Question Blocked PSU Fan - - adding side fans ?

[dansdiy]

Hey everyone, this is my first post. Been lurking for a longtime and it seemed like a good place to post a question like this.

I'm building a 2U server (with a custom chassis - using 3D printed parts and a 2U rack mount shelf) for my homelab and while deliberating power supply options just decided to get a (used) regular ATX PSU - obviously the fan will be blocked because it will be against the top of the chassis. I figured I'd just cut holes in the side of the sides of the PSU case and add 80mm fans because apparently I like doing things the hard way... I originally purchased a Corsair "Shift" PSU (with connections on the side), because it seemed like it would be best to blow straight through the PSU case, in one end and out the other. The ebay seller ended up sending me a regular RM850x - I couldn't exchange it for the Shift model because they didn't have one, so I didn't bother returning it.

Point is, I've got the thing open and now I'm deliberating on fan placement. As pictured, the PSU has clear plastic insulation running along the sides of the board. Pretty sure I'd need to cut this to get good flow. So I have a few questions I was wondering if anyone here could provide their opinions on. Firstly, is this plastic entirely necessary? Seems like it obviously makes the thing safer, it should make it harder for an arc to reach the case - but I've seen PSU's without this insulation. It seems obvious that I need to cut it, but I guess rather than cut it all down, the question is where should I cut it?

This seems to come down to two factors: which components are most likely to arc, and which components need the most cooling.
-I was originally going to put two 80mm fans on the left side of the PSU (first picture) since that would allow intake from outside of the server chassis. But in terms of components that might arc, my rudimentary knowledge tells me those big ass caps are likely candidates, but I could be mistaken.
-I could put one fan in the top left position and only cut back the clear insulation up to the capacitors -this should provide good air flow to the MOSFETS with heatsinks near the middle. This seems like a low risk option because the part I'd be exposing to the case by removing insulation is the PWM board (which I think controls the switching circuits, not the fan). I could pair this with another fan on the rear exhaust vent (second photo), but from what I know axial fans are better at blowing than sucking, despite these particular fans being labelled as "optimized for static pressure", so I'm not sure about the exhaust fan option.
-I could also have an intake fan on the other side (drawing from within the chassis), where I'd assume a safe position would be bottom right due to it mostly being occupied by the 120v connection and also being relatively open to air flow - the top right position is kinda blocked by the board that the original fan plugs into (however I assume that is a safer component to un-insulate as well). This fan formation might even help create a swirling effect inside the case that might help cool all the components (last photo).

I only bought two fans, but I'm still open to just putting two on each side - and maybe even a fifth fan on the back exhaust. I feel like all that might be overkill, but it might provide enough air flow that I might not need to cut back any insulation - just the air forced over the insulation might be enough at that point.

So many options! Which do you think is best based on the relevant factors? I'm leaning toward the swirl configuration plus an exhaust fan... combining photo two and three.

Jeeze this post really turned out a lot more wordy than I anticipated. I appreciate any and all of you who take the time to read it!

 

[Lutfij]

Welcome to the forums, newcomer!

Firstly, is this plastic entirely necessary?
It's to prevent an arc from happening when you have the metal shroud encasing the PSU's innards.

Adding the design for your case would also help us two fold. That being said, I'd make sure there's air flowing on top of the heatsinked areas. As it stands, where the fans are illustrated to go, most of your airflow is obstructed due tot he capacitors. 80mm fans in a concurrent day PSU would result in us having to deal with inefficient PSU's from 2 decades ago. What I meant by inefficient PSU's, look at the temp graph of a PSU and how it's efficiency drops when temps rise.

because apparently I like doing things the hard way
If I were in your shoes, I'd have gotten an SFX PSU. You're just creating problems to solve when you could've easily solved the design of your case without introducing new problems.

My 2 cents on the matter.

 

[Misgar]

Firstly, is this plastic entirely necessary? Seems like it obviously makes the thing safer, it should make it harder for an arc to reach the case - but I've seen PSU's without this insulation.

