Building a Wall-Mount Metal Frame for Desktop Watercooling Components

[Windows7ge]

How very frustrating...

 

So as it turns out through further testing even the module previously discussed won't be sufficient to directly control the fans...

 

This was my discovery after going through the trouble to wire it in.

 

 

Which by the way, when the old potentiometer died re-opening this box it reeked of burned electronics.

 

Now I do have one other tool that I've used in the past to drive computer fans via DC and it works very well, it's also much smaller. It's known as the LM2596 but there's a problem.

 

 

For size comparison.

 

 

The issue is that little blue box with the brass screw on it is known as a trimmer pot. And that is how you adjust the output voltage. It is very smooth, granular, and proportional, how much you turn it is consistent with a given voltage. Weather you're at 5V 7V or 12V. The problem is you need a screw driver to turn it and that's not ideal for on the fly changing of fan speed.

 

In the past I've tired changing out the trimmer pot for a potentiometer like we've been using in this project thus far and it kind of works but the granularity is gone. 1mm you're at 11.5V the next mm you're at 9V 2mm later it tanks to almost 0V. Serious PITA. Maybe @James Evens would be willing to throw in his two cents on how we might drive this little buck converter with a potentiometer.

[Windows7ge]

47 minutes ago, James Evens said:

Ping me in approx. 2 weeks time. Have a PCB and components in the mail for two circuits:

1. another one with a  potentiometer:

Read the datasheet. The problem you might run into is the limited rotation and with that large changes over small angles. A solution which is ugly but works are two trimmer pots like 100k and 10k for fine adjustment. Alternatively use one with a full or even two rotations. Linearity of poti is good enough (a metal spring scraping over a carbon film).

Also the small SMD ones I used in the past where more a set once type (specs guaranteed for 10 rotations) then continually adjustable.

 

Btw. you can buy or  print (never tried but if you give it shoot print it laying down) knobs for the blue trimmer pots.

 

2. What you are probably be interested in: Biasing the feedback voltage and with digital/software control. For that you have a µC and some DAC (in this case a R2R ladder but PWM with RC filter would work too) and OpAmp. 

 

(3.) Now the large but and why you don't need any of those: You can PWM 3 pin fans. In your case even easier as you use 4 pin fans so you don't need a mosfet at all. Just a PWM signal.

from the intel specification:

Most fans work with PWM slower then 21 kHz but generating 25 kHZ on any µC should be possible. Additionally you want a opto isolator.

PWM would probably be the smartest way for me to go. Would have to read up on it though since my technical skills stop at analog. Anything digital and I don't know what I'm doing but that would enable me to use a potentiometer for fan control. I've been told it can be achieved with a micro controller. Can probably find some youtube videos on the process.

[DJ46]

20 hours ago, Windows7ge said:

The issue is that little blue box with the brass screw on it is known as a trimmer pot. And that is how you adjust the output voltage. It is very smooth, granular, and proportional, how much you turn it is consistent with a given voltage. Weather you're at 5V 7V or 12V. The problem is you need a screw driver to turn it and that's not ideal for on the fly changing of fan speed.

 

In the past I've tired changing out the trimmer pot for a potentiometer like we've been using in this project thus far and it kind of works but the granularity is gone. 1mm you're at 11.5V the next mm you're at 9V 2mm later it tanks to almost 0V. Serious PITA.

Could you make a custom knob that has a flathead bit or just a flat piece of metal on the other side and turns the screw on the trimmer?

It might require desoldering the pot and attaching leads to it and the PCB but it would give you the amount of control you are used to.

[Windows7ge]

5 hours ago, DJ46 said:

Could you make a custom knob that has a flathead bit or just a flat piece of metal on the other side and turns the screw on the trimmer?

It might require desoldering the pot and attaching leads to it and the PCB but it would give you the amount of control you are used to.

I thought about this last night and almost exactly your idea to boot. Unfortunately due to the absolute tiny nature of the brass screw I can't think of any way of attaching a 3D printed part that would withstand the test of time.

