Richard said:
I recently install a ASUS motherboard in my old case and see a message
"Power Fan below threshold" from the ASUS utility software. There is a
CPU fun and a CHA fun for the case. I can not find the POWER fun for
the POWER supply.
Does that mean the Power Supply Fun stop working? Do I need to buy a
new Power Supply 300W?
Thank you in advance
There are several options for power supplies.
A cheap power supply might have no monitoring capability at
all. If there is no monitoring cable on the PSU, then the
PSU cannot talk to the motherboard.
If the PSU has a two wire cable, with a three pin connector
on the end, that is a fan monitor cable. The two wires are
"tachometer_output" and GND. When this PSU cable is connected
to a fan header on the motherboard, the motherboard should
be able to measure the RPMs. It does this by measuring the
period between pulses.
If the PSU has a two wire cable, with a two pin connector
on the end, this could be a thermistor output. A negative
tempco thermistor, nominal resistance 10K ohms at 25C, with
a beta of 3435, would be the typical component used. Most
monitoring chips and associated software would expect this
kind of component, and the temperature is computed with
the assumption that the thermistor is one of that type.
The thermistor inside the PSU allows the motherboard to
measure the temperature of the PSU, and with the right
software, you might be able to raise an alarm if the PSU is
too hot. Recent motherboards don't seem to have the
necessary two pin header - my P4B has a TRPWR header, but
I don't think my other boards have one of those.
Since the fan speed is typically measured by measuring
the period between pulses from the fan, the measurement
register has a limit to the amount of counts it can hold,
before it overflows. The slower the fan goes, the more
counts that accumulate in the counter. The operation
looks something like this:
| <--- Motor pulses twice per one fan revolution ---> |
_____________ _____________ ___
___| |______________| |____________|
_ _ _ _ _ _ _
_| |_| |_| |_| |_| |_| |_| |_
| <- seven counts in total-> |
counter = 0x07 hex
The monitor chip makes those tiny measurement pulses, and the
frequency of the pulses has to be selected to give good
resolution and good min to max range. The pulses are created
by a chain of divide-by-two flops, and the BIOS sets up the
right value for the divide chain before Windows boots. I think
both the BIOS and the OS/applications have access to the
monitor chip, and it could be that the OS/applications are
not allowed to change the divider chain (there is no standard
for the interface that I've heard of, so this is just a guess).
Now, imagine a thought experiment. The BIOS designer at Asus
expects the fan to spin at 3000 RPM. They set the divider so
that the 8 bit counter (max value 255 decimal) will catch
128 pulses in the time it takes the fan to generate one full
pulse. Now, the customer gets the motherboard, and plugs in
a PSU with a fan that spins at 1500 RPM. When spinning at 1500
RPM, the pulses are stretched out by a factor of two. Instead
of 128 pulses accumulating in the 8 bit counter, a total of
256 accumulate in the counter. That is one more pulse than
the counter can hold, so an overflow bit is set (or equivalent
thereof).
When the software detects an overflow, it doesn't know if the
total counts are 256 or a gazillion (when the fan stops). The
fan speed is somewhere between 0 and 1500RPM, but due to the
measurement method, there is no way to know. In a situation
like this, the BIOS reports "0 RPM", even though you can see
your fan is still spinning.
So, in some cases, there may still be room in the divider
chain, to adjust the frequency of the measurement pulses, and
bring the pulse period measurement back within the 1..255
count range limit of the counter. But whether the software
is allowed to do that or not, is something I cannot answer.
You may ask the question, "why do they measure the period
of the pulses, when they could count the revolutions ?".
The answer is, the period measurement method allows the
speed of the fan to be computed faster. The measurement
time is bounded by the time it takes to overflow the
counter. The question one has to ask, is why the silicon
chip designers could not use a larger register to accumulate
counts, like maybe a 16 bit register. With sufficient register
bits, there wouldn't even be a need to adjust the divider
chain.
The RPM value you see, is not the value extracted raw from
the chip. The readings are averaged, to reduce the jitter
in the values seen. Averaging the values increases the
latency in seeing a change in fan speed, but it avoids
the RPMs being annoyingly unreadable by a human.
For monitor chips that sit on the SMBUS, rather than an
LPC bus, it is possible for two software programs to
interfere with the readout of registers from the monitor
chip. When this happens, a program like MBM5, Speedfan,
or Asus Probe, might report the fan at one moment is 0 RPM,
and at the next instant is 10000 RPM. The SMBUS has no
semaphore to prevent simultaneous accesses by two programs,
and the serial data on the SMBUS is cut off in mid access.
The result is garbage values seen by the software.
Much of the above is supposition and assumptions on my
part, as the datasheets do not describe in clear English
how this stuff works. I hope (most of) the above is correct.
At least it may help explain why programs readout "0 RPM"
when the fan is still spinning.
HTH,
Paul
