Controlling an externally powered 12V PC fan by fan output PWM pin?

[incognito]

To avoid breaking anything I'm trying to find out beforehand can I use a standard PC 4pin 12V PWM fan powered externally, most likely by a regular 12VDC wall charger, speed controlled by a motherboard 4pin 5V PWM fan output PWM pin alone? The reason why not to use only the connector is there at the end, as 5V PWM fans in required form factor seems to be practically nonexisting, aside from the original Delta AFB0705MC-00. So, 12V + and - would come from elsewhere and I would only connect the PWM and RPM pins from the fan connector.

According to white papers PWM control pin should always be the same 5V regardles of fan nominal voltage, be it 5, 12 or 24 V:
https://www.intel.com/content/dam/support/us/en/documents/intel-nuc/intel-4wire-pwm-fans-specs.pdf
https://noctua.at/pub/media/wysiwyg/Noctua_PWM_specifications_white_paper.pdf

To provide more background here, I've taken up a project trying to replace a noisy, low quality and poorly performing original fan of a monitor with G-Sync module inside, which needs active cooling. I've already tried to replace the original fan with a Noctua NF-A8 5V PWM, but the 25mm thickness of the fan was too much to fit inside. Noctua doesn't have any slim 5V models, so that leaves standard 12V as only option. If speed control could be viable by the PWM pin alone, I would next try NF-A9x14 12V PWM, which should physically fit inside easily.

Here's a video showing the NF-A8 5V PWM running from that fan output:
https://photos.google.com/share/AF1QipPKI1k3EwxKTpncRfAWzckV5PMLKG4x1dAh2nDvxdv5bLVpmmDG5c0_fWOnhZ2IrA/photo/AF1QipPYAq9AYLNfq35FEyE1mJ2JwrdJUrS1zspHqifN?key=NVc5SmVOX2hSdl9kMmJ1Tl9lb0hqeHk1NVJodnZR

In case you're interested I'm describing the project in more general terms here:
https://forums.overclockers.co.uk/threads/lg-38gl950g-3840x1600-g-sync-144hz.18840788/page-19

[dobsonr741]

Is the concern the new fan will mess up the mobo when hooking up?
Two options:
1. Study the new fan in isolation, to see if the PWM output is indeed 5V, and the PWM in not pulling up to 12V, and it properly controls PWM @5V input PWM signal.

2. Build a current limiting/voltage limiting interconnect. A 1K resistor and a 5.1V zener on both lines would do.

[DavidAlfa]

4-pin fans have permanent power and adjust their speed from the pwm signal, so yes, it should work.
Just don't mess up when wiring, power on the wrong pin might easily kill your motherboard.
Remember to connect together the grounds from the PSU, mobo and fan.

[Jwillis]

Powering the fan externally is not a problem. But you may have to have Both the PWM(drive) and Tach(sense) wires connected to mother board. The PWM needs feedback from the fans tach sense to maintain "speed" control from the PWM.  And as DavidAlfa pointed out the grounds have to be at the same potential for the fan to work from an external power supply.

[Nominal Animal]

Typical wall chargers in Finland are insulated ones, but I would personally use something like Mean Well IRM-02-12 (datasheet PDF) connected to the display mains input; TME and Mouser have the IRM-02-12S model (34mm × 22mm × 16mm, 34x27x16mm if including the legs; only four of the pins, two at each end are used) in stock for under 10€ apiece, and it seems to me soldering wires directly to it with suitable heatshrinking would make it easily placeable inside some nook or cranny in the enclosure.  (As it is 75% efficient, it'll generate at most 0.67W of waste heat.)

These are "class II" isolated, meaning there is no capacitor from either output to the inputs, so connecting the DC negative output to the motherboard ground should be safe.  IRM-02 only provides up to 2 watts or 167 mA at 12 V DC, but NF-A9x14 only consumes max. 1.3 watts.  If you want more headroom, look at IRM-03-12 (3 W) or IRM-05 (5 W).

[incognito]

Excellent. I couldn't even have hoped for a better input and suggestions. I might wait a few days if Noctua decides(/dares) to comment my enquiry by email (basically same what I've asked here) before proceeding with this. Also excellent suggestion by Nominal Animal. I love the idea of not having to run extra wire to the monitor from an outlet, but nevertheless I might first try it out with an 12V wall adapter to see if it works, as I have one ready with free wire ends.

