Switch Case Mods (Acoustic)

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
Alright, so I'm copying some of this over from the end of another long thread, but the situation is this:

I have a 72 port 10G switch. It sounds like a leer jet. I want it to be quiet.

Attempt #1 (swapping out fans for 40mm Noctuas):

While converting the Sunon fans to the Noctua's I made the change one module at a time, checked the switch via terminal and the led's on back, and each time it was happy. Temporary end result = silent switch that was maintaining a temp of ~31DegC!





What I did not do was reboot the switch....

Now having rebooted, I'm locked out due to "fan failure". I replaced 6x 4 wire PWM fans with 6x 3 wire non PWM fans... because it didn't lock me out as each group was swapped, I made the (incorrect) assumption that it would let me run like that as long as temps stayed in check.

Apparently it will not.

Does anyone have a technical paper explaining exactly how PMW works? I have to assume that the switch is determining fan function based on the PWM signal... so what's the best way to feed it false info? Is it simply a voltage signal (I'm assuming not.. hence the "pulse" in PWM".

I'd rather go back to AOL dialup than listen to the Sunon fans run 17K RPM.

Other Possible Options:
1. Water cooling the damn thing (complicated/overkill/cool & silent)
2. Building/buying a new chassis that allows for larger fans and transplanting the internals (also a pita, current internals are well engineered with ducting/flow control... hate to tear that up).
3. Building a 2U cover for the existing base chassis... would have to rebuild internal ducting, but not too bad.
4. Building a plug and play hot swap adapter for each of the three fan modules allowing for a larger fan (overhanging the main chassis). This appeals to my inner engineer and is probably the cleanest/coolest option... would take some fab work to do it right. I can get one of our packagers to cut it out if I put together the drawings, but still not a get it tonight option.
5. Cutting a hole in the top cover just in front of the existing fan modules and mounting two PWM Noctua industrial 120mm 3K RPM fans... This is easy and definitely a get it done immediately option... but it feels akin to putting "spinners" on a 911 Turbo... such a nice switch; hate to rice it out. This would also be pulling intake air from the top of the rack which is not ideal... I suppose I could add some ducting to pull from the front of the rack, but now we've got spinners, neon under chassis lighting, and a wing out back :eek:

Pictures of the switch in question:









Who is selling these??


Any thoughts?
 
T

ttabbal

Active Member
Mar 10, 2016
747
207
43
47
Typically, PWM for a fan would be a control signal to change the fan speed. This is done by using a square wave, and varying the duty cycle. There are a ton of references around the net about PWM. I suspect what's happening with your switch is the power on test changes the speed and checks the RPM sensor to make sure that the fans do what they are told. Yours don't, so failure.

The only likely way to shut the complaining down is to use proper 4-wire PWM fans. That might also be a problem if the switch wants to see particular RPM readings. Faking that would likely require a microcontroller with a program that reads the PWM duty cycle and reports back modified RPM sensor readings.
 
P

pricklypunter

Well-Known Member
Nov 10, 2015
1,714
521
113
Canada
If the Mainboard is actually looking for specific ranges of RPM being reported back when it changes the PWM output to the fans, you're sol and will need PWM controlled fans, or as has already been mentioned, perhaps program a small Microcontroller to fake it. If on the other hand it's just wanting to see a non too accurate pulse train, then you could achieve that by throwing together a wee board with a 555 timer chip and a few passives on it and feed the output of that back to the inputs on the board. With a tweak or two over a standard circuit you could also probably fake the PWM signal using a 555 if so inclined :)
 
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
ttabbal said:
Typically, PWM for a fan would be a control signal to change the fan speed. This is done by using a square wave, and varying the duty cycle. There are a ton of references around the net about PWM. I suspect what's happening with your switch is the power on test changes the speed and checks the RPM sensor to make sure that the fans do what they are told. Yours don't, so failure.

The only likely way to shut the complaining down is to use proper 4-wire PWM fans. That might also be a problem if the switch wants to see particular RPM readings. Faking that would likely require a microcontroller with a program that reads the PWM duty cycle and reports back modified RPM sensor readings.
Click to expand...

Thank you for the response! I don't think the switch is looking for a specific PWM reading as I've swapped in a few different PWM fans of varying RPM ranges and the switch is happy. I've had a similar experience countless times when swapping 4 wire for 4 wire.

