Nyuk Nyuk Nyuk!
Of course it is an ordinary resistor. You can use any of several types,
whatever is available at your local electronics store. The value isn't
that critical. Like 5% is OK, 1/8W or 1/10W or even 1/20th of a watt
are safe power ratings. The only detail, is the form of the resistor,
either surface mount or through hole (axial leaded type). The resistors
on the motherboard are surface mount, and for this application SMT would
be the best if you can find some.
http://www.digikey.com (Enter P10.0KHCT-ND in the search box
The resistors are 10 for $0.90)
http://dkc3.digikey.com/PDF/T042/0914.pdf (link valid until catalog update)
The way this resistor is being applied, means it is named "programming
resistor" or a "strapping resistor". The idea is, a pin on an IC can be
used for two purposes, where one purpose exists at "T=0 seconds" and
the other purpose exists after that.
Let us draw a circuit...
+3.3V
+3.3V ---X |
10K |
X----/\/\/\---+ \ (Internal
"strap" | / 240K
GND ---X | \ Pullup
------------- | / Resistor)
| | |
| | | 22 ohm
PCICLK0 |----+--+--/\/\/\---------> Clock to PCI slot
| Series
| Damping
------------- Resistor ("R99")
When the circuit is running in its normal mode, PCICLK0 will be
enabled as an output. A square wave at 33MHz would be coming out
of that pin. The purpose of the 22 ohm resistor, is to match the
load at the end of the line, which in this case would be a PCI
plugin card. The sum of the output impedance of the pad, plus the
22 ohm resistor, is attempting to match the transmission line
impedance of the clock track. So, this is all that is needed of
the circuit, to do the normal stuff...
-------------
|
| 22 ohm
PCICLK0 |----------/\/\/\---------> Clock to PCI slot
| Series
| Damping
------------- Resistor ("R99")
Now, the other condition to worry about, is what happens at powerup,
i.e. when time is at zero seconds in the life of the IC. The PCICLK0
pin is not enabled to drive out at that time, and is instead an input.
It can sense a logic "1" as a 3.3V level, or a logic "0" as a 0V
level. The chip actually has a resistor of value 240K ohms inside
the chip, and the end of the resistor goes to +3.3V. This places a
weak 3.3V level on the pin, and that is the reason that no FS3 circuit
is required. So, with no modification to the circuit, FS3 always sees
a logic "1" caused by the weak (high resistance) internal resistor.
The high resistance is important. When the PCICLK0 is driving its square
wave, which is most of the time, the driver is so strong, it can
"override" the value the 240K pullup resistor was making at time t=0.
And, that is the trick of using "strapping" resistors, is the values
of the resistors are selected to "overcome" one another.
Our friend "P2B" has selected a 10K ohm resistor for a reason. It
is a stronger resistor than the 240K ohm resistor inside the chip.
When the 10K ohm "strap" is added to the circuit, and the other end
of the resistor goes to either +3.3V or GND, the 10K ohm resistor can
override what the 240K ohm resistor is doing. So, when a jumper plug
is used to connect the left hand end of the 10K ohm resistor to GND,
for example, the 10K resistor is saying to the PCICLK0 pin, "I want
a logic zero please". Since the 10K resistor is strong compared to
the 240K resistor, the 10K resistor "wins". There is actually a voltage
divider action, but I'm trying not to get too technical here.
But, of course, the PCICLK0 driver is much stronger than either the
10K ohm "strapping" resistor, or the 240K pullup resistor inside the
chip, so when the chip is finished initializing, the PCICLK0 will drive
out its square wave clock signal, and the other resistors just
"go along for the ride".
A detail that wasn't pointed out on the web site, is that another
function of the 10K ohm resistor, is it is separating the high
frequency part of the circuit (the "square wave"), from those nasty
long wires used to connect to the three pin header which is hot melt
glued to the PCB. It is important that the 10K ohm resistor be fastened
as close as possible to the 22 ohm resistor. The best resistor for
that purpose, would be a surface mount (SMT) resistor, of size
0603 or so.
This wire can be long.
|
|
v 10K
X----/\/\/\---+
"strap" |<------ Keep this wire physically short,
| as it can upset the high frequency
------------- | signal.
| |
| | 22 ohm
PCICLK0 |----+-----/\/\/\-----
Surface mount resistors are not a lot of fun to work with. They
tend to stick by surface tension, to the tip of the soldering iron.
In this case, an SMT resistor could be reflow soldered to the
existing junction where "R99" is soldered to the motherboard.
Most hobbyists will have access to the older through hole resistors,
the axial ones with the wires sticking out of the end. These should
not be soldered directly to the board! The problem is, the wire on the
resistor is stiff, and after soldering is finished, if you accidently
put any stresses on the resistor (tugging), you can actually rip a
copper pad off the motherboard. I recommend using some thin wire,
such as 30 gauge wire wrap wire (kynar insulation is easier to strip
than teflon). The purpose of the thin wire, is it bends easily and
that wire will "give" first, rather than the copper pad.
The max stub I would have off that circuit, is perhaps 1/2". That
is the length of wire plus the lead on the resistor, as shown
in the above figure with the "Keep this wire physically short"
arrow next to it. That is why soldering an SMT resistor directly
to R99, keeps the equivalent wire length to well under 0.1".
If you feel these conditions are too difficult to meet, the
only downside of using longer wires at the 22 ohm resistor, is
that whatever PCI slot or PCI device is driven by PCICLK0, may
get misclocked. Simply not using the PCI slot it connects to is
one solution. However, if the PCICLK0 signal is used to clock
the Southbridge, then you must maintain the high frequency
quality of the signal, if you expect the board to boot.
The datasheet for the clock gen goes over some of these details.
http://www.icst.com/products/pdf/ics9250-08.pdf
So, now you know the "secret" of strapping ICs. Many ICs use
this same principle, of using output pins as inputs at time
t=0, and latching or sampling the value "strapped" to the pin
for that short time. As long as the "strapping resistor" is weak,
the normal output circuit never knows it is there.
HTH,
Paul
