At 02:43 PM 4/12/00 +1000, you wrote:
>Dan Michaels wrote:
>
>> assuming the circuitry at the node may be represented by a simple
>> Thevenin equivalent, would not the node more or less act as a low-pass
>> filter, following the BW calculations previously given?
>
>>      +-------Rth------+
>>      |                |
>>     Vth             Cstray
>>      |                |
>>     gnd              gnd
>
>  Rth in this circuit is the series combination of the circuit trace
>resistance (tens of milliohms) and the driver internal resistance.  A
>PIC which can drive 50 mA into a short circuit has a resistance of 100
>ohms, so Rth is in total, 100 ohms.  Dedicated bus drivers (without blue
>caps!) can be lower again.
>
>  So, your time constant is quite short.  If you put 10 kilohm in
>parallel with Cstray, it won't affect much at all.  Certainly, 10 kilohm
>in series will slow things right down, but I don't see where this was
>proposed.
>--
>  Cheers,
>        Paul B.
>

Yes, exactly what I said last time, when looking at a driver output
node. For this case, an extra 10K in parallel simply folds into Rth,
for no significant effect on bandwidth or much else.

However, to repeat again, the example I gave previously for a
"general" node, where let's say Rth is on the order of 10K [rather
than 100 ohms], and the stray cap is 10 pF, and cross-coupling is
not a serious factor, then F(3dB) ~ 1.6 Mhz. **This** is the case
where stray capacitance will have some effect. **This** is the
point I have been trying to make for several days. Not **every**
node is a low-impedance node.

In some cases, I have actually used SIP resistors for purposes
other than terminating a driven buss. In some of my analog ckts,
where R values are in the range 5K - 100K, stray capacitance is
clearly a factor. Dropping selected R values, at nodes where no
physical caps are present, can markedly improve bandwidth.

best regards,
- Dan Michaels
Oricom Technologies
===================