The issue here is that all of the PWM 40mm fans I've been able to source are extremely loud. While they serve a purpose for testing, they really don't address the primary issue. Generating a false PWM signal can't be so complex that it's unachievable.. it has to a function of something we already have available; just have to determine and implement the right modifier. We do this sort of thing on PLC systems and engine tunes/sensors/feedback loops all of the time. I suppose I'll read up on pulse width modulation and report back. In a worst case scenario I suppose there is always the possibility of feeding a PMW signal to the switch from an independent fan that is not actually providing cooling to the case while letting the quiet 40mm fans run at full speed...
 
Last edited:
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
pricklypunter said:
If the Mainboard is actually looking for specific ranges of RPM being reported back when it changes the PWM output to the fans, you're sol and will need PWM controlled fans, or as has already been mentioned, perhaps program a small Microcontroller to fake it. If on the other hand it's just wanting to see a non too accurate pulse train, then you could achieve that by throwing together a wee board with a 555 timer chip and a few passives on it and feed the output of that back to the inputs on the board. With a tweak or two over a standard circuit you could also probably fake the PWM signal using a 555 if so inclined :)
Click to expand...
Really need to know more about the logic in the switch.. For instance: I can pull the 6k/8k PMW fans out of a Supermicro 7048R-TR, SC847, or Chenbro SR107 w/SM board and replace them with 300-1300 RPM PWM fans.. no problem; it sees that the fans are running some % duty cycle and life goes on.

I'll do some additional reading and/or testing and report my findings.. I know it's impractical and an unnecessary hassle, but the water cooled 1U server above looks PRETTY.

....Maybe what I REALLY need is a water cooled server rack plumbed for individual devices.... Sounds reasonable right?
 
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
It would also be pretty slick to rip the guts out of the Gnodal switch and mount them under the MB tray in the SC847 chassis (replacing the rear 2U back plane and drive caddies.. 24 disk shelf in front, 72 port 10G switch in back.... :)
 
P

pricklypunter

Well-Known Member
Nov 10, 2015
1,714
521
113
Canada
That would be an utterly wild mod, you never know, could start a trend, but what to call it...aha...a SuperGnodal :p:D

Having thought a little more about matters, It finally sank in that you are replacing 4 wire fans with 3 wire ones. So there is a tach signal already being sent back to the board on pin3, it's actually the PWM control signal that's missing. So either the board is looking for a mirrored signal replicating that control voltage, looking for a much higher pulse frequency (faster fan speed) or it is doing something else, perhaps just simply monitoring the fan's current draw within a window detector :)
 
T

ttabbal

Active Member
Mar 10, 2016
747
207
43
47
Yeah, that's why I considered that it might be wanting a response from the fan. It changes the control signal, and wants to see the fan tach speed reported back. Exactly how picky it is about that is an open question.
 
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
sherwood said:
Not sure if those switches are the same but do they need 5v fans?

Amazon.com: Noctua A-Series Cooling Fan Blades with AAO Frame, SSO2 Bearing (NF-A4X10-FLX 5V): Computers & Accessories

You are using 12v (at least from the pic I think it says 12v) not sure if that matters but some of my switches needed 5v not 12v.

Haven't seen anything online about the fans in those switches in my limited browsing.
Click to expand...
The original fans are definitely extremely highspeed Sunon 12V fans
PN: PMD1204PPBX-A



 
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
pricklypunter said:
That would be an utterly wild mod, you never know, could start a trend, but what to call it...aha...a SuperGnodal :p:D

Having thought a little more about matters, It finally sank in that you are replacing 4 wire fans with 3 wire ones. So there is a tach signal already being sent back to the board on pin3, it's actually the PWM control signal that's missing. So either the board is looking for a mirrored signal replicating that control voltage, looking for a much higher pulse frequency (faster fan speed) or it is doing something else, perhaps just simply monitoring the fan's current draw within a window detector :)
Click to expand...

Strongly considering it.. Will need to look into potential EMI issues though...
 
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
Delta appears to have some interesting options as well... Although the last two (highlighted) have to be errors:
 
A

AJXCR

Active Member
Jan 20, 2017
565
96
28
35
Edit- All of the Delta fans in the list above are 2 and three wire with the exception of the AFB0412HB (although this appears to come in both 3 and 4 wire variants). They also make an Afb0412VHB that is 4 wire, but it's quite a bit louder.
 
You must log in or register to reply here.
Top Bottom