1 / 6
Cisco NXA-FAN-30CFM-B and N3K-3232C Fan Silencing — zdjęcie 1
1 / 6

Cisco NXA-FAN-30CFM-B and N3K-3232C Fan Silencing

Zamów wydruk tego modelu

Obliczymy ostateczną cenę na podstawie wybranego pliku. Sprawdzisz ją i zaakceptujesz przed płatnością.

Zapisane na tym urządzeniu.

This is a WIP for replacing NXA-FAN-30CFM-B fans in a Cisco switch.

Ocena
0 / 5
Polubienia
0
Pobrania
3
Wydruki
0
Wyświetlenia
21
Oceny
0

Aktualizacja u źródła:

Informacje o modelu

Źródło
Printables
Kategoria
Computers
Licencja
Creative Commons — Attribution — Noncommercial

This set of files is provided with use at your own risk warning, the switch will have much less cooling when used as I intend.  I did not do anything to the PSU fans which are still somewhat noisy because on my N3k-3232, the PSU fans are used for cooling the bottom side of the mainboard and the bottom row of SFP cages.

This file package is a work in progress, although since it meets my MVP, I didn't plan to continue development.  If I were to continue refinement:

  • perhaps cleaning up the code in the scad file would help other permutations,

  • and adjustment to what I have used, the depth of the mount should be increased slightly to allow assembling the connector to the cowling for easier installation

  • a recess for allowing a nut to mount the fan (its currently too tight to use a screw and nut)

  • and a small cutout for cable routing added to the spacer because the two pieces of the original connector are needed to provide a flat surface to tighten things together (this is because for the connector to sit flat with the screws holding it, the original connector's cover is still needed.

Additional to crediting the creator of the fan spacer, I was also inspired by the author of this Noctua fan swap: https://imgur.com/a/quiet-fan-replacement-cisco-3850-fan-modules-CSs5CSL

Unlike that person, I was able to get a PWM fan to operate with the Nexus N3K-3232c running NX-OS 9.3, I don't expect any firmware pitfall to exist though.

I determined the pinout of the original NXA-FAN-30CFM-B fan when oriented as if you are looking into the unit left to right is (also note this fan is the red style, so in this orientation it would be blowing in your face if it were running:

  1. Inlet Tach (Blue)

  2. 12V (powers both stages of the fan) (one is red, other orange)

  3. PWM (connected to both stages (one green, other white)

  4. GND (Inlet), also has a loop over to pin 8 (black)

  5. Outlet Tach (yellow)

  6. Outlet GND (brown)

  7. No Connection

  8. Presence (was connected to pin 4) (black)

A few things to be aware of here are

  • the original fans I had don't seem to follow a typical 4-pin fan color code, or even similar fans' datasheets from Nidec

  • The switch will report "Failure" if the tach signal is missing on the second stage of the fan.  I never tried driving only the outlet tach to see if the behaviour was different, but I do know that with tach only on the inlet connection, it shows "failure". My switch will tolerate failure conditions and still allow logging into it, such as missing one fan, or missing a tach signal on one stage of the original fan.

  • without connecting pin 8 to ground, the fan was shown as "not present"

  • I used some blue single stage fans for donor connector parts, they are keyed so I needed to break off the incompatible key.

  • I think my switch supports either direction of fan (blue or red) so perhaps the presence connection identifies fan direction, I don't know for sure, but I did see other postings that those other people seemed to have a fan that had a different jumper connection.

Let's start putting it together...

If your switch is similar to my Nexus N3K-3232c You will need

  • two spacers printed (printed with the central column facing up, no supports)

  • one cowling printed as designed (printed with the flat face on the bed, with supports)

  • one cowling printed to be mirrored using your slicer on the axis to result in the overhang being opposite (also printed flat face on bed, with supports)

  • two of the shorter screws from the original fan

  • two 92mm PWM fans, such as the Artic P9 PWM PST model I used

  • soldering tools and accessories

  • heatshrink

  • Extra fan connector pins, such as Molex P/N 22-01-2037-P

Material: I suggest using something with a higher glass temperature, like PETG which I used since over time it could deform.

On the first of two fan connections, I used the PST connector of the fan to get GND to pin 8.

I then wired one Arctic P9 PWM PST Fan to the inlet connections which are the first four connections, and split the tach signal to a wire (going to the other connector) and to the outlet tach.

The pins inside the connector can be removed as the imgur link discusses.  They are typical fan pins, in a custom plastic assembly.  The Artic fan's pins were compatible, and any additional pins I used Molex P/N 22-01-2037-P pins and my ratcheting crimper to install them.  I suggest for the frustration saved you probably want to buy a proper crimper.

For the second connector, I wired pin 4 to pin 8, because without it the switch reports the fan is missing.  I then wired that loose Tach wire from the first connector to the two tach pins.  The wire between the two connectors was long enough to allow putting each one in separately (was required because the existing design doesn't have enough of a stub out the cowling and would need a spacer or to be slightly longer to continue to be compatible with the original screws, which was my original intent.  This two-piece construction allows you to still use the fan grill for EMI shielding.

To mount the fan I used zipties.

I then used the shorter of the three original screws to tighten the spacer to the cowling, and that's what holds the assembly inside the chassis

To install the assembly I grounded myself to the chassis by holding on to the chassis, and slid my finger in the fan opening pushing the connector in and confirming engagement of the connector to the switch, then did the same to the second connector.  I would not twiddle the connections with the power on since the switch would be down half its airflow, and if a pin pops out there is potentially a lot of power available in that area of the motherboard (poof!).

One final check the fan spins freely, then powered it on.

Its a new switch to me so I don't know what typical is, but at idle and how I'd originally set fans to min fan speed, its been running for a while and has at least 21deg C of headroom to minor alarm as per "show environment temp".  The fans are probably not running at full speed, rather 40%.

I should also note that computer fans aren't always good for all orientations, on my setup they are dangling the armature on the bearing, so if they prove unreliable they might need to have the orientation changed so they are more vertical to help the bearing last longer,

Tagi