True to scale the mesh down.
Configured 5 simulations
Upload and start all jobs (non-blocking)¶
# Upload and start all jobs without waiting
job_ids = []
for sim in sims:
job_id = sim.run(wait=False, check_cache=True)
job_ids.append(job_id)
print(f"Started {len(job_ids)} jobs: {job_ids}")
Info : Reading 'palace-sim-w2.0/palace.msh'...
Info : 9203 nodes
Info : 69098 elements
Info : Done reading 'palace-sim-w2.0/palace.msh'
Info : Reading 'palace-sim-w2.0/palace.msh'...
Info : 9203 nodes
Info : 69098 elements
Info : Done reading 'palace-sim-w2.0/palace.msh'
Cache hit: reusing job 01a0f822-34d5-7fe3-b8e3-5d9fe8b3670e
Info : Reading 'palace-sim-w6.0/palace.msh'...
Info : 9838 nodes
Info : 74046 elements
Info : Done reading 'palace-sim-w6.0/palace.msh'
Info : Reading 'palace-sim-w6.0/palace.msh'...
Info : 9838 nodes
Info : 74046 elements
Info : Done reading 'palace-sim-w6.0/palace.msh'
Cache hit: reusing job 01a0f822-3ae8-7633-beef-bb96ec13bfdb
Info : Reading 'palace-sim-w10.0/palace.msh'...
Info : 9784 nodes
Info : 73826 elements
Info : Done reading 'palace-sim-w10.0/palace.msh'
Info : Reading 'palace-sim-w10.0/palace.msh'...
Info : 9784 nodes
Info : 73826 elements
Info : Done reading 'palace-sim-w10.0/palace.msh'
Cache hit: reusing job 01a0f822-41e6-7c51-9015-cbe238d204db
Info : Reading 'palace-sim-w14.0/palace.msh'...
Info : 9643 nodes
Info : 72801 elements
Info : Done reading 'palace-sim-w14.0/palace.msh'
Info : Reading 'palace-sim-w14.0/palace.msh'...
Info : 9643 nodes
Info : 72801 elements
Info : Done reading 'palace-sim-w14.0/palace.msh'
Cache hit: reusing job 01a0f822-47cb-7600-88ff-0bdf0d991b1d
Info : Reading 'palace-sim-w18.0/palace.msh'...
Info : 9918 nodes
Info : 74998 elements
Info : Done reading 'palace-sim-w18.0/palace.msh'
Info : Reading 'palace-sim-w18.0/palace.msh'...
Info : 9918 nodes
Info : 74998 elements
Info : Done reading 'palace-sim-w18.0/palace.msh'
Cache hit: reusing job 01a0f822-4ddd-71a3-a549-8f617d822423
Started 5 jobs: ['01a0f822-34d5-7fe3-b8e3-5d9fe8b3670e', '01a0f822-3ae8-7633-beef-bb96ec13bfdb', '01a0f822-41e6-7c51-9015-cbe238d204db', '01a0f822-47cb-7600-88ff-0bdf0d991b1d', '01a0f822-4ddd-71a3-a549-8f617d822423']
Wait for all jobs to complete¶
import gsim
# Poll all jobs concurrently, download and parse results
results = gsim.wait_for_results(job_ids)
Waiting for 5 jobs...
palace-5b5227f5 [####################] 100% complete elapsed 0m 00s
palace-37990c45 [####################] 100% complete elapsed 0m 00s
palace-2a9da1de [####################] 100% complete elapsed 0m 00s
palace-1215f504 [####################] 100% complete elapsed 0m 00s
palace-f9f6ddc2 [####################] 100% complete elapsed 0m 00s
Extracting results.tar.gz...
Downloaded 12 files to sim-data-palace-5b5227f5
Extracting results.tar.gz...
Downloaded 12 files to sim-data-palace-37990c45
Extracting results.tar.gz...
Downloaded 12 files to sim-data-palace-2a9da1de
Extracting results.tar.gz...
Downloaded 12 files to sim-data-palace-1215f504
Extracting results.tar.gz...
Downloaded 12 files to sim-data-palace-f9f6ddc2
Plot S11 and S21 comparison¶
import matplotlib.pyplot as plt
fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(6, 6))
for w, sp in zip(widths, results, strict=True):
ax1.plot(sp.freq, sp.s11.db, label=f"w={w:.1f} um")
ax2.plot(sp.freq, sp.s21.db, label=f"w={w:.1f} um")
ax1.set(
xlabel="Frequency (GHz)", ylabel="|S11| (dB)", title="S11 — Return Loss vs Width"
)
ax1.legend()
ax1.grid(True)
ax2.set(
xlabel="Frequency (GHz)", ylabel="|S21| (dB)", title="S21 — Insertion Loss vs Width"
)
ax2.legend()
ax2.grid(True)
plt.tight_layout()