 

Work for a while? Yes. Keep working and not fall off after a couple hundred rotations? No.

 

Tonight I swapped out the big module for the LM2596. The wiring method is identical. One big upside to the LM2596 is it's ability to maintain a set voltage even as the load changes. Something the larger module seems to have trouble with.

 

Currently printing a box to put the LM2596 in that will be screwed to the side of the current control box. It won't be the prettiest and will definitely look out of place but I'm not in the mood to fully disassemble and redesign the box to accommodate it. Maybe in the future if I'm particularly bored one afternoon.

[DJ46]

8 hours ago, Windows7ge said:

I thought about this last night and almost exactly your idea to boot. Unfortunately due to the absolute tiny nature of the brass screw I can't think of any way of attaching a 3D printed part that would withstand the test of time.

 

Work for a while? Yes. Keep working and not fall off after a couple hundred rotations? No.

Yeah, a 3D printed knob probably wouldn't hold up. What I had in mind is some form of enclosure for the trimmer that would hold it in place and also have a slot for the outer knob that you turn, which would have a thicker slot inside for a flat piece of metal or a hex hole for a screwdriver bit which would actually turn the screw on the pot.

 

It's probably an overbuilt mechanical solution but in my mind a lot easier than the electronic solution of making a custom PWM controller. If I had to use PWM I would probably just look for a way to attach the PWM signal generator Noctua makes to the side of the box and split the signal to all of the fans with power coming from somewhere else.

[Windows7ge]

8 hours ago, James Evens said:

Sorry to disappoint you this would be the first china module I know of using a genuine LM2596. Most of them use a IC which doesn't even has the right switching frequency.

Oh it likely is a China knock-off that they're just calling the LM2596. It's just a knock-off that's cheap as chips and works well for driving computer fans or other trinkets directly.

 

8 hours ago, James Evens said:

Polypropylen and like mentioned make sure it the force are distribute within the layer. Theoretically you could also glue it.

For the Potentiometer problem: did you tried printing a gear so you can get multiple knob turns onto a single potentiometer revolution??

I can't even get my printer to print ABS. I don't know if Polypropylene is possible. Everything you see is PLA. I did see online they make what appear to be brass knobs. I could line up the trimmer pot with the potentiometer hole and just keep one nearby. Kind of like a tiny screw driver.

 

And unfortunately no, I just jank wired in a B10K potentiometer. Same value as the trimmer pot. It has about a 270° range of rotation. Bit annoying that about 75% of its rotation only decreases the output voltage by about 1V but after that it just jumps off a cliff.

 

Using gears as a form of reduction is a very nice idea but there is no physical space for it. And due to my own inexperience in the design of this control box any multi-turn potentiometer would have to be in the same form-factor as the one in my blurry pictures and I'm thinking those don't exist.

 

So I dug my own grave. All because I didn't test the electronics before designing everything around what I thought would work OK.

 

1 hour ago, DJ46 said:

Yeah, a 3D printed knob probably wouldn't hold up. What I had in mind is some form of enclosure for the trimmer that would hold it in place and also have a slot for the outer knob that you turn, which would have a thicker slot inside for a flat piece of metal or a hex hole for a screwdriver bit which would actually turn the screw on the pot.

 

It's probably an overbuilt mechanical solution but in my mind a lot easier than the electronic solution of making a custom PWM controller. If I had to use PWM I would probably just look for a way to attach the PWM signal generator Noctua makes to the side of the box and split the signal to all of the fans with power coming from somewhere else.

I think I know what you mean. It would basically guarantee that the knob couldn't fall off because it's being actively clamped inside the enclosure. That could work but I know it'd be at least a full days worth of time to design and tolerances would have to be just right for everything to fit. If I was going to go that far I would probably just go James Evans route and gang up multiple potentiometers of different values in series to have a bulk & fine adjustment knob.

 

If I were to go the PWM route to be quite honest I might as well plug it directly into the motherboard. Just let the BIOS control the fans at that point.

