Question Case fans run at full tilt if there is no motherboard in the case ?

[Furi0usGe0rge]

I have a 4u case that has hot swap 120mm fans (in caddies) with standard female 4-pin pwm connectors.

When slotted into the case, the male PWM connectors that they plug into are powered by a 4-pin molex.

I don't know what kind of fan controller is built into this thing, or what else explains this mystery, but if I power these fans by shorting the 24 pin atx connector to start the psu, the fans run at full speed always.

So my first question is: why is this? Normally you can't control fan speed using just a molex power connector and the presence alone of a motherboard.

I took a multimeter to one while running with a mb in place, and got 12v, so whatever is slowing the fans it's somehow not a reduction of voltage.

(I also have some 2U supermicro cases which seem to do the same thing, though that backplane assembly uses pwm connectors that are basically just extensions, so it makes sense.)

Secondly, I bought a fan controller, anticipating this would be the cheapest way to address the problem. (I'm doing this because I'm turning cases into SAS enclosures, in case at this point you're wondering.) That's when I realized that the fan controller has a 4 pin male pwm header, while the case fans' backplane has that "magical" 4 pin female molex, and for reasons that are obvious when you think about it, an adapter for that doesn't exist.

So I can take the fans' 4 pin female pwm connectors out of their fan caddies, and thread them thru the fan backplane assembly, thereby bypassing the 4 pin molex.

Or I can make my own PWM header to molex male, and risk the certainty of eventually accidentally creating a short curcuit which potentially damages hardware or even data.

My second question is: any other ideas?

Thanks for reading.

 

[Paperdoc]

Let's start with how a 4-pin PWM fan operates. Its connector provides these signals:
Pin 1 - Ground
Pin 2 - Fixed +12 VDC power supply
Pin 3 - Fan speed signal (2 pulses per rev) sent back to host header
Pin 4 - PWM signal to fan
Inside each such fan there is a small chip that modifies flow of current from the fixed 12 VDC supply (Pin 2) through the windings to control speed. So the fact that you measured a fixed 12 VDC value on Pin #2 is NORMAL. Speed control is NOT done by changing this voltage.

What you MAY be seeing is a kind of built-in fan HUB, a device that is really designed particularly for use with 4-pin PWM fans. What a HUB does is replace one signal source. It draws power for the +12 VDC lines (pin 2 of all fan connections) directly from a PSU output - often from a 4-pin Molex connection to the PSU - and draws NO fan power from a mobo header. This avoids the normal limit of 1.0 A max current load from a mobo header, and the limit of power from Hub to fans now becomes the limit of a Molex output from the PSU, which is well over 10 A. Thus the 4U unit can use MANY fans with no limit from the mobo header. The Hub may also ensure that the Ground lead (Pin 1) is connected directly to that same Molex. The Hub will still take the PWM signal from a mobo host header Pin 4 and share it out to all Hub fans, but that is NOT a signal overload situation. Regarding the speed signal on Pin 3, a host header can only deal with that from ONE fan, so the Hub will send back to the host header the speed of one of its fans and ignore the speeds of all others.

So, the 4U module may well have a Hub circuit built into its backplane with power to it from a Molex output from the PSU, PLUS a cable to connect the backplane to one fan header of a mobo to pick up the required PWM signal. The fan caddies may be arranged to plug into backplane sockets so that they each get all the signals from the built-in Hub system. IF there is NO mobo present and hence no cable connection to the 4U backplane to provide the PWM signal, then all fans will always operate at full speed.

You have purchased a third-party Fan Controller for your system, recognizing that it will not have a standard mobo to link to. So, how to get the required PWM signal from the Controller to the backplane? Well, look for how that is done NORMALLY when you have a mobo - is there a cable to go to a mobo fan header?

