wafer.space Community - @mole99 Is there any ESD protection in

4:56 a.m.

@Alf The template project uses the pad cells as provided by the PDK. All of the digital pads have ESD protection. See here for example: https://gf180mcu-pdk.readthedocs.io/en/latest/IPs/IO/gf180mcu_fd_io/tri_state_2.html This should cover HBM, CDM and MM: https://gf180mcu-pdk.readthedocs.io/en/latest/IPs/IO/gf180mcu_fd_io/esd.html Each power/ground pad has one ESD protection circuit, the corners have two: https://gf180mcu-pdk.readthedocs.io/en/latest/IPs/IO/gf180mcu_fd_io/power.html The analog pad has only HBM protection diodes, additional ESD protection must be added: https://gf180mcu-pdk.readthedocs.io/en/latest/IPs/IO/gf180mcu_fd_io/analog.html

Great, thanks!

Hi again, just to clarify, are the "input" pins in the template what you refer to as analog inputs?

8:52 p.m.

Also, the clock and reset line are a bit oddly placed in the package, down in the corner. A mid chip location feels like a better place for routing and minimize clock/reset delta over the chip. Was there some particular thinking of the location?

Hi @Alf, no, the NUM_INPUT parameter controls the number of gf180mcu_fd_io__in_c pads, which is a digital input pad. https://github.com/wafer-space/gf180mcu-project-template/blob/a4805b8bb19a960c107ed6f199936e6442d52207/src/chip_top.sv#L242C9-L242C29 The analog pad is gf180mcu_fd_io__asig_5p0. I will add another parameter NUM_ANALOG that instantiates analog pads. No, there was no particular thinking of the location besides having them at pad_0/pad_1. You can easily swap pins by editing the LibreLane configuration: https://github.com/wafer-space/gf180mcu-project-template/blob/main/librelane/config.yaml It might make sense to move those closer to the center in the template. Just do understand this better: Does this really make a difference for clock/reset delta? The pads are directly next to each other. Could it be because the clock root buffer for the clock tree is located somewhere in the middle of the chip, and the reset does not form a tree?

Hi, tnx! I was not thinking of delta between clock and reset but overall clock propagation. Perhaps it is less critical for these nodes and frequencies but in general it is desirable to try and clock the whole chip as simultaneous as possible. If the clock comes in at lower corner then those blocks will clock sooner than the ones at the top.

Hi Alf, that won't happen since OpenROAD builds up a balanced clock tree that is fed from root clock buffer which will be located somewhere in the middle of the chip. You can view the clock tree using the OpenROAD GUI. OpenSTA runs to ensure that setup/hold is met for all flip-flops. So the only difference would be the distance from the pad to the root clock buffer. Of course, it is better if that distance is smaller, but I don't know how critical that really is.