To make sure I'm on the same page with everyone else, I've attached the schematics.

[MathWizard]

I wouldn't use external power from another supply. I'd just add a supply wire, back to 1 of the 12V sata or ATX cables. I'd even cut open an unused SATA cable, and solder right onto the 12V wire. What PSU is in your PC ? If it has any free GPU connector, or older style 4-pin cable, just twist up or fold some wire into a 12V pin or 5V if thats what you need, and tape it in place. Or with a header connector, and tap 12V off the mobo and run that single wire over to your fan Vcc.

[incognito]

This project is for a monitor. My PC is several meters away from my desktop and only optical display cables run from it to my desk, so unfortunately your suggestion in not viable option here. It also sounds a bit unpractical, at least in my opinion. Would you care to comment why you wouldn't use external 12V power supply? Thanks.

[MathWizard]

OK I missed that it's for a monitor. If it was me, I'd probably open up the monitor and look for a 12V rail, but yeah if the fan is a standard fan, and the wallpack is an isolated SMPS, like most are, then it should be fine.

[Siwastaja]

To make sure I'm on the same page with everyone else, I've attached the schematics.

Yeah, that should do it. There are theoretical failure modes with power sequencing (i.e., fan receiving the 5V control signal without having power supply present) but it's pretty far-fetched so I would not stress about it; such a product should be able to deal with it, else they would be failing left and right.

[incognito]

OK I missed that it's for a monitor. If it was me, I'd probably open up the monitor and look for a 12V rail, but yeah if the fan is a standard fan, and the wallpack is an isolated SMPS, like most are, then it should be fine.

How likely do you think it is that a 12V rail can be found inside the monitor? In a case there is one, is it unnecessary to interconnect the ground from that rail between the fan and the fan output connector, or just use the sole +12V line to the fan and only run GND from fan to fan output?

edit: Side note: There is that another 12VDC auxiliary fan, that's worth mentioning, but as it's a non-PWM fan that line is up only when that secondary fan runs, which is hardly ever, but perhaps that's a place where near to start looking for?

[incognito]

Typical wall chargers in Finland are insulated ones, but I would personally use something like Mean Well IRM-02-12 (datasheet PDF) connected to the display mains input; TME and Mouser have the IRM-02-12S model (34mm × 22mm × 16mm, 34x27x16mm if including the legs; only four of the pins, two at each end are used) in stock for under 10€ apiece, and it seems to me soldering wires directly to it with suitable heatshrinking would make it easily placeable inside some nook or cranny in the enclosure.  (As it is 75% efficient, it'll generate at most 0.67W of waste heat.)

These are "class II" isolated, meaning there is no capacitor from either output to the inputs, so connecting the DC negative output to the motherboard ground should be safe.  IRM-02 only provides up to 2 watts or 167 mA at 12 V DC, but NF-A9x14 only consumes max. 1.3 watts.  If you want more headroom, look at IRM-03-12 (3 W) or IRM-05 (5 W).

Is there any solution with similar output, but that can take ~20VDC input? If I recall right the monitor's DC adapter output is 19V or close by. That would allow me to solder the transformer's input lines onto the board's main power.

[Eraldo]

Why not just use a buck converter or a linear regulator like 7812. They are pretty small and can fit inside the monitor. As for the 7812 it might need some cooling so it would be best if you hold it down with screws or even solder it in the metal chassis of the monitor if it has one and is not close to the monitor display.
The buck converter you can find it pretty cheap on aliexpress or even on amazon (but costs more) and that one heats less.

[Nominal Animal]

Is there any solution with similar output, but that can take ~20VDC input? If I recall right the monitor's DC adapter output is 19V or close by. That would allow me to solder the transformer's input lines onto the board's main power.

Is the input ground directly connected to the fan ground?

If it is, a DC-DC converter like R-78K12-0.5 that Mouser sells for under 3€ apiece, ought to work well.  According to the datasheet, it is very efficient at 20-to-12V conversion, too, promising 90% efficiency at the 100mA mark (and better for any load above, up to 500mA).  If you can add the EMC filter according to the datasheet (two 10µF capacitors, one 1µF capacitor, and one 100µF inductor, on the input), even better.  It's tiny, too, just 8×12×11mm plus the three legs.

