ICX7450 or 6610 Fan Mod Metal Case Lid + PULSE Generator Roundup & Mods

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

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
Got a lot of great ideas from these threads:





I just purchased am ICX7450-32ZP. It's been my dream switch for many years, but I always knew it would be too loud. Others have outlines the pinout of the fan connector and how to trick the power supply into thinking that it has a stock fan, so I am going to post the differences in my mods. Firstly, I purchased a nonfunctional ICX7450-48 to gain an extra top cover (I did NOT want to use acrylic), power supply, fan, power supply blank, fan blank, and other miscellaneous parts. I purchased the necessary Molex fan connector for the fan housing and crimp pins for the connector. I also used this pulse generator and set it up on my oscilloscope for 50% duty cycle and 650Hz.

1718250008191.png

I immediately replaced the caps with high quality Nichicon. However, the fan mod sill failed at first because the pulse generator gets very finicky close to 50% duty cycle and as I approached 50% the waveform is lost completely. I thought I had it in a good spot, but it would run a while and then fail. Luckily, the 50% duty cycle widely believed to be "needed" by the Brocade switch to fool the fan sensor is not really exact. I increased the duty cycle to 70% and everything has been fine since.

In my mod I also retained the green/amber LED in the original fan board.

In my junk 7450 I removed the rails for the power supplies by drilling out the rivets. This allowed me to make a couple practice holes in the case on the back. I tried a new to me tool called a "nibbler" to cut the circle holes for the 140mm fans. I must say, where has this tool been all my life! And Amazon had it on sale for $20.


I also used the scrap case to screw the top back into when I made my real holes in the top cover. This did a perfect job holding the top steady and providing clearance for the nibbler tool (especially since the power supply rails had been removed earlier).

This was really a dry run and I plan to redo the whole mod. I'm going to replace the trimmers on the pulse generator with fixed value resistors, make a second Molex connector that is neater than the first one, and generally clean up the wiring. You can see that I just bunched up the fan wires near the fan control circuit and I'd like these to be short and to the point.

Here are photos:

1.jpg

2.jpg

3.jpg

4.jpg

5.jpg

6.jpg

9.jpg

7.jpg

8.jpg

I can just plug the fan box into the chassis with the pulse generator in it as you would an ordinary fan box. Only difference is the wire that runs from the pulse generator to the power supply fan plug, and the power wire that runs to the fans. I may actually pickup 12V someplace else on the final design.
 
Last edited:

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
I should have posted this. Running about 20 hours now:

Code:
SSH@ICX7450-32ZP>show chassis
The stack unit 1 chassis info:

Power supply 1 (AC - PoE) present, status ok
        Model Number:   23-0000142-02
        Serial Number:  BLD2M41MTTS
        Firmware Ver:   32.67
Power supply 1 Fan Air Flow Direction:  Front to Back
Power supply 2 not present

Fan 1 not present
Fan 2 ok, speed (auto): [[1]]<->2

Fan controlled temperature: 42.8 deg-C

Fan speed switching temperature thresholds:
                Speed 1: NM<----->72       deg-C
                Speed 2:       62<-----> 85 deg-C (shutdown)

Fan 2 Air Flow Direction:  Front to Back
Slot 1 Current Temperature: 29.3 deg-C (Sensor 1), 43.0 deg-C (Sensor 2), 39.0 deg-C (Sensor 3), 29.0 deg-C (Sensor 4), 36.5 deg-C (Sensor 5)
Slot 2 Current Temperature: 32.2 deg-C (Sensor 1)
        Warning level.......: 70.0 deg-C
        Shutdown level......: 85.0 deg-C
Also, I put a temp probe at the power supply outlet. The air coming out of the PS after the mod is about 29.5C, before the mod it was about 33.0C.

Everything is cooler and I have the fans running at about 65% speed (guessing).

Here is my fan control:

1718321355881.png 1718321378266.png

And yeah, I added a heat sink to the transistor.
 