Having spent time analysing PSU designs in industrial and MILSPEC systems for Safety Critical components, including Creepage and Clearance distances, I have to say I'm slightly confused by the high sides of this clear plastic sheet.
https://resources.altium.com/p/high-voltage-pcb-design-creepage-and-clearance-distance

I can understand a PSU manufacturer might include an insulating sheet under the PCB, if their Quality Control is insufficient to ensure all component leads are cropped adequately and don't protrude too far from the PCB down towards the metal base plate.

Perhaps they include this insulation in case wires break loose from the PCB and touch the inside of the case. Unlikely, but who knows what went through the designer's head or the engineers vetting their design..

As you've spotted, not all PSUs include these high sidewalls of insulating material. I cannot recall ever seeing this in any of the PSUs I've examined.

As for arcing, it takes roughly 3,000 Volts per millimetre before an arc occurs in dry air. With the highest voltage in an ATX PSU being typically 340V DC (240V AC x 1.414), we're nowhere near the 3kV needed to jump 1mm. Gaps between components and the sides of the case are greater than 1mm. Go figure.
https://kindle-tech.com/faqs/what-is-the-voltage-for-arcing

I'm inclined to agree that a standard ATX PSU is not ideal in a low profile 2RU server case. Have you considered chopping a hole in the lid to allow the fan to breathe?

However (and this is my personal opinion) I'd have no qualms about trimming back the plastic side panels to permit better ventilation. You still might not achieve ideal ventilation with side fans, but if they create enough air flow, you should be OK.

When all said and done, it's your own life (and your family's) that's potentially (pun intended) at stake.

I'd feel much happier if you bought a "proper" server PSU. Of course they're bound to be silly prices if you buy new, but the risk is far smaller. Take care.

 

[geofelt]

Can you supply a photo of the case and the proposed locations of parts within it?

On most pc's the most power is used by the graphics card under high load.
Where will your power draw come from? You have a 850w psu of which most of the 850w may not be used. Under those conditions a psu fan may not run at all.
You DO need some airflow. But, possibly simple openings may do it , even without fan assist.

 

[dansdiy]

Thanks for your input everybody!

That being said, I'd make sure there's air flowing on top of the heatsinked areas. As it stands, where the fans are illustrated to go, most of your airflow is obstructed due tot he capacitors.

Those heatsinks are definitely my main concern, however I think even if the air is being forced around the capacitors, since it will be flowing toward the vent in the back it should pass over those heatsinks reasonably well. Since this GPU doesn't have the ability to communicate temperature conditions (my old AX1200i had this feature) I just might need to add a sensor in somewhere to log and monitor temps, at the very least to give me peace of mind.

I'd have gotten an SFX PSU.

Darn, that looks like it would have been a good fit. I obviously didn't do enough research because I didn't know that form factor existed.

I cannot recall ever seeing this in any of the PSUs I've examined.

Well that's reassuring. My reasoning was similar, that a very high potential and seemingly already catastrophic situation would need to happen before that insulation would become necessary.

 

[dansdiy]

I'm inclined to agree that a standard ATX PSU is not ideal in a low profile 2RU server case. Have you considered chopping a hole in the lid to allow the fan to breathe?

I agree it's not ideal, but I saw other people doing it with flow-through style ATX PSU's (where the air flows from front to back rather that coming in from the side, this design doesn't seem very common anymore for ATX PSU's, I literally only found one and it wasn't a modular PSU) and figured I could just modify mine to a flow-through design. I did consider a hole in the lid, however my rack is not very big (12U) and depending on what I get into it might very well become full, so I didn't want to have to dedicate a whole unit to air space.

When all said and done, it's your own life (and your family's) that's potentially (pun intended) at stake.