If it is not, an isolated DC-DC converter like Mean Well SPBW03F-12 that Mouser sells for under 11€, ought to work well.  It is slightly less efficient, but not too bad.  The input EMC filter is just a 18µF inductor and a 6.8µF ceramic capacitor.  This one is 22×10×12mm, with four of the eight legs NC (not connected, but cut and sleeve the stubs in heatshrink).  Because the minimum load is 11mA, you should add a 1kOhm resistor between the outputs, unless the fan always turns when there is power.  The resistor does waste 144mW of power into heat, so better select a quarter-watt or better rated resistor, and put it somewhere the little bit of heat is easy to dissipate.

Because the load is inductive, I'd also put a flyback diode in reverse on the output in both cases, i.e. cathode to positive output, anode to negative output.  For the same reason, the 0.05 to 0.1 Vpp noise on the output of the DC-DC converter shouldn't be a problem at all.

I'm only a hobbyist myself on the electronics side, so if any members here point out anything I missed here, I'd definitely listen.  I do try to carefully read the datasheets, and I did do some electronics courses at Uni, but I do not have the practical experience many of the members here do.
I do like to use digital isolators and isolated dc-dc converters in my own projects, mostly dealing with microcontrollers (Teensies mostly!), but stuff like EMC/EMI is way outside my skill set.  I don't even have suitable equipment to test (not even a proper oscilloscope, only an AD2), so I just follow the datasheet recommendations, prefer conductive cases for my projects, and hope for the best... 

[incognito]

Is there any solution with similar output, but that can take ~20VDC input? If I recall right the monitor's DC adapter output is 19V or close by. That would allow me to solder the transformer's input lines onto the board's main power.

Is the input ground directly connected to the fan ground?

How can I know this? I connected multimeter negative end to the fan ground and searched the board for a 12V rail. Didn't found one. Instead I found plenty of ~19,5V rails, which is exactly what the main power adapter outputs. Also found several 3V rails and a couple of 10,2V. I think I might've tested the main power connector's legs and the ground from those showed zero or practically near between the fan ground. I'll have to double check that, if that's how to make sure what you were asking for.

But today I got to the point of testing in practise what I started the thread for and it looks like it works as planned  . You can read my more lengthy update with videos here:
https://forums.overclockers.co.uk/threads/lg-38gl950g-3840x1600-g-sync-144hz.18840788/post-36440001

[Nominal Animal]

Is there any solution with similar output, but that can take ~20VDC input? If I recall right the monitor's DC adapter output is 19V or close by. That would allow me to solder the transformer's input lines onto the board's main power.

Is the input ground directly connected to the fan ground?

How can I know this?

When powered off, check the continuity or resistance between the fan ground and the negative/ground side of the 19V/20V rail you intend to use.
If you have a low-resistance connection between all the negative/ground sides you intend to use (on the TV board), then you don't need isolated DC/DC conversion.

https://forums.overclockers.co.uk/threads/lg-38gl950g-3840x1600-g-sync-144hz.18840788/post-36440001

Some DC/DC converters like non-isolated Cui V78E12-500-SMT-TR have a separate Remote On/Off input.  That particular converter has one that when grounded turns off the converter, and when pulled to 3.2V - 8V (or left floating/unconnected) enables the converter.  Thus, if you were to connect the GND pins (3 and 7) to the ground, VIN+ pins (1 and 2) to the 15-36V input, VOUT+ pins (4 and 5) to fan 12V, and the board +5V supply for the fan to remote (pin 10) with a pull-down resistor to ground (say, 10k), the 12V PWM fan would only be powered whenever the 5V PWM fan would be powered.  (The adjust pin, pin 6, is left floating for no output voltage adjustment.)

Perhaps that would let you use Deep sleep and save power when not in use?

The V78E12-200-SMT-TR does need a tantalum (or low-ESR electrolytic) capacitors: 10µF between VIN+ and GND (rated at least 50V), and 22µF between VOUT+ and GND (rated at least 25V).  The input side EMC seems to be an LC filter; see the datasheet.  I'm hoping more experienced members will help with exactly what kind of filtering is needed/appropriate here.