  • Like
Reactions: MrGuvernment

MrGuvernment

Member
Nov 16, 2020
47
11
8
Love this, thank you for posting.As someone with a 6610 just sitting waiting to be used (moved to a new house and want 40G to my TrueNAS and a breakout to my pfsense firewall) I need to get on board and either do the fan mod trick, or at least mount some 240mm fans on the lid and wire them in direct vs using an external power source.

didnt know about a nibbler (diff brand name but prob same)
 

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
Love this, thank you for posting.As someone with a 6610 just sitting waiting to be used (moved to a new house and want 40G to my TrueNAS and a breakout to my pfsense firewall) I need to get on board and either do the fan mod trick, or at least mount some 240mm fans on the lid and wire them in direct vs using an external power source.

didnt know about a nibbler (diff brand name but prob same)
I think all those nibblers are made by the same company. When i got mine Amazon had it on sale and it was even cheaper than AliX.

Don’t forget that you’ll need to use something as a pulse generator to fool your switch into thinking there are fans in it running around 650Hz. I actually just ordered four different pulse generators from AliX to test them all; cost me less than $10 for all of them together.

I’ve also been experimenting with black aluminum duct tape. I blocked off the top row of vents (1 of the 2 rows) above the NIC ports and partially blocking the side vents. This was to try forcing more air through the power supply (I removed fan from power supply too). To my surprise, the air temperature coming out of the PS actually went up 2C. So i think there is a balance between pressure and free flow through the entire unit.
 

MrGuvernment

Member
Nov 16, 2020
47
11
8
Ya, I know someone else did the mod where they just added the 240mm fans on top, which gave enough cool air, the PSU fans and other fans never rev'd up high enough to be annoying, that was the route I was considering, but would also be nice to do it all and properly like this and how you nicely mounted everything internally...
 

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
When I get all these I’ll post results with photos on my oscilloscope. We want about 650Hz and about a 50% duty cycle square wave.

Arduino. I have this here but didn’t use it. Personally I think this is overkill and not needed for this job. All we need is a square wave with about a 50% duty cycle. But some have found this more reliable. I think it’s do to low quality caps and trimmers on the pulse generator boards, not because this is better.

Board I’m using now. Works fine but flakes out as I approach a 50% duty cycle. I think the problem is just low-quality trimmers. Near 50% the wave form just disappears and the LED goes out, but 70% is very stable and still works fine to fool switch. The frequency also changes each time I measure it with the scope and doesn’t change with every turn of the trimmer. I ordered a second and will replace trimmers with fixed value resistors to get correct frequency and duty cycle. I’ll post results. The waveform is a very nice square wave though (I did change the caps before I even used it. The caps on it are crap and one was even badly dented).

And I ordered these:

I like this one because it does not come assembled. I can use higher quality caps right off and solder fixed value resistors.

This one has an LCD display showing duty cycle and frequency. It also has solid state buttons to set both. I’ll see how accurate it is on the oscilloscope. May be a great option especially if you don’t own on oscilloscope. However, there are chips in both sides of the PCB so it should be mounted in a way to allow for an air gap on both sides of the board (shouldn’t just glue it down).

This one I like because it’s the simplest and smallest design. Also $0.78 USD (though they are obviously all cheap). Very similar to the one I’m using now and I expect it will need better caps and trimmers removed.
 
Last edited:

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
Ya, I know someone else did the mod where they just added the 240mm fans on top, which gave enough cool air, the PSU fans and other fans never rev'd up high enough to be annoying, that was the route I was considering, but would also be nice to do it all and properly like this and how you nicely mounted everything internally...
I find the stock fans unbearable even at their lowest speed. That said I have sensitive hearing and the network rack is right next to my theater room now, and will be in a rack next to both my office and theater room in the new house. I can hear the stock switch through the wall even when the fans run on low. I could also hear the ICX6450 with 40mm Noctua fans through the wall as well. Office that would have been OK, but unacceptable in my theater room.

I even had to modify my ICX6450 with a 120mm fan because for me that switch was too loud, and it’s the quietest switch except for the passive ones.
 