[Windows7ge]

1 minute ago, James Evens said:

Oh that problem. Is it old? Maybe try some contact spray.

Otherwise check with a multimeter the resitance in various positions.

I could try that but I'm thinking the reason it's doing this is because the range isn't dynamic. This module supports up to 40V DC input IIRC so I'm thinking because my input is only 12V the 12V and lower range doesn't start until the pot/potentometer is 3/4th through its rotation.

 

That's one flaw of these modules. The input voltage doesn't appear to become the maximum of the range. The maximum is pre-determined.

[Windows7ge]

20 minutes ago, James Evens said:

Oh that problem. Is it old? Maybe try some contact spray.

Otherwise check with a multimeter the resitance in various positions.

You did just give me an idea though. What if I added a fixed resistor to one opposing leg of the potentiometer?

 

In theory it would increase the range of rotation 0 - 12V if it thinks the pot has already be rotated higher than it actually has.

[Windows7ge]

Well, I made a box and screwed it to the side of the existing box. It does very much look like an after thought of which I can't share a picture of at the moment because the camera that takes crappy mid-range pictures (good for macro and far away pictures, garbage for most of the mid-distanced pictures I want to take) batteries are temperamental. It "takes the picture" then just shuts off. When I turn it back on it doesn't save the picture...

 

Maybe it's for the best, you can never see what I'm taking a picture of anyways with how it's always blurry.

 

I really need a half decent camera. Phone camera was great for this but it shit the bed...need a new phone.

 

ANYHOW, I digress. For the time being we have a "working" fan controller (which I may or may not put the time into diagnosing later..., I'm thinking now though if I determine where 12V starts on the potentiometer I could use a lower resistance potentiometer and solder a resistor equal to the difference to the input. This would make the LM2596 think the range is 0 ~ 40V when it's really 0 ~ 12V and with a lower value potentiometer give me more granular controller over a larger turn radius.) I'm debating weather or not to run the PWM pin wires...I mean, I don't plan to need or use them right now but in the future...for now I think I REALLY wanna know if I have any leaks. So next on the agenda is filling the loop with distilled water and seeing where it drips out of.

 

Hopefully nowhere but I think that's setting my hopes a little too high. I have a spare GPU block if disassembling the one permanently ruined a gasket. I pray cleaning out the CPU block didn't hurt it's gasket because I have no spare of those and who knows what XSPC would tell me if I asked for replacements (which they probably wouldn't have).

 

We will see. Either Friday or Saturday.

[Windows7ge]

Quite a bit of trouble has both happened and been worked around since the last update and a good part of the solutions are thanks to James Evens's input.

 

We got the loop filled. It's just distilled water. May or may not add a glycol agent & corrosion inhibitor when it's back in stock.

 

 

As it turns out the dismantling of both of the waterblocks without suitable replacement gaskets (O-rings) did result in the XSPC GPU block leaking near the power delivery area.

 

The good news in this is I have a second block that I did not disassemble for exactly this reason. The CPU block appears to be behaving itself but it is significantly newer so the gaskets likely still had the elasticity to seal properly. Which is good.

 

I've already mounted gradularthe spare GPU block (that doesn't leak) and the system has been purging air for about the past 24hrs. Aside from the GPU block issue which I knew might happen there have been no other (un)expected leaks.

 

Now about the electronics. James Evens had brought up the idea of a bulk & fine adjustment knob for smoothing out the granularity of the DC output voltage adjustment knob. So turning it 1mm doesn't make it jump off a cliff.

 

His input made me wonder if the design of this LM2596 module in itself which accepts a maximum of 40V DC input didn't support a dynamic range when adjusting the potentiometer. It appeared to have this dead zone (about 75% of the rotation) where nothing would happen. I wondered thanks to him if this "dead space" was 13V - 40V.

 

And after some testing today. This was confirmed.

 

I removed the trimmer pot off a spare I had:

 

 

And wired in three potentiometers with the following values: 5K, 1K, 10K Ohms using the center tap on the left and right pots to adjust the min and max values that the middle pot could use.