A similar isolated DC/DC converter is Cui PQQ6W-Q24-S12-S, with input ground pin 1, VIN+ pin 2, control pin 3 (board +5V supply to the fan with a pull-down resistor to ground/pin 1, requires fan ground and pin 1 to have low-resistance connectivity), VOUT+ pin 6, and VOUT- pin 7.  Pin 4 is the gap, and pin 8 is not connected.  Some capacitors and possibly a common-mode choke on the input is needed with this one.

[incognito]

Is there any solution with similar output, but that can take ~20VDC input? If I recall right the monitor's DC adapter output is 19V or close by. That would allow me to solder the transformer's input lines onto the board's main power.

Is the input ground directly connected to the fan ground?

How can I know this?

When powered off, check the continuity or resistance between the fan ground and the negative/ground side of the 19V/20V rail you intend to use.
If you have a low-resistance connection between all the negative/ground sides you intend to use (on the TV board), then you don't need isolated DC/DC conversion.

https://forums.overclockers.co.uk/threads/lg-38gl950g-3840x1600-g-sync-144hz.18840788/post-36440001

Some DC/DC converters like non-isolated Cui V78E12-500-SMT-TR have a separate Remote On/Off input.  That particular converter has one that when grounded turns off the converter, and when pulled to 3.2V - 8V (or left floating/unconnected) enables the converter.  Thus, if you were to connect the GND pins (3 and 7) to the ground, VIN+ pins (1 and 2) to the 15-36V input, VOUT+ pins (4 and 5) to fan 12V, and the board +5V supply for the fan to remote (pin 10) with a pull-down resistor to ground (say, 10k), the 12V PWM fan would only be powered whenever the 5V PWM fan would be powered.  (The adjust pin, pin 6, is left floating for no output voltage adjustment.)

Perhaps that would let you use Deep sleep and save power when not in use?

The V78E12-200-SMT-TR does need a tantalum (or low-ESR electrolytic) capacitors: 10µF between VIN+ and GND (rated at least 50V), and 22µF between VOUT+ and GND (rated at least 25V).  The input side EMC seems to be an LC filter; see the datasheet.  I'm hoping more experienced members will help with exactly what kind of filtering is needed/appropriate here.

A similar isolated DC/DC converter is Cui PQQ6W-Q24-S12-S, with input ground pin 1, VIN+ pin 2, control pin 3 (board +5V supply to the fan with a pull-down resistor to ground/pin 1, requires fan ground and pin 1 to have low-resistance connectivity), VOUT+ pin 6, and VOUT- pin 7.  Pin 4 is the gap, and pin 8 is not connected.  Some capacitors and possibly a common-mode choke on the input is needed with this one.

Excellent suggestions, again. Although the last two paragraphs were a little too advanced for me to yet fully understand.

I'm on vacation and visiting home only every now and then, thus the project having seldom updates. It just came to my mind how about using a 5-12 DC/DC step-up converter from the original fan output? Looking from the pictures the original AFB0705MC-00 5V fan is rated 0,5 A, which would add up to a nominal 2,5W power:
https://photos.google.com/share/AF1QipPKI1k3EwxKTpncRfAWzckV5PMLKG4x1dAh2nDvxdv5bLVpmmDG5c0_fWOnhZ2IrA/photo/AF1QipO9OrxWUDgjt2vLzup03-kYYhp4n8jp5iIwm8Y4?key=NVc5SmVOX2hSdl9kMmJ1Tl9lb0hqeHk1NVJodnZR

Though a table here with a slighly different model suffix (*-00F0R) says 0,3A and 1,1W:
https://www.delta-fan.com/technology/frameless-fans/AFB0705MC-00F0R.html

As NF-A9x14 PWM is rated 1,32W max. do you see this possible, without major risks? How much of a waste power would such a converter introduce?

Another interesting option would be to take power from one of the three USB ports backside with this I just stumbled upon:
https://www.spelektroniikka.fi/p24522-usb-dc-dc-step-up-muunnin-5v-12v-14-35mm-dc-plugi-paassa-en.html

I don't know yet, but I think there's a good chance those ports would be powered monitor turned off but deep sleep disabled (fan PWM control is present) and powered down deep sleep enabled, which would be ideal, stopping the fan when PWM control is cut off in deep sleep.