Last edited:

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
Love this, thank you for posting.As someone with a 6610 just sitting waiting to be used (moved to a new house and want 40G to my TrueNAS and a breakout to my pfsense firewall) I need to get on board and either do the fan mod trick, or at least mount some 240mm fans on the lid and wire them in direct vs using an external power source.

didnt know about a nibbler (diff brand name but prob same)
One more thing, you have me thinking. I run 10GbE optical to my unRAID server right now. I use it to serve media to the whole house (four of us) and it runs VM's that I use as my primary computer (I'm typing on a Windows 11 VM right now on it). I've never needed more than 10GbE, but using a 40GbE would be fun... and possibly useful. I have no trouble serving four simultaneous UHD Blu-ray REMUX's with 10GbE. The highest bitrate I have seen on a single stream is with the JVC demo disc which runs at about 95 MB/sec (super high quality UHD at 60Hz).

The unRAID server runs both PLEX and KODI so it is heavily utilized. I have PLEX DVR with two HDHomeRun FLEX's (so eight tuners) for all our live TV with an OTA antenna (loads of channels here). Everything else is KODI. My internet speed is 3.5GbE (it's awesome, I only pay for 1GbE but they don't cap it!).
 
Last edited:
  • Like
Reactions: MrGuvernment

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
With all my free time, I'll test them soon.

IMG_1447.jpg

The trimmers are also as I expected, at the bottom of the clockwise range as the value approaches zero ohms, the trimmers cut out. I suspect for duty cycle a 1-2 ohm fixed value resistor will work, possibly even just a jumper.
 
  • Like
Reactions: MrGuvernment

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
The unit on the top left and the DIY board on the top right are the exact same circuit. In my final design I used the one in the top left only because it's a bit smaller. I stripped all unnecessary components for our application, removed the jumper block and soldered in a jumper wire, swapped junk caps for Nichicon, swapped the variable resistors with fixed values, removed the NE555P socker, changed the LED to a green surface mount, and cleaned everything else up. The resistor values I used are:

Duty cycle 150 ohm.
Frequency 1,000 ohm.

That got me very close to a ~50% duty cycle and landed at ~635Hz and I had those resistor values on hand. Changing caps did also change frequency quite a bit. The ICX7450 switch accepted this willingly. Here are photos of my final.

IMG_1462.jpg
Please note that the actual P-P voltage is 12.2VDC (probe was set to 10x and scope to 1x).

IMG_1458.jpg

IMG_1461.jpg

IMG_1465.jpg

IMG_1467.jpg

I will say I think that the top center board is actually better than the other two trimmer designs. Had I tested it on my scope first, I would have used that one. I still would have changed the capacitor and trimmers for fixed value resistors no matter what board I used. I'll follow this post shortly with a mini review of the other two boards.

IMG_1471.jpg
 
Last edited:
  • Like
Reactions: MrGuvernment

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
The top middle was the best circuit design out of the box. It could do 50% duty cycle unlike the other circuit and was very easy to dial in the frequency. If I use it, I will definitely replace both 100uf capacitors with high-quality caps. I would also replace the variable resistors with fixed value resistors. However, even though this board uses the same lousy variable resistors, it's design lets it hit 50% duty cycle with room to spare so it may be safe to just leave them alone especially considering the duty cycle slop the brocade switches tolerate. If you set it to 55-60% duty cycle I don't think there is any chance the waveform will ever fail from the trimmer being turned too low.


Please note that the actual P-P voltage is 12.4VDC (probe was set to 10x and scope to 1x).
IMG_1470.jpg
 
Last edited:

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
If you don't have an oscilloscope, I'd say the pulse generator with the LCD screen is the way to go. It displays the duty cycle and frequency right on its screen. There are four micro switches to increase or decrease frequency and duty cycle. The LCD display matches my oscilloscope perfectly. The only down side to this PCB is that it has IC's on both sides and that it is the most complex. I worry about complexity and things breaking. If you use this one, put it on stand offs and screw or glue those down to allow air circulation under the PCB.

