RE: [SI-LIST] : Output driver versus internal logic switching noise

Peterson, James F (jfpeterson@space.honeywell.com)
Thu, 28 Aug 1997 12:40:02 -0400

If Larry is right about internal logic generating a large portion of the
chips current compared to the simultaneous switching of out buffers,
then the old way of doing ground bounce analysis based on simultaneous
switching activity is a WAG at best. (An SI wizard once told me you also
need to account for the current conducted through the clamping diodes
when input signals undershoot and overshoot...)

>----------
>From: Larry.Smith@Eng.Sun.COM[SMTP:Larry.Smith@Eng.Sun.COM]
>Sent: Thursday, August 28, 1997 11:46 AM
>To: si-list@silab.Eng.Sun.COM
>Subject: Re: [SI-LIST] : Output driver versus internal logic switching noise
>
>+
>> I'have been reading in several switching noise related papers that
>> traditionally the only important source of noise in ICs was due to the
>> simultaneous switching of the output drivers but that in today circuits the
>> switching noise generated by the internal logic would be also very
>> important. Does anybody have quantitative confirmation of this? It is any
>> study of the ratio between internal logic/output driver switching noise in
>> any commercial IC or ASIC?
>>
>> Thanks in advance
>> Jose Luis Gonzalez
>
>It is true. Just take a look at the power dissipated by internal logic
>circuits and IO drivers and ask yourself which one has the capability of
>making the most noise. IO circuitry may dissipate a few watts but
>internal circuitry on some modern processors and ASICs can dissipate 10's
>of watts. It is now common for the internal circuitry to run at a lower
>power supply voltage (2.5 or 2.0 volts) compared to the IO circuitry
>(3.3 volts). Not only is there more energy to make noise in the
>internal circuitry, but it is likely to be more sensitive because 100 mV
>of noise is a bigger proportion of a 2.0 volt power supply voltage.
>
>Noise transients in the internal circuitry tend to be code dependent.
>For example, a processor can be actively doing something for a few clock
>cycles and then stop doing it for a few clock cycles. This creates a
>clock sub-harmonic. Various forms of energy conservation also create
>noisy power surges in internal circuitry when circuits are activated or
>deactivated. My personal opinion is that power distribution and noise
>management will be the next big challenge for the package/semiconductor
>industry. It is likely to surpass simultaneous switch as the major
>problem to be solved.
>
>There is a special session in the upcoming Electrical Performance of
>Electrical Packages conference (San Jose, October 27-29, 1997) to deal
>with this topic.
>
>regards,
>Larry D Smith
>Sun Microsystems
>