AM2631: AM263 hardware design: AC PDN Z target

Hi Zane,

Sorry for the delay. Please find my responses below

Zane Wei said:

Q1.What is the source of the PDN control frequency of the three modes of operation Fmax?

I will reach out to our experts and get details on this

Zane Wei said:

Q2. Three different modes of operation Ztarget numerically, baseline1 frequency (≤200MHz) and impedance range (≤12mΩ) cover the remaining two modes of operation, is it directly according to the standard of baseline1 to control PDN impedance?

Apologies, I am not sure if I understand this correctly. But if you are referring to the 3 different Transient cases, the three transient cases are as follows

VDD BASELINE1 - This is just a very pessimistic transient model where it assumes change of power consumption from Zero to Max power consumption. This is not a realistically possible scenario and hence is not considered.

VDD XTAL_PLL1 - This is a realistic scenario where the Clock source changes from XTAL(25MHz) to PLL(400MHz).

VDD WFI1 - This is another realistic scenario where in the 4 R5F cores come out of waiting for Interrupt state.

In the above, the second one is chosen since that is the most power consuming, practical use case.

Zane Wei said:

Q3.1: The official version of PDN impedance does not meet the standard of less than 12mΩ within 200M?

Q3.2: Several different PDN curves are used to lower the anti-resonant point by changing the decap capacitance, is it necessary to optimize the recommended decap configuration in the actual design?

Q3.3: Why is the Ztarget set here 36mΩ?

As mentioned earlier the VDD XTAL_PLL1 transient is considered since that is the most power consuming, practical use case. Hence a target of 36mΩ is taken.

Zane Wei said:

Q3.4: Despite the relaxation of Ztarget to 36mΩ, the impedance in the low and high frequency bands is still exceeded?

Lower bands are not much of a problem, since the Buck operating frequency would accommodate the variation of voltage and account for it.

Higher bands(about 50-80MHz or higher) are not optimizable on PCB and must be optimized on package.

Zane Wei said:

Q4. Please to provide the specific control standard, frequency range and Ztarget of 1.2V nuclear power.

We recommend the same 36mΩ target based on VDD XTAL_PLL1 transient use case.

Zane Wei said:

Q5: The 3.3V supply current is small and the PDN impedance standard is less stringent, but our design does not differentiate between digital and analog power supplies, but rather directly combines. Trouble provides the combined PDN impedance control standard, frequency range and Ztarget

Can you please elaborate more on this?

Thanks,

Tejas Kulakarni

Hi Jiang,

jiang chayce said:

removing decoupling capacitors in order of their loop inductance values from highest to lowes

Yes, this is the right way of approaching this solution, but it highly depends on the stack-up, layouting, polygon pours or power planes of that particular board. The example given in the AM26x Hardware design guideline is from our EVM designs.

Hence, I would not be able to comment on as to whether only once capacitor is sufficient or not. If there is confidence in your simulation setup and the results concur to the target impedance needed, I would say it should be good enough.

Thanks,
Tejas Kulakarni