Even Deeper into Rabbit Hole - 4 Pin PWM with External Power Precautions

As from my previous post, I discussed the dangers of a wrong PWM connector used on D5 pumps. Let's dig a bit deeper shall we?

For a refreshment, here's the papers (again) that I'll keep mentioning throughout this post:

- Noctua PWM Specifications White Paper:
https://noctua.at/media/wysiwyg/Noctua_PWM_specifications_white_paper.pdf

- Intel 4-Wire Pulse Width Modulation (PWM) Controlled Fans:
https://docplayer.net/23812689-4-wire-pulse-width-modulation-pwm-controlled-fans.html

What can possibly go wrong?

Now, as far as I can remember, not only pumps that were externally powered from molex or sata, but also a number of PWM fan hubs.

"Wait a sec, are you trying to say that fan hubs are also unsafe?!"

They are generally safe, but not without any cautions. Basically there are 2 types of hubs that can be plugged into MB's PWM fan header:

• Simple fan splitter cables are safe, but you also should understand the current capability for your motherboard's fan header to avoid damage.
• Powered fan hubs are also safe, because you're not taking power from the MB. But you also has to be more careful when plugging the input to the motherboard.

For now we will be focused down to the powered hubs. They usually took power externally by plugging in to molex or sata connector. And most of them works by simply splitting the PWM/Control signal from the MB, take a TACH/Sense reading from one of the fans to the MB, and give power from the external connector. They usually have only 2 wires on 4 pin PWM connector to the MB. Some of them also features a big electrolytic capacitor to smooth out the power to the fans.

The implementation for splitting the PWM signal this way is already different from the specification, intel suggests adding buffers if intended to drive multiple fans with a single PWM output (Intel specification, page 11: "Considerations for Motherboard Designers"). As the intel specification, PWM control output from the MB is an open-drain or open-collector, and the pull-up is located in the fan. Noctua, on the other hand, stated that there are no special requirement for driving the PWM signal, they shows that either push-pull and open-collector outputs are fine (Noctua PWM specifications, page 8: "Driving circuit"). So I guess it's okay to say that although different, this splitting approach is safe enough. As long as the pull-up circuits are identical between all connected fans (in another words, using the same fans for a single hub).

My trusty homemade portable PWM fan tester,

driving the PWM signal straight from the MCU output pin.

Moving on.

Now let's bring up the connector precaution again. What will happens if the 4 pin connector is plugged wrongly onto the MB header? Let's refer to my previous post again for the diagram and schematic of plugging the connector on an offset. This time, the current will flow from the +12v pin from the MB header, into the TACH input on the fan hub, out to the TACH input on the connected fan (usually the first header or the different colored one on the hub), into the transistor/mosfet inside the fan, and out to the ground via external supply (molex/sata). Poof goes the transistor.

Wrong current path

Granted, this have a high possibility only when the manufacturer is using a wrong part for the 4 pin PWM connector like in the case of my D5 pump, and I've never see a hub with wrong connector, yet. But! It doesn't mean that it can't happen on a proper connector as well.

Whomp!

Alright, alright, this is obviously a worst case scenario. You had to jam it pretty hard to get into this position. That plastic is pretty hard to bend and surely you'll feel it if you plugged it wrongly. I can only imagine this situation might happen if someone (presumably not really experienced in building a computer) is in a hurry.

Then again, this is all depending on how the fan hub manufacturers implement the driving circuit. Most of the time, especially with cheaper ones, they tend to use this direct PWM signal splitting method. There are few exceptions though, like Phanteks' Universal Fan Controller that can take both PWM and DC signal from the MB to control the fan speed, in which obviously they have some active circuitry to split the control signals.

Summary

So the conclusion is always pay attention to the fan connector. Not only for the pumps, but for anything that needs to be plugged into PWM fan header and be externally powered. And as I said on my previous post, do be aware with devices with improper connector.

On the very least if something goes wrong, there will be just a puff of magic smoke released from the transistor/mosfet, and your RPM reading will be gone. But at the worst case, it could be a fire hazard (if the transistor fails by keep shorting the collector and emitter, and if the motherboard didn't have a current limiting capability on the fan headers).