[Nominal Animal]

As NF-A9x14 PWM is rated 1,32W max. do you see this possible, without major risks?

With a suitable step-up DC/DC converter, sure, if the original fan is rated at 2.5W.

If you have a DMM, you could always measure the current (of the original fan), although it may involve cutting one of the fan wires (because to measure current, the DMM has to be in series with the load).  I don't really see any major risks here, but then again, I too am only a hobbyist.

How much of a waste power would such a converter introduce?

I can't find too many suitable DC/DC step-up converter modules, but Mouser does have MPS mEZD41501A-B, which accepts an input voltage between 4.5V and 10V, and outputs 12V, at maximum 1A output current (datasheet PDF; note that the block diagram on the first page is of the innards of the converter, not how you use it; it actually does not require any external components at all –– except possibly for the reverse flyback diode on the output, since the load is inductive.)

The NF-A9x14 fan is rated at 0.11A (110mA @ 12V = 1.32W), and the mEZD41501A-B efficiency curve states 92%+ efficiency there.  Thus, the input power is 1.32W/0.92 ≃ 1.44W, and the generated waste heat the difference, or 0.12W –– very, very little, since this DC/DC converter is quite efficient.

This DC/DC converter is easier to use than the others I've referred to, because it is intended to be used by DIY'ers and as-is in products, according to the datasheet.  On the other hand, Mouser sells it for 13.70€, so quite a bit more than the alternatives... but I really like this option myself.

Another interesting option would be to take power from one of the three USB ports backside with this I just stumbled upon:
https://www.spelektroniikka.fi/p24522-usb-dc-dc-step-up-muunnin-5v-12v-14-35mm-dc-plugi-paassa-en.html

True, that should work too.

USB is well regulated.  The 5V fan supply can be noisy, because fans generally don't mind a bit noisy supply, so I am not sure if wiring one of those (any USB-to-12V step-up converter) would work when connected to the 5V fan supply.  From an actual USB, they tend to work quite well.

I don't know yet, but I think there's a good chance those ports would be powered monitor turned off but deep sleep disabled (fan PWM control is present) and powered down deep sleep enabled, which would be ideal, stopping the fan when PWM control is cut off in deep sleep.

It ought to be easy to test, with any USB gadget or power bank!

That said, I personally prefer the MPS mEZD41501A-B solution.  You simply connect its input to the fan 5V supply and ground to the fan ground on the board, and output and ground to the new fan.  The flyback diode between the output and ground, as well as the pull-up resistor and voltage divider for the tach signal you can even dead-bug (use through-hole resistors and diode, solder them "in air", but cover with heatshrink tubing).  I personally would make a small board though, probably in EasyEDA, with surface-mount resistors and components (that I'd order from Mouser as well, to fill up the 50€ minimum order), and try and put transparent heatshrink over the entire assembly for good measure.

I think it would be nice to have a standard 12V PWM fan connector on one end, plus a pigtail to the original 5V PWM fan connector, making it a module others with the same or similar problem could use – 5V to 12V PWM fan converter.  If you want, I can try and draw one up based on the MPS mEZD41501A-B, so we could ask the more experienced members here to review and check it and make suggestions if necessary?  (All my designs are in public domain.  All I ask is you pass the help forward.)

[incognito]

I think it would be nice to have a standard 12V PWM fan connector on one end, plus a pigtail to the original 5V PWM fan connector, making it a module others with the same or similar problem could use – 5V to 12V PWM fan converter.  If you want, I can try and draw one up based on the MPS mEZD41501A-B, so we could ask the more experienced members here to review and check it and make suggestions if necessary?  (All my designs are in public domain.  All I ask is you pass the help forward.)

Yes, that's actually a great idea, if you have time in between at some point. 

PS: I'll stay tuned, as it's a few days away anyway until I get back to this project in practise.

