Zynq Ultrascale+ VCCINT_VCU decoupling

[jt94096]

I would like to understand the justification for the VCCINT_VCU decoupling listed as a requirement in DS583 v1.19 Table 1-12.

The table lists three 470uF polymer capacitors in addition to roughly 1000uF of ceramic capacitors. I would prefer not to use the polymer capacitors, mainly due to board space. The polymer caps won't help much with step changes in load and my supply already performs well in this respect with much less ceramic capacitance. Therefore I don't see what benefit the polymer caps would provide.

Thanks - Jason

[simon]

Hi @jt94096 

If you don't follow decoupling capacitors in table 1-12, you should do PI simulation to meet our spec(the performance of the decoupling networks presented here)

[jt94096]

Thanks Simon,

Does Xilinx have PDN sim results for one of their dev boards such as the ZCU106 that can be used for comparison? Or perhaps an impedance spectrum mask? I am not an expert on PDN simulation but I expect that the regulator performance also plays a role at the lower end of the spectrum.

Given the challenges and costs (in time and money) of obtaining accurate post-layout PDN sim data, we also want to have a sound engineering basis for the PDN design. Our regulator supplier has stated that they see no need for this much bulk decoupling and in fact it tends to be detrimental to the performance of the regulator.

Jason

[IAM_NOT_IAM]

if you have higher DC/DC switch frequency like 2MHz, i think you can reduce the number of 470uF cap