Set up OptimizePI

Hi Kaydee, you can model the VRM at the starting point of the net VDD_MB and set the observation point at U1_VDD_GND. Additionally, you may need to ensure a return path exists for accurate impedance calculation.

Hi Sumith

Tool / version

• Sigrity 19.0, OptimizePI flow

• Board has been used in PowerDC already and DC path is correct.

Topology

• External power board provides VDD through a connector J5 on my board.

• Net from connector is called VDD_MB.

• From J5 the power goes through two resistors R10 (DCR ≈ 0.2 mΩ each) to the net VDD.

• Net VDD powers the main IC U1.

• So the path is: external VRM → J5 (VDD_MB) → R101/R102 → VDD → U1.

OptimizePI setup (follow your guideline)

1. In Net Manager / P-G classification wizard, both VDD_MB and VDD are classified as PowerNets, GND as GroundNet.

2. In Define VRM Models, I created a VRM model and assigned:

   • Power net = VDD_MB

   • Return net = GND

3. In Impedance Observations, I created two observation points:

   • J5_VDD_MB_GND

   • U1_VDD_GND (Model Name = @Open@)

4. Analysis type = What-If, frequency range 100 kHz – 1 GHz.

Problem / plot

• The self-impedance at J5_VDD_MB_GND (red curve) is flat and very low (~1 mΩ) across the whole band, which looks like the VRM impedance – this seems correct.

• But the self-impedance at U1_VDD_GND (blue curve) starts at very high impedance (~10³ Ω) at low frequency and then decreases with 1/ω, with a resonance dip around a few hundred MHz.

• This shape looks like “no VRM connected, only planes + decaps”.

• In other words, from the U1 node OptimizePI appears not to “see” the VRM at J5, even though in the physical design they are connected through R101/R102 and I can simulate DC IR drop in PowerDC.

Question

What is the correct way to set up the VRM and nets so that the VRM at VDD_MB is seen from the load node U1_VDD_GND?

• Do I need to do anything special in Select Nets / PowerTree in this older (19.0) OptimizePI flow to make VDD_MB and VDD part of the same rail?

• Could component filtering be removing R101/R102 or otherwise breaking the connection between VDD_MB and VDD in the OptimizePI circuit?

Hi Sumith

Tool / version

• Sigrity 19.0, OptimizePI flow

• Board has been used in PowerDC already and DC path is correct.

Topology

• External power board provides VDD through a connector J5 on my board.

• Net from connector is called VDD_MB.

• From J5 the power goes through two resistors R10 (DCR ≈ 0.2 mΩ each) to the net VDD.

• Net VDD powers the main IC U1.

• So the path is: external VRM → J5 (VDD_MB) → R101/R102 → VDD → U1.

OptimizePI setup (follow your guideline)

1. In Net Manager / P-G classification wizard, both VDD_MB and VDD are classified as PowerNets, GND as GroundNet.

2. In Define VRM Models, I created a VRM model and assigned:

   • Power net = VDD_MB

   • Return net = GND

3. In Impedance Observations, I created two observation points:

   • J5_VDD_MB_GND

   • U1_VDD_GND (Model Name = @Open@)

4. Analysis type = What-If, frequency range 100 kHz – 1 GHz.

Problem / plot

• The self-impedance at J5_VDD_MB_GND (red curve) is flat and very low (~1 mΩ) across the whole band, which looks like the VRM impedance – this seems correct.

• But the self-impedance at U1_VDD_GND (blue curve) starts at very high impedance (~10³ Ω) at low frequency and then decreases with 1/ω, with a resonance dip around a few hundred MHz.

• This shape looks like “no VRM connected, only planes + decaps”.

• In other words, from the U1 node OptimizePI appears not to “see” the VRM at J5, even though in the physical design they are connected through R101/R102 and I can simulate DC IR drop in PowerDC.

Question

What is the correct way to set up the VRM and nets so that the VRM at VDD_MB is seen from the load node U1_VDD_GND?