Upon arrival, the power supply pins for my LCD were not pushed down all the way when it was soldered. It probably would have worked just fine, but I quickly fixed this. I also cut off the long leads from the LCD screen on the bottom side.

Here are pics:

IMG_1455.jpg IMG_1456.jpg

IMG_1469.jpg

Please note that the actual P-P voltage is 13.0VDC (probe was set to 10x and scope to 1x).
IMG_1468.jpg

Nothing to replace on this PCB. Other than fixing the power supply pins on mine that's all I did except cut off the extra-long leads on the back side. This thing works wonders.
 
Last edited:

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
All this really makes me wonder how accurate any of this really needs to be in order to fool the switch. I doubt many (anyone) else has put their PCB on an oscilloscope to set it. Both the frequency and duty cycle are no where close to what is required, so if you just buy a pulse generator and put it in, you are nowhere close to the correct frequency or duty cycle. The duty cycle needs to be very close to the bottom range of the trimmer resistor, but when you get about five full turns from the bottom, the variable resistor trimmer runs past its range and the waveform signal goes dead. I bet most people who have done this mod with a pulse generator have their duty cycle above 70% because that's where it is in about the middle of the trimmer's range. Also, the frequency when on the correct jumper across the variable resistor's 10kohm range, is between 50-1,000Hz. If you just spin it, you'll land anywhere in that range. How do you have any idea if you are near 650Hz?

If you have used a pulse generator and had issues, it's most likely because you're too far out of range on duty cycle and or frequency. I also think in a switch of this quality, that quality components should be used. The cheap junk caps must be changed, they start out bad and they will not last. Also, the variable resistors are the worst I have ever seen. I got the above four boards for about $10 on AliX... what can one expect? You need to polish these boards up a bit if you want them to work well and for the life of your switch.

If you don't have an oscilloscope, then I highly recombed the generator with the LCD screen.
 
Last edited:

MrGuvernment

Member
Nov 16, 2020
47
11
8
Toys toys and more toys!

I did get my fans in, but as you noted, even with the fans over the hot parts, leaving the stock fans as is, it still has some decent volume out of it (I still need to cut the lid and mount the fans vs just sat them on top to test for sound) And I only have a single PSU and single fan tray in mine.

So I may look to venture into doing this properly as you have done to minimise the noise.

My final set up I am thinking to do is
  • Main Workstation ----> 40Gb to ICX6610
  • TrueNAS ----> 40Gb to ICX6610 (same VLAN as my main workstation)
  • PFSense ----> 4x10Gb Break out to ICX6610

Since PFSense starts to limit out around the 10Gb mark with out extreme tweaking, thinking if I split the 4 x 10GB across 2x dual port 10GB sitting in their own PCIe 3 x16 slots (Just upgraded my pfsense to a Dell 5820 from the old HP SFF it was running on) I got some extra PCIe slots to work with now and for some reason am fixated on tweaking the crap out of pfsense to milk ever last drop of raw speed I can out of it (vs just doing ACL's on the BrocadeICX to get wire speed..lol what fun is that!)

Then I can put my specific high traffic VLANs on their own interfaces or LACP ones to assure any cross VLAN traffic that will happen, has as much bandwidth to abuse as it can.
 
  • Like
Reactions: CIR-Engineering

CIR-Engineering

I am a functional adult?
Jan 14, 2021
153
59
28
48
Chicago USA
www.cir-engineering.com
What is your line speed? I thought I was killing it at ~3.5GB. I've got OPNsense on an HP T740 thin client with a two SFP+ port Mellanox board (I have forgotten which model). RJ45 transceiver goes to the company modem at 5GB, and passive DAC cable to 7450 at 10GB.

I'm only using a single PSU now with the fan removed and a wire going from the signal generator to the fan plug in the power supply. I also have only one fan tray with fans removed. My only cooling is with the two 140mm Noctua fans. The exhaust from the 7450 power supply actually went down a couple degrees C after the fan mod. You can't get a temperature report on the PS from the Brocade, so I put a digital thermometer at the PS exhaust before and after the mod. Air exiting the PS now is around 30C.