From: Adrian Shiner ([email protected])
Date: Tue Nov 09 1999 - 12:06:03 PST
Fud ( rock on Larsen) for thought...Silly question maybe, who is designing
the circuit? Don't the circuit designer & the PCB layout designer work
together? Carry on working separately & we will be suffering from hardware
designed like some software operating systems. Has anyone ever seen the
hardware timing requirements for some of these operating systems. I know in
the real time industrial computing world that these things matter. Perhaps
the software game is still in the wooden waggon wheel era.
Adrian
----- Original Message -----
From: D. C. Sessions <[email protected]>
To: <[email protected]>
Sent: 09 November 1999 19:14
Subject: Re: [SI-LIST] : More Micro Noise;-)
> [email protected] wrote:
> >
> > I'm back...
> >
> > We have decided to take a look at the spread spectrum concept for one of
the
> > boards, however, a problem still exists for the other...
> >
> > This board has one micro feeding 3 DSP's, in the DSP's there are loads
of
> > registers that change at the same time. So I have very strong spurs...
> >
> > Since I can't tolerate clock jitter, but I could tolerate a skewed
clock,
> > what if I build in say a 1/8th of a clock cycle delay between each big
> > current hog. i.e. The micro transitions on the rising edge of the clock,
this
> > rising edge is delayed slightly ( 1/8th ), then it triggers the first
DSP
> > then delayed again before triggering the 2nd DSP and so on until after
the
> > 1/3rd it's time for the micro clock to go high.... and so on.
> >
> > While it most likely won't affect the radiation from the chips, it
should
> > reduce the dips on the power planes... Am I way off base here?
>
> It's a good idea. I'd change it in one regard, though, by not using the
> same phase interval between all of the parts. That way you can also
reduce
> the inphase harmonics as well. (Obviously a bit of math is in order
> depending on the number of devices you're using.)
>
> For static logic the #1 power transient comes from the clock buffers in
> the flip-flops themselves, and although there's usually some difference
> in the magnitude between the rising and falling clock edges it's not
> large thanks to the several stages of inversion. Therefore, you don't
> get much benefit from using opposite clock edges. Quadrature would be
> OK, aside from piling up energy in the even harmonics (CMOS has some odd
> -- or perhaps I should say even -- harmonic content.)
>
> I'd have a look at the spectral distribution of your emissions and plan
> the phasing to avoid the peaks you already have.
>
> --
> D. C. Sessions
> [email protected]
>
> **** To unsubscribe from si-list: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE
si-list, for more help, put HELP. si-list archives are accessible at
http://www.qsl.net/wb6tpu/si-list ****
>
**** To unsubscribe from si-list: send e-mail to [email protected]. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****
This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:44 PST