Prev: Quad Core Processors
Next: ATX PSU
From: kony on 2 Apr 2008 03:18
On Tue, 1 Apr 2008 12:33:47 +0200, "DanielEKFA"
<danielekfa(a)yahoo.com> wrote:

>Hi there :)
>
>I was hoping someone could help me clarify if this idea I got is a good one
>or not. Yesterday, I just got a Thermalright Ultra 120 CPU cooler to which I
>can attach two 12-inch fans. I'd like the motherboard (an ASUS P5K-E
>WiFi/AP) to be able to throttle these two fans in the same way it did the
>stock cooler's fan, so that when the CPU is cool, I get as little noise as
>possible, while when it runs hot, it'll throttle up the fans to more
>effectively cool down the CPU.

As you reported in your subsequent post, the processor runs
really cool... ok, but is this at full load? The
distinction is, full load is the more important parameter
and unless you are overclocking a lot, you should not need
two 120mm fans on that heatsink, would be making this more
complex than necessary, perhaps more complex than beneficial
since a CPU doesn't need to run as cool as possible, only
cool enough to stay stable and promote good lifespan (which
is many years longer than you'd be likely to use it).


>
>Thing is, it's not like there's two connectors for the CPU cooler's fans,
>there's just the one. Now, I was thinking that it might be possible to
>simply connect both fans to the same connector. But then I thought, "how
>would that affect the fan rotation readings," which, AFAICT, are read
>through the yellow wire, while the black and red are ground and +12V
>respectively.

The sample rate the board uses to detect RPM would determine
how much of these overlapping RPM signals would be detected.
It would tend to register higher than the RPM of one fan,
but possibly not a summary total of both fans' RPM. It
would be an undesirable result either way so one of the
fan's RPM leads would be isolated, not plugged in.


>
>Then I start thinking about how this throttling is done in the first place,
>and I'm *guessing* that the yellow wire reports back rotational speed in
>some way, while some mechanism lowers the voltage (or something equivalent
>to that) on the power wires, effectively controlling the fan. So, I'm
>thinking I could just share the ground and +12V wires between two fans,
>while using just one's yellow wire to report back speed.

Yes, but what you lose is the fine level of control
possible. There is a minimum and maximum RPM the fan
controller circuit could cause with a given fan, then when
having a second one the minimum is still possible, but is a
higher duty cycle of the entire possible control range per
the same fan(s) RPM, and the maximum power the circuit can
deliver is still the same so it's roughly divided between
the two fans - you can't have both fans spinning at full RPM
that (only) one would have if the system were to get hot
enough to cause the circuit to try to output full power.

This could effectively cancel most of the gain in having two
fans, but to back up a moment we would also want to consider
if the two fan combo, or one fan alone, would ever run at
full RPM - if the CPU ever got hot enough to cause that
which it may not, especially not if you aren't overclocking
and it's a typical room temp, decent case cooling.


>
>Okay, so my questions are the following:
>A) Would this work? (I realize that I'd only be able to detect the possible
>death of the one fan which is connected to the yellow wire, but as they're
>both cooling the same object, I'm willing to accept this)

Yes you will retain a certain level of control, just a lower
max RPM, and the steps of increase (a granular increase
typically, not truely linear compared to infinitely small
steps in CPU temp increase) would be broader. In practice
it may not matter so much because a CPU tends to change temp
rapidly going from mostly idle to mostly loaded, seldom
would it stay around 30% load for example, unless you had a
very fixed scenario like encoding a video in realtime as it
came in over a streaming (cable TV or whatever) connection.


>B) Would the motherboard be able to support this, power consumption wise,
>i.e. would I be in risk of blowing out a capacitor or worse? How many watts
>would be safe to draw from these onboard connectors?

The fan control circuit is typically consisting of an opamp
driving a transistor. It would be possible to overheat the
transistor if it weren't 'sunk to a large enough area of the
board copper or if case temp was very high, but it's largely
a matter of just how much current you're using, how power
hungry these fans are at max output. Generally, a fan trace
going to some header elsewhere on the motherboard is good
for about 500mA max, but for the CPU it can depend on the
board design, it might be only 500mA total or could be quite
higher in capacity.

Under no situation would you blow out a capacitor with this
config, it'd be the transistor or a copper trace
delaminating, oxidizing or burning through over time. If
your fans are each rated for about 350mA or less it does not
seem likely to be possible to cause a problem, fans tend to
be rated for peak current while above I'm speaking of
continuous current. Since the fan control circuit is, by
design, limiting current, in a normal throttled operating
mode it would likely be safe to use fans rated for even
upwards of 1A peak power but to be conservative there is no
need to test or risk this, certainly no need to have such a
fan let alone a pair of them for processor cooling.