[incognito]

While waiting if Nominal Animal has at some time point a chance to design the step up board are these Aliexpress solutions passable or something to avoid:
search results: https://www.aliexpress.com/w/wholesale-USB-step-up-5V%252F12V.html?catId=0&initiative_id=SB_20230614105518&SearchText=USB+step+up+5V%2F12V&spm=a2g0o.home.1000002.0

USB step up  module: https://www.aliexpress.com/item/1005004782722294.html?spm=a2g0o.productlist.main.1.9875193fOy3xEd&algo_pvid=ee5080aa-b3b8-4c4a-a6e8-7453a735cf19&algo_exp_id=ee5080aa-b3b8-4c4a-a6e8-7453a735cf19-0&pdp_npi=3%40dis%21EUR%211.62%211.38%21%21%21%21%21%40210218bf16867689191416261d0743%2112000030468028036%21sea%21FI%214218132387&curPageLogUid=wFEw3UmdToos

And general 5V to 12V module: https://www.aliexpress.com/item/4000197777372.html?spm=a2g0o.productlist.main.3.9875193fOy3xEd&algo_pvid=ee5080aa-b3b8-4c4a-a6e8-7453a735cf19&algo_exp_id=ee5080aa-b3b8-4c4a-a6e8-7453a735cf19-1&pdp_npi=3%40dis%21EUR%211.71%211.49%21%21%21%21%21%40210218bf16867689191416261d0743%2110000001116164371%21sea%21FI%214218132387&curPageLogUid=8j1H9sJx60bV

Does it look like the latter module includes everything possibly required here, like the flyback diode?

[mariush]

Can't believe nobody suggested this already.

Looking at the video in first post, the monitor has external power supply, which is probably 12v or higher. most likely in the 18-19.5v range

It would be MUCH easier to get a step-down regulator to produce 10-12v for the fan from this higher voltage, directly from the barrel jack input.  There's a big diode near the barrel jack connector and there will be some capacitors there rated for 25v or higher - your 19.5v (or thereabouts) voltage will be on the positive of those capacitors, so you could just solder a wire to that side to get your voltage... or on one end of that big diode.

You can get cheap step-down regulators from eBay, for example : https://www.ebay.com/itm/134428816503   = just tape the bottom so it won't cause short circuits and then you can use some double sided adhesive tape to stick it somewhere inside the case.

If the 5v fan only starts sometimes, you could take advantage of that by using a transistor to connect the negative wire of the fan to ground.  Have a wire from the 5v of the original fan go to base pin of a transistor, with a small resistor (10-100 ohm) in series to protect the transistor, then negative wire of fan goes to collector, and emitter of the transistor goes to ground.  Now the fan will start only when there's 5v on the original fan.

Otherwise, you could still control the fan's speed by adjusting the voltage of the step-down regulator.
For example, the step-down regulator uses a 10kOhm potentiometer to adjust the voltage.
The output voltage can be calculated with formula Vout = 1.23 ( 1  + R2/R1) where  one of the resistors is the potentiometer and the other one is a resistor near the potentiometer, probably the resistor with 331 on it, meaning 330 ohm. 

So, let's say you want the fan to work with voltages between 6v and 12v, that would mean  R2/ 330 = 6/1.23-1  = 3.88 => R2 = 3.87 x 330 = 1280 ohm   and R2/330 = 12/1.23 -1 = 8.76 => 2890 ohm so you have a range of 2890 - 1280 ohm = 1610 ohm between these voltages.

But you could do it the other way and instead of a fixed 330 ohm resistor, you could have maybe a 180 ohm + a 220 ohm thermistor (when it's cold it's 220 ohm, the warmer it is the lower the resistance).
So for example, at 25c the thermistor will be 220 ohm so you set the potentiometer on 1800 ohm and you get Vout = 1.23 ( 1 + 1800/400 ohm ) = ~ 6.75v

Example thermistor : https://www.digikey.com/en/products/detail/epcos-tdk-electronics/B57164K0221J000/652169

and you can see in the 1305 column on page 7 of datasheet:



Temp   Rt/R25

25.0  1.0000
30.0  0.8418
35.0  0.7108   --- 156.2 ohm -> vout = 1.23(1+1800/336.2) = 7.8v
40.0  0.60317
45.0  0.51419 --- 112.2ohm => vout =  1.23(1+1800/292.2) = 8.8v
50.0  0.44037
55.0  0.37824 -- 81.4 , vout = 9.7v
60.0  0.32636
65.0  0.28333 -- 61.6 , vout =  10.39v
70.0  0.24697
75.0  0.21573 -  46.2 , vout = 11v
80.0  0.18908
85.0  0.16649 -
90.0  0.14709 - 30.8 , vout = 11.73v

[Nominal Animal]

I think it depends on how stable the 5V on the board is.  Fans do not need a very clean supply voltage at all, so it may not be well regulated, and may be noisy; that can affect the DC-DC converter as well.