I reckon it's not that serious of a situation, given the PSU has an auto-shutdown feature should it overheat. This should prevent it from ever getting close to a fire-starting temperature. Regardless, you've inspired me to add in a temperature sensor. I have a raspberry pi and some sensors for that already, I even have some relays attached to one that were set up to remotely hard-reset GPU mining rigs, so I'll dig that out and figure how to use it to trigger a shut-down past a certain temp. If you think this is a waste of my time compared to buying the correct PSU, I'm unemployed right now so you're wrong, my time is valueless and I have too much of it, lol. Plus it would be cool to log temps and see if my airflow solution was an overall success. I imagine 3 fans should be adequate. I could just run them at full speed, but I was thinking about trying to wire them to the PSU's fan controller.

 

[dansdiy]

Had to break up my responses due to spam detection...

Can you supply a photo of the case and the proposed locations of parts within it?

On most pc's the most power is used by the graphics card under high load.
Where will your power draw come from? You have a 850w psu of which most of the 850w may not be used. Under those conditions a psu fan may not run at all.
You DO need some airflow. But, possibly simple openings may do it , even without fan assist.

You're probably right, the load shouldn't be to intense, at least at first until I start playing around with new things. At the moment the only plans I have for this server is to run VM's for security camera software, a b.t.c. node, lightning node, BISQ node (maybe), Plex, and for general learning/homelab stuff - I have some IT experience but want to learn more. My NAS is a separate unit. The GPU's only job will really be transcoding. Here is a photo of a mock up of the parts. Please be gentle, it's just a mockup. That's a 2U rackmount shelf. I'm still working on the brackets for mounting everything. GPU will be mounted in a suspended position basically as shown. There will be a lid, as well as a front and backplate. And yes, the PSU is going to blow out the front...

 

[geofelt]

What is the cpu?
Would integrated graphics do the job?

 

[dansdiy]

What is the cpu?
Would integrated graphics do the job?

Just an i5- 8600k
GPU is an old 1060 my buddy had laying around. To be honest I'm not sure about integrated graphics transcoding, I'm totally new to running Plex. Pretty sure even a 1060 is overkill. But it wasn't being used so I figured I'd give it a purpose. As Kurt Vonnegut said in The Sirens of Titan: "The worst thing that could possibly happen to anybody would be to not be used for anything by anybody." lol

 

[dansdiy]

Thanks for the input everyone, I pulled the trigger and ended up with this

I thought about how I might track the PSU temps for peace of mind and I remembered that I have a handful of ATtiny85 devices from a prior project, as well as a thermister from the old hot end of my 3D printer (I upgraded to an E3D v6), and an MLX90614 contactless IR temperature sensor from an old project. I figure the thermister will be good for attaching to the switching MOSFETS to monitor their temps directly, since they generate the most heat. The IR sensor will be mounted to the inside top of the PSU "looking down" to monitor the overall temperature of the rest of the components.

I borrowed someone else's code from the internet for utilizing the thermister to output temperature data via virtual serial over USB for logging to an application on the server (https://www.hackster.io/najad/phone-thermometer-using-digispark-attiny85-991942).

As for the MLX90614 I couldn't find exactly what I needed so I just had ChatGPT write me a similar bit of code for that, then had it combine the code with the other code I found so I can run both sensors from the same device. Actually I can probably have ChatGPT design a simple script to shut down the server if the temperature gets too high, so no need for a fancy serial logging application, a simple one will do. Won't be able to test it until the resistors I ordered arrive in a day or two, I'll post again when it's all completed. Below is the IR sensor already soldered to a cord and heatshrunk (left), the thermister (top) and two ATtiny85.

I know "vibe coding" gets a lot of hate, but the way it's helped me blow through this little side project while also learning quite a bit from it (I have VERY little background in digital electronics and coding, but I asked ChatGPT lots of questions along the way and it's responses were both thorough and easy to understand) is really cool. We'll see how well the code actually works but from what I experienced just giving ChatGPT feedback in trying to get a version that will compile without errors and getting it down to a size that will fit on the ATtiny85, it shouldn't be too hard to work with it to fix any problems.