Personally I would do as suggested above, remove one fan and
see how hot the system gets... keeping in mind that you may
have bios or software control over what the fan RPM vs temp
threshold is. I usually like to keep my 'sink fan running
at minimum RPM until CPU has exceeded 50C, sometimes higher
depending on the situation. If you only have one fan header
that has variable control, it could even be as desirable (if
you really want it to control two fans) to control one fan
on the heatsink and one fan on the case exhaust... otherwise
having the case exhaust fan at a fixed speed it isn't
responding to removal of the heat from the case in a
corresponding matter to heat generation.

From: kony on 2 Apr 2008 03:23
On Tue, 01 Apr 2008 13:53:44 GMT, Angelo Campanella
<a.campanella(a)att.net> wrote:


> For new fans drawing more current, a separate current feeding supply
>should be installed. If the applied voltage is switched (either 12v or
>zero in rapid succession) then your new control circuit is easy and
>cool. If it is analog partial voltage, then heat dissipation within your
> new control circuit will need to be considered (heat sink, etc). It is
>possible that where current switching (full 12 volts, part time) is
>used, it is at an ultrosound rate, then smoothed by a capacitor before
>being fed to the fan motor. Using this smoothed external signal for your
>control circuit results in an analog circuit requiring a heat sink. It
>is also possiblte that the switching is at a low frequency, and that you
>can use that as a new switched control signal. Put a scope on the 12v
>terminal and determine which it is.

Generally the configuration is something similar to a
programmable control chip as input to an LM386 opamp biasing
a transistor, filtered output with a few hundred uF
capacitor. The transistor is 'sunk to board copper, there
is not a very effective way to further heatsink it -
epoxying something on the top might mitigate epoxy casing
breakdown but it's not likely to be nearly the issue as that
high a current level as heating of the supply traces.


>
> Another possibility is to simply run the new fans at one reduced
>voltage (e.g. 5 volts), for low RPM silent operation. At least start out
>that way. Another possibllty is to run the two new fans in series,
>giving 6 volts each for the same result.

They can't be run in series, assuming typical DC brushless
design.
From: kony on 2 Apr 2008 03:26
On Tue, 1 Apr 2008 15:59:22 +0100, "GT"
<ContactGT_remove_(a)hotmail.com> wrote:


>The yellow (RPM) wire is not used to control the fan throttling system. The
>fan speed is controlled by CPU temperature and threshold values in the BIOS.
>There is a warning on fan rotation speed which can be set to 'alarm' or
>'shutdown system' or 'disabled'. So getting a speed reading from just 1 fan
>would be fine and shouldn't cause confusion
>


Normally this is true, but it wouldn't be impossible (and to
some, even a more desirable way to implement the fan
control) if the RPM signal were used to determine and change
fan RPM to a threshold level. For example, at 50C, do what
is needed to make fan RPM 3000. It would be a tradeoff
since to some the goal is a linear increase of fan speed to
temp in an attempt to keep the temp constant, BUT, there is
no real need to keep the temp constant and with the temp vs
speed control the granularity and thresholds tend to make it
less compatible with certain fans, mainly those that use
higher current.
From: Paul on 2 Apr 2008 04:32
DanielEKFA wrote:
> Hi there :)
>
> I was hoping someone could help me clarify if this idea I got is a good
> one or not. Yesterday, I just got a Thermalright Ultra 120 CPU cooler to
> which I can attach two 12-inch fans. I'd like the motherboard (an ASUS
> P5K-E WiFi/AP) to be able to throttle these two fans in the same way it
> did the stock cooler's fan, so that when the CPU is cool, I get as
> little noise as possible, while when it runs hot, it'll throttle up the
> fans to more effectively cool down the CPU.
>
> Thing is, it's not like there's two connectors for the CPU cooler's
> fans, there's just the one. Now, I was thinking that it might be
> possible to simply connect both fans to the same connector. But then I
> thought, "how would that affect the fan rotation readings," which,
> AFAICT, are read through the yellow wire, while the black and red are
> ground and +12V respectively.
>
> Then I start thinking about how this throttling is done in the first
> place, and I'm *guessing* that the yellow wire reports back rotational
> speed in some way, while some mechanism lowers the voltage (or something
> equivalent to that) on the power wires, effectively controlling the fan.
> So, I'm thinking I could just share the ground and +12V wires between
> two fans, while using just one's yellow wire to report back speed.
>
> Okay, so my questions are the following:
> A) Would this work? (I realize that I'd only be able to detect the
> possible death of the one fan which is connected to the yellow wire, but
> as they're both cooling the same object, I'm willing to accept this)
> B) Would the motherboard be able to support this, power consumption
> wise, i.e. would I be in risk of blowing out a capacitor or worse? How
> many watts would be safe to draw from these onboard connectors?
>
> Thanks in advance for any feedback :)
>
> Daniel

There are four fan headers on that motherboard. You can run the RPM wire from
one fan, to one of those unused chassis fan headers. The RPM wire from the
second fan, can continue to be connected to the four pin CPU fan header. (I.e.
You'd use two fan headers, to monitor two RPM signals.)