There are Youtube reviews on these modules, and although I have some myself, I'd be hesitant on putting one inside an expensive display...
(I do intend to use something like that on an Orthex SmartStore 45 box-to-soldering-fume-hood conversion with activated charcoal HEPA filter and a few old 3-pin 120mm Noctua/Gentle Typhoon fans.  As these are voltage controlled, I can just set up the output voltage to match the desired fan speed.  That is, I do believe these work just fine, I just don't really trust them inside anything expensive.)

That's why I like the mEZD41501A-B solution, even though it is nowadays Not Recommended for New Designs: it has such a wide input range, and is designed for exactly this purpose (DIY, 12VDC regulated output).  The datasheet also specifies its efficiency at low loads (12V, 110mA for the 1.32W fan), which definitely helps; and even shows the actual circuit and components on it.

A lot of existing DC-DC step-up/boost ones have very tight input (4.5V to 5.5V) and/or unregulated output.  Almost all the isolated ones (which obviously work for un-isolated uses as well) I can find in the <10€ class have an unregulated output, requiring a minimum of 10% of nominal load, or their output voltage may even double.  They also tend to be very inefficient at low loads.

(Thus, I took a bit of a detour into TI Power Designer to see how complex a fully DIY step-up from say 4-6V to 12V would be, but my design skills just aren't up to that level yet.  Also, the efficient chips tend to be legless packages, which I'm not confident soldering myself.)

If you already tested the 12V PWM fan with a separate supply, and it works fine, then the PWM and Tach signals do not need level shifting, all you need is mEZD41501A-B and optionally a flyback diode, say 1N4148, with its cathode (-) to the 12V output, and anode to GND.  No board needed, just solder wires directly (and protect with heatshrink); use a leaded diode.  I do believe Noctua (and other quality) fans either have one or equivalent circuitry built-in.

To do the PWM and Tach level shifting properly without inverting them, each needs a couple of resistors or a trimmer (as a voltage divider), a diode (in case the board pulls the Tach to higher VCC than the adapter, or fan the PWM), a couple of resistors, and a pair of signal N-MOSFETs.  Even the dual N-MOSFETS cost only a couple dozen cents each (under 3€ for ten in SOT-323-6, which is small but still hand-solderable; with many having the same pinout, for example DMN63D8LDW-13), so it's really just a matter of having to solder all the components.

(Because I am a hobbyist, I had to simulate the circuit using ngspice – I use Linux.  The PWM frequency is nominally 25 kHz, and at two pulses per rotation, 2000 RPM, the Tach output is 67 pulses/second.  So, these aren't exactly high-frequency signals.)

I need to do some component datasheet checks, and simulate the final schematic in ngspice before it is ready for review, though.

[Nominal Animal]

Attached is the schematic I'm working on; it still needs review, and I myself will try to simulate it in ngspice (using this DMN64D8LDW spice model) to check it works correctly in all situations.  After simulation, I'll check the footprints, create a board, and publish the entire thing.

Simply put, the core is the MPS mEZD41501A-B step-up DC-DC converter.  It requires no external components.
Note that the optional 1N4148 diode does not normally pass any current; it is there only if the fan motor causes an inductive negative spike on the supply.  (I suspect most PWM fans are brushless nowadays, so there is no need for such a flyback diode, but obviously it does no harm either.)

The host provides an open-drain PWM output, marked PWM_IN in the schematic.  Per the standard, this has to be pulled up to 5V or 3.3V in the fan.  R2-R3 is the voltage divider (from 12V), here configured for 5V.  The proper voltage depends on the host, so I am thinking R2-R3 could instead be a 10k multi-turn trimmer potentiometer, that one would adjust when making the board.  Possibly keep pads for 0805/0603 resistors and through holes for a standard multi-turn trimmer?  The diode ensures that even if the host does not provide a true open drain output, and pulls it to a higher voltage, the circuit will work just fine.  The PWM to fan is marked PWM_OUT in the schematic, and is a proper open drain output that can handle being pulled to anywhere between -15V and +15V in the fan.