• Do I need to do anything special in Select Nets / PowerTree in this older (19.0) OptimizePI flow to make VDD_MB and VDD part of the same rail?

• Could component filtering be removing R101/R102 or otherwise breaking the connection between VDD_MB and VDD in the OptimizePI circuit?

Thank you very much @Sumith

Topology

• External power board provides VDD through a connector J5 on my board.

• Net from connector is called VDD_MB.

• From J5 the power goes through two resistors R10(DCR ≈ 0.2 mΩ each) to the net VDD.

• Net VDD powers the main IC U1.

• So the path is: external VRM → J5 (VDD_MB) → R101/R102 → VDD → U1.

OptimizePI setup (follow your guild line)

1. In Net Manager / P-G classification wizard, both VDD_MB and VDD are classified as PowerNets, GND as GroundNet.

2. In Define VRM Models, I created a VRM model and assigned:

   • Power net = VDD_MB

   • Return net = GND

3. In Impedance Observations, I created two observation points:

   • J5_VDD_MB_GND

   • U1_VDD_GND

4. Analysis type = What-If, frequency range 100 kHz – 1 GHz.

Problem / plot

• The self-impedance at J5_VDD_MB_GND (red curve) is flat and very low (~1 mΩ) across the whole band, which looks like the VRM impedance – this seems correct.

• But the self-impedance at U1_VDD_GND (blue curve) starts at very high impedance (~10³ Ω) at low frequency and then decreases with 1/ω, with a resonance dip around a few hundred MHz.

• This shape looks like “no VRM connected, only planes + decaps”.

• In other words, from the U1 node OptimizePI appears not to “see” the VRM at J5, even though in the physical design they are connected through R10 and I can simulate DC IR drop in PowerDC.

Question

What is the correct way to set up the VRM and nets so that the VRM at VDD_MB is seen from the load node U1_VDD_GND?

• Do I need to do anything special in Select Nets / PowerTree in this older (19.0) OptimizePI flow to make VDD_MB and VDD part of the same rail?

• Could component filtering be removing R101/R102 or otherwise breaking the connection between VDD_MB and VDD in the OptimizePI circuit?

Thank you very much Sumith,

Topology

• External power board provides VDD through a connector J5 on my board.

• Net from connector is called VDD_MB.

• From J5 the power goes through two resistors R10(DCR ≈ 0.2 mΩ each) to the net VDD.

• Net VDD powers the main IC U1.

• So the path is: external VRM → J5 (VDD_MB) → R101/R102 → VDD → U1.

OptimizePI setup (follow your guild line)

1. In Net Manager / P-G classification wizard, both VDD_MB and VDD are classified as PowerNets, GND as GroundNet.

2. In Define VRM Models, I created a VRM model and assigned:

   • Power net = VDD_MB

   • Return net = GND

3. In Impedance Observations, I created two observation points:

   • J5_VDD_MB_GND

   • U1_VDD_GND

4. Analysis type = What-If, frequency range 100 kHz – 1 GHz.

Problem / plot

• The self-impedance at J5_VDD_MB_GND (red curve) is flat and very low (~1 mΩ) across the whole band, which looks like the VRM impedance – this seems correct.

• But the self-impedance at U1_VDD_GND (blue curve) starts at very high impedance (~10³ Ω) at low frequency and then decreases with 1/ω, with a resonance dip around a few hundred MHz.

• This shape looks like “no VRM connected, only planes + decaps”.

• In other words, from the U1 node OptimizePI appears not to “see” the VRM at J5, even though in the physical design they are connected through R10 and I can simulate DC IR drop in PowerDC.

Question

What is the correct way to set up the VRM and nets so that the VRM at VDD_MB is seen from the load node U1_VDD_GND?

• Do I need to do anything special in Select Nets / PowerTree in this older (19.0) OptimizePI flow to make VDD_MB and VDD part of the same rail?

• Could component filtering be removing R101/R102 or otherwise breaking the connection between VDD_MB and VDD in the OptimizePI circuit?