When there is room, there’s less reason to compact the layout.

There is one simple way to build in some room for changes without having a negative effect on performance while making the layout more flexible to “stretch”, to add in devices, or to make metal only changes to create new devices from old ones.

Don’t merge sources between devices, keeping each device aligned to itself.

What do I mean? For a basic illustration below, the two different inverters can merge sources if the space saved is needed, but otherwise, being separated by default will afford flexibility and ease of partial window selecting for stretching layout.

The Y-VARIABLE would adjust for the standard of using guard bars and the use of poly to M1 contacts to connect the P and N gates across the guard bars by M1.

This arrangement will make it possible to steal a P or N channel when performing metal only changes or, by separating the P and N and connecting the source together instead of to power and ground, make creating a XFR gate possible.

I am not saying separate normal series gates, like NANDS and NORS, or sensitive capacitance nodes like when two XFR gates should be combined. However, the addition of buffers after a latch stage or in a delay chain can add enough ‘inner space’ in the layout as long as there is enough space in the floorplan.

Even the XFR gate leading to a latch is subject to separation. You may ask, “How much capacitance does it really add by separating the XFR gate from the latch, or the output of the feedback inverter? Isn’t it possible a little extra capacitance on a ‘storage’ node is really not a deficit?

You also may ask, why not separate the latch inverters? That’s possible, though, in my experience, I chose not to do that for my default, but would have an alternate version ready in case. I chose, as my default, not to separate the XFR gate too far from the driving inverter of the latch, while simplifying the interconnect keeping the M1 runs close between the P and N devices.

Moving the feedback inverter out of the critical path that could be an option when planning bit slice-type layout OR, when it’s needed to compact the layout in the X-direction. Since the sticky feedback gate is not a critical path, it is not unusual to place small slow gate outside of the critical path and bring the XFR gate next to the latch driver.

Conveniently, separating the OD between the XFR gate and the latch also provides space for the connections to the XFR gate inputs, which will require alternating interconnect to a M2 bus running between the P and N. (Note to yourself: A CLK and CLKB pair always alternate when supplying inputs to a series of XFR within the same datapath).

Another area that may be suitable to separate are large and asymmetrical NANDS or NORS. These can be very large devices or where the high speed input is reduced to reduce the gate loading, in either case, a standard series configuration where both gates are the same size is not the case. An increase in source drain capacitance will ensue, so this should be taken into consideration (talk to the design engineer) since the gate may be sensitive to the added intermediate parasitics depending on the application.

Very large buffers can be another area where a gap can be inserted by separating a normally merged device. This gaps can also provide a space to run a M1 connection to the guard bar if the tap frequency needs help.

In general, gates can be poly connected in pairs with double contacts, but not merged all together in one giant poly comb. Keeping poly gate to pairs only is to further build flexibility for mask changes into the layout. In other words, having ALL the poly connected together for a large device may lower the resistance but may also be ridiculously redundant and more restrictive when the circuit designers are searching for ways to make a metal mask change and cutting up a big buffer without changing poly can help keep costs down.

***

Vacation for the eyes

What kind of vacations are good for your eyes? One that you cannot take any monitor with you.

I was a trainee on a sailing ship for two weeks in 1996. Although I got seasick for a couple days straight, my eyes felt great.

Another idea is go to a dude ranch and ride horses chasing cows and mend fences all day. I managed to do that in 1992, that was really fun and a was another monitor-less experience. What about you, what was one of your classic eye soothing vacations?

***

Feel free to Like, Share, Follow, and Comment (please, no AI driven scripts)

CKY

***

Copyright © 2023 Challen Yee / ICLayout.blog  Some Rights Reserved

Disclaimers