The fan provides an open-drain Tach output, marked TACH_IN in the schematic.  Per the standard, this has to be pulled to 12V.  The Tach output to the host is proper open-drain one, that can be pulled to anywhere between -15V and +15V by the host.

The 100 ohm resistors at the gates of the MOSFETs are there to filter out VHF noise picked up by the long leads.  I suspect their value needs to be smaller, something like 10 ohm or 22 ohm or so.

Because a single N-MOSFET inverts the signal, we cascade two of them.  I chose this over a more common BJT transistor, to keep continuous current draw low, and because dual N-MOSFETs are cheap and plentiful in tiny SOT-323-6 packages, while still being hand-solderable.  This particular one is available at LSCS, JLCPCB, Mouser, and others; including JLCPCB assembly service, if you don't want to solder these yourself.

[Nominal Animal]

Looking at the attached EasyEDA 3D renders of the board, I think I'll replace the 3-pin SOT-23-3 diodes (dual common cathode, NXP BAT54C,215) with SOD-123; and the 6-pin dual N-MOSFET in SOT-363 with 3-pin N-MOSFETs in SOT-23-3, for easier soldering.  The board itself is only about an inch square.

Simulations indicate this should work pretty well.

R1 and R2 on the backside are only used if the blue multi-turn pot is not used.  For 3.3V PWM pull-up, R1=2.75k and R2=7.25k; for 5V, R1=4.2k, R2=5.9k.
The idea is to connect 12V and GND, pot, and the diode, and then adjust the pot until PWM_IN (blue input wire) reads 3.3V (or 5V) or a bit less.

[Nominal Animal]

And here is the schematic and board files, in Public Domain.

Both the schematic and the board design needs some critical review!
Any and all comments and suggestions and fixes are welcome and appreciated.  The circuits themselves are trivial, just open drain level shifting without inversion.  The important magic, boosting 4 VDC to 9VDC to 12 VDC at high efficiency, is done by the mEZD41501A-B module.







R1 and R2 are alternates in case one does not want to use the trimmer marked R3.  My reasoning for the design choices:

• D1 is an optional freewheeling diode.  I suspect most standard 12V PWM fans are brushless nowadays, and don't need one, or have the equivalent built-in to their control circuits.
• D2 ensures that even if the host were to use a nonstandard circuit to drive the PWM signal, exceeding the pull-up, nothing bad happens.
• 100 Ohm resistors at the gates of the MOSFETs have a dual purpose: one is to filter out high-frequency noise picked up by the long wires, and the other is to limit the current to the gate (when switching), as the PWM and Tach signals are not expected to handle more than 5mA of current.
• The pull-ups and voltage dividers are all 10 kOhm, for both simplicity and because it leads to reasonable currents (0.33mA to 1.2mA, depending on pull-up voltage).
• In the silkscreen, the 100 ohm resistors have a continuous surrounding rectangle, and the 10 kOhm ones have a broken outline.  The board is small, and I didn't want to use extra space for just the designators.
• Board is offset like the MPS mEZD41501A-B module is.  The idea is that you can carefully trim the module pins, and just stack the two boards.
• All components except the 6-pin mEZD41501A-B DC-DC step-up converter were chosen to have JLCPCB assembly counterparts (either same, or sufficiently similar).  The JLCPCB assembly should be able to handle all the SMD parts, but I did try to make the entire design hand-solderable using just a soldering iron (or hot air) with a fine or conical tip.
• The NX138AKR N-MOSFETs may seem like overkill for this, but note that its gate can be ±20V with respect to source (always ground here); plus the gate-source threshold voltage (Vgs) and on-state resistance (Rdson) are so low, it works reliably even if the gate never exceeds 1.5V.  In other words, this choice makes this robust: even if the fan or host do Funny Stuff with the PWM or Tach signals, it is Okay.

The basic PWM fan control used nowadays is standardized based on Intel 4-wire PWM Fan Specs (Intel PDF), simply because there are so many PWM-controlled fans in computers –– not just desktop ones, but in servers also.