help: monitor power supply lines

[Robin Laing]

Tim wrote:
> On Tue, 2008-04-22 at 10:38 -0600, Robin Laing wrote:
>> First, don't fully trust the voltages provided by the motherboard.  If 
>> you need accurate voltages, you have to adjust (Calibrate) the voltages 
>> to an accurate reading.  I ran through this with two power supplies that 
>> the MB said were good but had 5V rails that were .5 volts low.  One DVD 
>> burner later I found out.
>>
>> Install lm-sensors and either ksensors or gnome-applet-sensors.  You can 
>> adjust the accuracy of the sensors by playing with the configuration files.
>>
>> http://www.lm-sensors.org/
> 
> Though you need something accurate to calibrate against...
> 
> The power supply is internally regulated.  If it's set to provide 5
> Volts from its output terminals, that's what it does.  However, if
> there's a drop across the wiring and there's less than 5 Volts at the
> motherboard, the power supply will not know about it.  It doesn't sense
> at that point of the circuit.
> 
> I don't know if it's still the case, but some computer supplies only
> really regulated some of the voltages.  e.g. It'd keep the 5 Volts at 5
> Volts, but the 12 Volt supplies were less controlled.  Since some
> supplies were derived from a common source, it's presumed that adjusting
> one keeps the rest in step, but that doesn't always happen in practice.
> 
> Some motherboards may have voltage regulators on the board, but that
> could only help against too much voltage, they can't increase supply of
> what's not there.
> 

Not fully true.

The power supply is supposed to regulate the power.  That is why 
monitoring is important with a calibrated system.

If there is a significant drop on the wiring between the power supply 
and the motherboard, there is a major current draw to cause the 
resistance in the wire to drop that much voltage. V=I*R  And R is very 
small.

In my case the power supply monitor on the mother board stated that the 
voltage was at 5.1 volts, even in the BIOS.  When I checked with a volt 
meter, the voltage was closer to 4.5 volts.  This was on one of the 
unused leads from the power supply.  Using your analogy, the 
motherboard/ps wiring actually increased the voltage.

In my case, the power supply has a known fault.  The sense line circuit 
wasn't working properly and needed modification to allow adjustment to 
get the voltages correct.  I searched the web and found many people with 
the same issue and how to modify the PS to make it work.  The cost of 
shipping the PS back under warranty was more then the PS was worth.

Here is a web site explaining the mod.

http://forums.extremeoverclocking.com/t97195.html

Of course when you adjust, you measure on the board with the system running.

The accuracy of any reading is only as good as the sensors and 
conversion factors used.  If you look at the lm-sensors package, you can 
modify the conversion factor to get them accurate.

 From "/etc/sensors.conf"

VOLTAGE COMPUTATION DETAILS
# ---------------------------
# Most voltage sensors in sensor chips have a range of 0 to 4.096 Volts.
# This is generally sufficient for the 3.3 and CPU (2.5V, for example)
# supply voltages, so the sensor chip reading is the actual voltage.
#
# Other supply voltages must be scaled with an external resistor network.
# The chip driver generally reports the 'raw' value 0 - 4.09 V, and the
# userspace application must convert this raw value to an actual voltage.
# The 'compute' lines provide this facility.
#
# Unfortunately the resistor values vary among motherboard types.
# Therefore you may have to adjust the computations in this file
# to match your motherboard.

The file goes into more details on how to calculate the values and 
compensate for the differences.

As it states for the actual software, you have to adjust the values to 
provide an accurate reading in /etc/sensors.conf


-- 
Robin Laing