 

 

As it turned out setting a minimum of 1.4K Ohms gave an output of about 6.5V. Good for direct driving some fans that you want quiet. Adjusting the pot on the right..did nothing, and as it turned out removing the pot on the right also changed nothing. So I don't know why this module even has a third leg if it doesn't use it.

 

Anyhow. If I had or if a 1.25~1.5K Ohm potentiometer exists/existed (because a 30 second Google search didn't show anything promising in the form factor I need) the range of 6.5V - ~11.5V is around 1500 Ohms.

 

To set a minimum voltage I needed a 1.4K Ohm resistor. A patient or prepared person would order or wait for a pack to be delivered but I figured hey, I can repurpose the trimmer pot to do this.

 

 

And I then just wired it in series with the correct leg on the LM2596 & replaced the 10K pot on the fan controller with a 1K.

 

 

With this 1K Ohm pot and a 1.4K Ohm resistor in series with the right input. I can adjust not only the output voltage from 6.5V - ~10.00V but the entire rotation of the potentiometer can be used and it is VERY granular. Much much finer output voltage control with this arrangement.

 

If I need the fans to go faster than they can on 10V I might as well hit the switch that shorts the input to the output anyhow so this will work fabulously.

 

Next up. Designing brackets that I can use to fix this whole thing to the wall. Additionally I'm going to look into replacing the thermal pads I'm using on the GPU block. They look to all be 1mm thick to me.

[Windows7ge]

2 hours ago, James Evens said:

now let's look again at your picture (picture + guess):

I've bought a number of these over the years from a couple different sources. Where you have CFF marked which is an unpopulated component there is sometimes a surface mount LED in other copies of this board. That's what that is.

 

Unfortunately the LED is in parallel with the output so if you need really low voltages the LED goes dim.

So I'm kind of glad this one doesn't have it otherwise I would have probably desoldered it myself.

[Xychic]

On 6/9/2021 at 6:13 AM, Windows7ge said:

PWM would probably be the smartest way for me to go. Would have to read up on it though since my technical skills stop at analog. Anything digital and I don't know what I'm doing but that would enable me to use a potentiometer for fan control. I've been told it can be achieved with a micro controller. Can probably find some youtube videos on the process.

I have made a PWM controller from an Arduino nano, if you still wanted help with creating a PWM control circuit I would be more than happy to help. Just PM me!

[Windows7ge]

Sorry for suddenly going MIA on this project. I had a change in occupation and it wrecked 99% of my available time forcing me to heavily prioritize more important things that needed my attention.

 

So we left off with the whole loop filled, running, fan controls figured out and working smooth. @XychicThe offer is appreciated and maybe in future project's I'll take you up on that but for now direct DC control works fine. Just needed to reverse engineer the LM2596 circuit a little bit and figure out how to create an offset.

 

So in the time I've been gone I got the whole apparatus up and running on my desktop. Mounting it to the wall came with some challenges starting with the wall I chose only had a stud every 24". The whole jig is only around 19" wide. So using metal brackets from the old wooden setup I bought some extra aluminum rod and 3D printed some brackets.

 

 

 

The machine screws I ended up finding and using were way too long but on one had that gives me more options for clamping or other possible functions so I'm still debating weather to leave them or chop the excess off.

 

 

All of the wall mounts are 3" screws driven into studs so this thing isn't going anywhere. I might try and source some zip ties that are long enough to go around the H shaped blocks just for extra security but so far they've not had any issues. Still need to source the proper screws for inside the fan controller because apparently when I thought I ordered them I didn't?...anyways this concludes this project.

 

The performance isn't much to talk about. About what you'd expect for 2x 480mm rads. Heat rising from one into the other doesn't have a dramatic impact on performance. The second radiator's effectiveness is just negated slightly. Nothing to worry about. Barely a noticeable change.

 

I'll make one additional post here for people who may be interested in my next upcoming project when I get it started. It involves multiple 10Gig networking, fiberoptics, iPXE network boot and iSCSI so anyone who feels interested you'll hear about that in the not so distant future.