Q-fan, as far as I know, is not a closed loop system - the chip controlling
the speed of the fan, does not check to see if the fan is following commands
or not. There will be a check for "zero RPM", which is a threshold value, but
other than that, Q-fan delivers a voltage in response to the measured
temperature. It doesn't care what RPM the fan responds with, when the
voltage is applied.

The motherboard has no selection function in the BIOS, for three pin versus
four pin control method. The motherboard doesn't appear to support the old
12V control method. It looks like a PWM-only CPU-fan header. That means,
a retail Intel fan, having the four pin interface, listens to the PWM
signal, and the speed is controlled that way.

The fans you've placed on your cooler, could be three pin fans, which means
right now, they probably don't demonstrate any control at all. If you connect
one of them as a test, it would probably running at full speed no matter
how Q-fan is configured for the CPU-Fan.

The following, is a PWM fan 120mmx120mm and 25mm thick. At full bore, it is
96CFM, and that would probably be pretty loud. Since it is PWM, connecting
it will be easy. The PWM output of the motherboard, can be connected to
more than one fan in parallel. So no circuitry needs to be designed, to
do this. It is just wiring.

Rexflo DF1212025BH-PWMG 120mm Case Fan - 96 CFM 42.60 dBA 12V at 0.3A - 4 pin w. PWM. $18
http://www.newegg.com/Product/Product.aspx?Item=N82E16835705025
http://rexususa.com/fans/21225-4.html


CPU-FAN PWM -----------------------------+------------------------+
| |
PWM-Fan1 PWM-Fan2

RPM -----------------------------+
|
RPM-Fan1

+12V -----------------------------+------------------------+
| |
+12V-Fan1 +12V-Fan2

GND -----------------------------+------------------------+
| |
GND-Fan1 GND-Fan2

Spare-Chassis-Fan-Header

RPM ------------------------------------------------------+
|
RPM-Fan2

For the +12V and GND, those don't even have to come from the CPU-FAN header,
if you don't want. They can also come from a Molex disk drive cable, as it
has +12V and GND. That is an alternative, if you think the fans will draw
too much current for the motherboard rating. Since the above Rexflo 120mm
fan is 12V @ 0.3A, two of those at 0.6A is probably OK. The motherboard
manual claims up to 2 amps can come from a fan header, which seems a
bit aggressive on their part. So the 0.6A total is probably not too
much, if you want to run both fans using the +12V and GND pins of the
CPU fan header.

So nothing needs to be designed, if you can find a PWM equipped fan like
the Rexflo above. Just a matter of doing some wiring. It helps if you
have an electronics store, that happens to carry fan connectors. I got
a bag of pins and the plastic 3-hole bodies, so I can make my own three pin
ends when I need them. You'd need a three pin end, to make the single wire and
connector needed for the "Spare-Chassis-Fan-Header" connection.

Otherwise, the wiring of the other two fans, can be done with the
harnesses already on them. You could even chop off the four pin on
one Rexflo fan, and use it as a mechanism to make the connection to the
"Spare-Chassis-Fan-Header", but it would look a bit ugly having
a four pin connector on a three pin header. You might forget how it
should be oriented, and it might fall off some day.

HTH,
Paul
From: GT on 2 Apr 2008 06:01
"kony" <spam(a)spam.com> wrote in message
news:n3d6v3pgqm6u71kll5t6otsom7sg1qmqlu(a)4ax.com...
> On Tue, 1 Apr 2008 15:59:22 +0100, "GT"
> <ContactGT_remove_(a)hotmail.com> wrote:
>
>
>>The yellow (RPM) wire is not used to control the fan throttling system.
>>The
>>fan speed is controlled by CPU temperature and threshold values in the
>>BIOS.
>>There is a warning on fan rotation speed which can be set to 'alarm' or
>>'shutdown system' or 'disabled'. So getting a speed reading from just 1
>>fan
>>would be fine and shouldn't cause confusion
>>
>
> Normally this is true, but it wouldn't be impossible (and to
> some, even a more desirable way to implement the fan
> control) if the RPM signal were used to determine and change
> fan RPM to a threshold level. For example, at 50C, do what
> is needed to make fan RPM 3000.

Absolutely - not impossible, but that would require the BIOS to have
knowledge of the thermal dissipation power of the heatsink and the CFM of
the fan, plus one or two other values concerning case temperature and
airflow through the heatsink. Not as simple as you think.


First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: Quad Core Processors
Next: ATX PSU