Can you "splice" a fans speed sensor to more than one connection?

[spyrule]

Hello,

So I've recently revived an older Dell DCS 6005 server as a temporary hold in for a failed NAS, until we can budget for a proper NAS.

In the interim, I've modified the chassis, took out 1 whole sled, cut the Top lid to accept 1 140MM Noctua ippc-3000 fan pointed down at each sled, and then removed the stock chassis fans that are rediculously loud, and replaced one 80mm fan on each sled pointed to the back, to create a front-to-back pressure direction. With this setup, my CPU's are in the ~30-35'c range, and the chipset float around ~40'c range under test load.

One of the issues that I'd like to solve is that the 80mm fans run at 2200rpm, which is above the min amount needed so that the BCM threshold alarm isn't triggered. Except that I only have two fans. Is it possible (or safe'ish) to splice the fans speed sensor wire into the un-used fan header sensor, so that the BMC thinks all 4 fans are running at the same speed ?

 

[nthu9280]

Hello,

So I've recently revived an older Dell DCS 6005 server as a temporary hold in for a failed NAS, until we can budget for a proper NAS.

In the interim, I've modified the chassis, took out 1 whole sled, cut the Top lid to accept 1 140MM Noctua ippc-3000 fan pointed down at each sled, and then removed the stock chassis fans that are rediculously loud, and replaced one 80mm fan on each sled pointed to the back, to create a front-to-back pressure direction. With this setup, my CPU's are in the ~30-35'c range, and the chipset float around ~40'c range under test load.

One of the issues that I'd like to solve is that the 80mm fans run at 2200rpm, which is above the min amount needed so that the BCM threshold alarm isn't triggered. Except that I only have two fans. Is it possible (or safe'ish) to splice the fans speed sensor wire into the un-used fan header sensor, so that the BMC thinks all 4 fans are running at the same speed ?

I think you can update the thresholds using IPMI. But don't know answer to your original question

Sent from my Nexus 6 using Tapatalk

 

[spyrule]

Yeah, the only problem is that the thresholds don't survive total power loss. Unfortunately the alarms also activate before even the BMCs POST is completed, so resetting the threshold before the alarms isn't possible. Hence, why I'm looking to trick the motherboard into thinking that 4 fans are connected, when in fact only two are. (I'd wire one fan each to one additional sensor).

 

[spyrule]

So in case, anybody is ever wondering if this can be done... the answer is.... YES! It seems as though each jump you make, you lose ~50rpm, but for me, that was well within the safe zone for me. Considering that I have my fan running at a constant 1500rpm, and the sensor threshold is 1000.

This solves my BMC from complaining. Especially after I did set the threshold to 1rpm and after 48hrs of continuous use, the BMC would check the thresholds, see that it was too low, and reset it to 1000rpm. Now it doesn't know any better. best of all worlds in this case.

 

[spyrule]

Here is a picture of how I made mine.

The only part that might confuse some people is the brown connector. It uses one of the 4 prong white connectors (right besides it as an example), where I have soldered the 4 short pins together to create a single circuit. This simply made for a cleaner, easier way to connect each wire without having to solder 4 separate wires together.

 

[pricklypunter]

You are probably losing signal integrity, translating into lost counts/ incorrect PWM feedback loop. This explains the missing/ incorrect RPM's. The likely reason is that the fan input is open collector, and by joining them all together you will be changing the resistor values on the board at each fan connector. This will cause the pulse edges to slowly fall rather than having the nice clean edge the inputs are expecting, it may also round off the bottom of the pulse if the fan is incapable of saturating the output transistor because of the loading. A fiddle here and there with some common resistor values would probably restore the signal to something more acceptable, it would be easier though to put a buffer between them to isolate the fan inputs, or opto isolate them. What you have done won't do any harm though and it's obviously working out for you being rigged like this, so this is just for information

Edit:
Well, that and the fan outputs will probably be overlapping each other, as the fans will neither start from the same degree of rotation, nor maintain exactly the same speed, plus any other little tolerances between them. So the output pulse will be widened, making the board think the fans are running slower than they actually are and adjust the PWM output accordingly. Worst case is when they all just overlap each other. Of course how sensitive that is depends on exactly how the fans are controlled, and I have never really looked at how they are implementing it, so can't really say much about it

 

[spyrule]

Well, actually, I only seem to lose 50rpm sense on the first hop, and every other after that they are identical. If this server was going into a high-temp, high use situation (I wouldn't be using this server tbh), however if I WAS, I wouldn't have used my quick fix. I'm actually attributing the 50rpm signal drop to my ok at best soldering skills.