Metal antenna ratio is not cumulative, but why?

[Willt]

Hi all,

Most of the material discussing antenna rule says that
metal antenna ratio (AR) is not cumulative (e.g. just count M2 alone, not M2 and M1 together).
But why?

Is this because the gate leakage discharges the metal which is already fabricated?
Maybe for more advanced technology, cumulative antenna ratio (CAR) is used for precise checking.

Any comment is much appreciated.
Thank you very much

Will

 

[rangermad]

Speculating a bit but I think the plasma etch only effects the current metal/via level of processing. It is cumulative at the current level. So it you have two M2 segments shorted with M1 then those segments can add up to violate the metal to gate ratio. Gate leakage is not enough to discharge the metal. P+ or N+ diodes are the most effective way to solve the problem.

 

[erikl]

Just an addition to rangermad's good explanation: AFAIK metal always has to be cumulative, but the rules allow to limit the area ratio calculation to a max. metal layer. Why? Because such a gate input - even if wired at higher metal levels to other inputs - perhaps at the next higher than the last considered metal level is controlled by an output, which in any case establishes a junction connection to the substrate, sufficient to discharge the collected ion charge.

I.e. metal layers must be (cumulatively) considered only up to one layer below the output layer feeding it.

 

[kapilsn]

Hi Erikl
I thought the cumulative ratio was not used since the charges accumulated on the metal are are flushed after every etch .

 

[erikl]

Re: Metal antenna ratio is cumulative, but why?

... since the charges accumulated on the metal are are flushed after every etch.

Absolutely correct, kapilsn, but ... at the next etch, the currently etched metal layer wire is already connected (by vias) to all the lower metal wires and (by contact) to the gate. So the - during this next etch - collected charges will be distributed to the currently free-etched wire and all lower metal wires down to the gate. That's why the antenna ratio calculation must always be cumulative from each layer down to the gate.

 

[kapilsn]

Re: Metal antenna ratio is cumulative, but why?

Hi Erikl,
I still had a doubt that the amount of charges getting developed will be only proportional to the area of the current metal getting etched . The metal which is already etched will not result in any extra charges . Why would the area of the already etched metal be important .

Kapilsn

 

[erikl]

Re: Metal antenna ratio is cumulative, but why?

... the amount of charges getting developed will be only proportional to the area of the current metal getting etched. The metal which is already etched will not result in any extra charges.

It is connected! Charges will be distributed all over them.

Why would the area of the already etched metal be important.

Because it is connected. Q=CV . Do you know connected wires (without current flow) which have different voltages at different locations? I.e. charges are distributed corresponding to the local V=Q/C=const. value.

 

[dick_freebird]

Re: Metal antenna ratio is cumulative, but why?

The "input current" is the plasma charge-throw as it
etches the present layer. Dry etch plasma can't get
at the already-etched layers as they are hard-masked
by the ILD between present layer and lower ones. A
lower attached layer may "share" charge but this does
not mean it -adds- charge or current. All of that is
received at the layer under etch.

The "antenna" is the periphery of the feature, just as
it it being cut free. Before this there is a shorting sheet
with many attached vias for plasma charge to exit, and
the not-etched feature is soft masked by photoresist.

Once the layer's features separate, plasma charge on a
line has only diodes (benign) or MOS gates ("programmable")
to get out through. You don't want randomly programmed
VT and you want latent oxide defectivity or outright gate
rupture even less.

So antenna rules, if done right, would check net
connectivity for any attached diodes, and only flag
nets that have only MOS gates connected.

Thing is, CAD dudes (and the very occasional dudette)
are lazy bastages and won't do the work. So you get
conservative rules that assume no help from diodes,
in some technology cases I have seen.

Capacitance doesn't really enter it. Etch times are long.
Interconnect capacitances are small.

You will charge up until either diode photoconduction
or MOS gate tunneling takes a balancing current.
So if you look carefully you ought to see antenna rules
working against a ratio of feature periphery to attached
MOS gate area (it's current density through the gate
that leaves trapped charge and oxide damage
proportionally).