BUT there is another BIG question. MANY third-party Fan Controllers do NOT provide a PWM signal system even though they do have 4-pin output headers. You see, the new 4-pin PWM fan system has several backwards compatible features. ONE of those is that, if the 4-pin fan is connected to an older 3-pin fan header that can use ONLY the older Voltage Control Mode, it can still work. In that older Mode, the header supplies a VARYING voltage on Pin 2 to control fan speed, but NO PWM signal on non-existent Pin 4. Given those signals, a 4-pin fan can NOT modify current flow using NO PWM signal, but it WILL run according to the Voltage supplied on Pin 2 and thus have its speed controlled. This is not ideal for this fan design, but it works. So a third-party Fan Controller maker can design their unit to do only the Voltage Control Mode and that WILL control the speed of either fan design. HOWEVER, that is a problem when you try to add a HUB into the system. The HUB cannot distribute the varying voltage from Pin 2 to its fans because it connects to its pin 2 outputs ONLY the fixed 12 VDC supply from the PSU. Further, IF the Controller maker designed this way, there is NO PWM signal on Pin 4 even thought the pin exists, so there is no source of PWM signal to send to the 4U unit's built-in Hub system.

So, your big question to answer is: how was that third-party Fan Controller designed? IF it really does use the newer PWM Mode control system, you can find the cable from your 4U unit's backplane that normally would go to a mobo fan header, and plug it instead into an output from the new controller.

However, IF the new Controller only operates as an older 3-pin Voltage Control Mode system, you cannot do it that way. Then what you could do is to ignore all connections of the fans in their caddies to any backplane sockets. Since the fans themselves appear to be normal 4-pin PWM fans with standard connectors, you could route all of their cables directly to the new Controller's outputs. CHECK three items before doing this:

1. What is the max current draw of each fan?
2. What is the max current output rating of each output port of the new Controller?
3. What is the max TOTAL load on that Controller for ALL fans plugged into it?

If the Controller can supply all the power required by all the fans mounted in the 4U unit, you can do it this way.

Of course, doing this will NOT give you the automatic control of fan speeds based upon measured temperature at a sensor on the mobo when there is NO mobo with its built-in Controller. HOW the new third-party Controller decides what speed the fans should run depends on its design, but many are simply manually-set speeds.

 

[Furi0usGe0rge]

Thanks for the thorough response. All of my confusion results from a bad assumption on my part:

The case I'm working on is very similar but not identical to the Norco RPC-4224., which I also own a couple of.

That case does NOT come with a PWM uplink cable, yet the fans don't run at full speed... or so I assumed...

When Norco became defunct, I sourced similar cases from China. One way these are NOT identical to the Norco is that the fans are about 10 mm wider.

As I should have realized from my experience with the wide 80 mm hot swap fans on Supermicro's 2U chassis, these new fans are not your Granny's gentle 1800 rpm desktop case fans. Instead these are your drunken Peepaw's angry, screaming, finger splitting, 'nam scarred, high pressure 5000 rpm industrial fans, like one finds on ASIC miners. Duh!

So, in fact both cases' fans are operating at max capacity whenever the PWM headers are not connected. I just never noticed it before now, since 120 mm 1800 rpm case fans like the Norco have are pretty quiet.

You are correct that my fan controller works by modulating voltage.

You are also correct that the new case does come with a separate cable for PWM, surely necessitated by how loud it is without one. It's actually 3 male headers. I found a thread on reddit about this fan backplane. It has a handy picture:

You can't see that when it's mounted in the case. You can barely get the molex on it.

One guy on reddit finally got it:

Three sets of PWM/tach signals from the mobo, one distributed to each 4-pin connector?

I did not mention this cable in my OP, though it does come with the new case, because I thought it was broken. I didn't understand why the headers would not power any of the fans. I could not get a 12V volt reading because it only passes through the PWM signaling, which was not present anyway.

As you suggested, I was making this work by threading the fans' PWM connector through the backplane assembly, and connecting it straight to my voltage regulating fan controller.

Unfortunately, it is a PCI mounted fab controller, and I fried two of them so far while troubleshooting tgese fans, by dropping them into the case with no motherboard, and creating a short circuit.

For my next attempt, I think I will make a 12V to 5V molex "converter" cable. If that's still too loud, I'll revisit fan controllers.

Thanks again!

 

[Paperdoc]

That photo you posted shows what appears to be three two-pin connectors at top. Or MAYBE that is two 3-pin connectors. Is that where the actual fans plug in? I am trying to grasp what connections are being made to the fans. Maybe there are connections made from those six pins to some backplane bus that the fans in their caddies connect to?

Making an adapter to feed reduced power from a Molex supply connector to fans has some issues you should understand. The first is that this means the fans are always running at one speed, and in some cases that may not be enough to provide all the cooling your system requires. The alternative of connecting each fan to the output of a Controller with manual settings possible leaves you the option of setting the fan speed to whatever you choose, even if that does means fixed speed until you decide to change it. But you CAN change it if your system need different cooling.

Secondly, there are three ways you can make power connections from a Molex to a fan. At the female Molex 4-pin connector, the RED wire is the +5 VDC supply, the BLACK wires (two in the middle) both are Ground, and the YELLOW wire is the +12 VDC supply. "Normal" uses are Red-to-Black for +5 VDC supply pair, Yellow-to-Black for a +12 VDC supply pair. But you also can do Yellow-to-Red (Yellow becomes +, Red becomes - or "Ground" but not really Ground) to get a fixed 7 VDC supply pair. For a fan, 5 VDC is VERY slow usually, and SOME fans are near to stalling with that power supply. Providing 7 VDC normally will guarantee the fan will not stall, but it will run somewhat faster, although not nearly as fast (less noise and air flow) as a 12 VDC supply.

 

[Furi0usGe0rge]

That photo you posted shows what appears to be three two-pin connectors at top. Or MAYBE that is two 3-pin connectors. Is that where the actual fans plug in? I am trying to grasp what connections are being made to the fans. Maybe there are connections made from those six pins to some backplane bus that the fans in their caddies connect to?

Yes, it's 3x2pin connectors, to send the PWM circuit to the 3 fans, from three motherboard headers.

If you could extend the previous image to the left, you would see the backplane with three 4-pin PWM fan connectors. The fans in their orange caddies are slotted into the "backplane assembly", such that connector lines up. Here you see the fans from the top on a similar case:

It's actually just as elegant as the Supermicro design (perhaps just copied).

The molex connection seen in the previous image can be found outside the bottom right corner of the right-most fan.

Making an adapter to feed reduced power from a Molex supply connector to fans has some issues you should understand. The first is that this means the fans are always running at one speed, and in some cases that may not be enough to provide all the cooling your system requires. The alternative of connecting each fan to the output of a Controller with manual settings possible leaves you the option of setting the fan speed to whatever you choose, even if that does means fixed speed until you decide to change it. But you CAN change it if your system need different cooling.

Secondly, there are three ways you can make power connections from a Molex to a fan. At the female Molex 4-pin connector, the RED wire is the +5 VDC supply, the BLACK wires (two in the middle) both are Ground, and the YELLOW wire is the +12 VDC supply. "Normal" uses are Red-to-Black for +5 VDC supply pair, Yellow-to-Black for a +12 VDC supply pair. But you also can do Yellow-to-Red (Yellow becomes +, Red becomes - or "Ground" but not really Ground) to get a fixed 7 VDC supply pair. For a fan, 5 VDC is VERY slow usually, and SOME fans are near to stalling with that power supply. Providing 7 VDC normally will guarantee the fan will not stall, but it will run somewhat faster, although not nearly as fast (less noise and air flow) as a 12 VDC supply.

Thanks again for the reply. I had no idea that the +12V and +5V could be combined to make +7, rather than a short. I will definitely try it both ways (5 and 7V).

I found that before I friend the fan controller, the industrial fans would spin faster than the desktop fans, at less than 5V, which I'm just estimating from how far I'd turned the knob. It could be that the Norco case was under-engineered, but my use of the disks is relatively modest, so I will live on the edge.

I once forgot to turn plug the fans in all together, and the case got too hot to touch. Some of the fs got corrupted, but all the hardware survived. That's when I learned that HDDs are more resilient than I thought.

I haven't checked, but I suspect that there are fan controllers with external thermometer inputs that can then send PWM data. I've seen asus motherboards that do something simialr, but in that case a separate cable interfaces the fan controller with the mortherbaord, and it is configured through the BIOS.

Since I'm using this case as a SAS enclosure, I could just send the PWM up to the headers for the backplane (the cases are already connected by an eSAS cable), but I wonder if that will be worth the effort, The drives are afterall in another case entirely, so what information of value could it use to set the PWM, other than CPU temp of the processors in the other case? Maybe that would not be so bad, since it will correlate pretty well with disk IO, but I'd have to calibrate it manually.