Down The Rabbit Hole - D5 PWM Pump Issue and Possible No RPM Fix

Background story:

I was completing my first time custom watercooling rig, it was a full EKWB set. EK Velocity DRGB cpu block, EK-XRES 140 Revo D5 RGB PWM pump res combo, EK Vardar ER fans, and EK CoolStream PE 240 radiator. Simple loop, with soft tubing. The reason why I choose D5, just because I may expand this loop in the future.

After the setup was complete, completed the leak test and everything else, I hooked it to my system. Everything works perfectly. PWM control and RPM readings are all good on both the Vardar fans and D5 pump.

Then I decided to tidying up the cables a little bit, and reconnecting all the fans/pump PWM to the motherboard. One thing I noticed, that suddenly my D5 pump run at full speed, and the RPM reading was gone.

0 RPM reading

The Problem:

After a closer inspection I saw that my PWM/tach connector from my D5 pump was connected to the motherboard in an offset position. I quickly reconnect it to the proper position, but alas, the RPM/tach reading was already dead. The PWM control still works though, but I have no RPM reading from the D5 pump anymore.

I quickly googled it, and found out that many people also reported that the RPM reading from their D5 pumps (through various brands) were gone. But I found no good answer and/or fixes for this issue. Furthermore, many of the results were from quite old forum posts.

I also asked my local EK distributor about this issue, they tested their pump, and also have no RPM readings. Which then got me thinking, "is this quite a widespread issue?"

Then I begin to diagnose it myself. At first, I noticed that the 4 pin PWM connector from the D5 is not the same as a common PWM fans. On normal PWM fans, the 'clip guide' / 'notch' or whatever it called (edit: it's called "Polarizing rib"), it is positioned on between the 3rd and 4th pin, while in D5 connector, the rib is all the way over the connector's width. This shows that D5 is using the more common 4-pin small molex connector, instead of the more standardized version which have that rib in the middle (like Molex 47054-1000 part number which mentioned on Noctua PWM Specification White Paper).

Left: Proper PWM fan connector | Right: D5 PWM connector

This connector, even though works, will have a chance to be connected on an offset like I already did. Now here's the catch, if the pump is powered by MB fan header, it will probably just fine (it just won't turn on). But on D5's case, when the pump is powered by external molex connector, it is just get messy.

Proper connector, minimal risk of plugging offset

Improper connector, it is possible to shift/offset the connector like this

As per Intel's standard for PWM (also mentioned on the Noctua White Paper), the RPM/tach output (pin 3) is an open collector (or open drain for a mosfet), it needs to be driven with a pull-up resistor. However, if by accident the connector is on an offset, this tach output will be connected to a low-impedance, 12v voltage source, to the transistor, straight to the ground via external molex power. Although I didn't smell any magic smoke escape.

Improper open collector driving, creating a low impedance path, thus a large amount of current is dumped into the transistor

Proper driving procedure of tach out signal by Noctua.

Noctua even have current limit to tach output on their PWM white paper

The Fix:

After searching around for posts where people disassembled their D5 pumps, I took a closer look at where the RPM cable goes. And sure does, it goes straight into a tiny SMD transistor inside the D5. Then I bought some generic NPN SMD transistors (picked up MMBT3904LT1 from my local electronic store, but any NPN signal transistor should work I guess, since RPM is not a high speed signal), cracked open my D5, and start probing around that transistor.

The suspecting transistor (red circle), and the RPM/tach signal cable besides it (blue cable)

(Credit to someone who shared this internal pic of his D5 pump, sorry I've been unable to track back the source, I think it was from overclock forum. Let me know for the credit! I totally forgot to take a pic of my D5 internals when I fix it, but it's pretty similar.)
(Also, yes, sorry for the confusion, D5 have this color flipped between PWM and tach cable.)

I don't have any schematic for D5, and I don't know how Laing/Xylem implement their RPM signal circuit, but after probing out, I was certain that the RPM cable runs straight into the collector of that transistor. Probing further, I found out that the base and emitter of that transistor was shorted, even though there is a 100k resistor between that pins.

MMBT3904LT1

So then I pulled out that smd transistor and replaces it with the new one. Reassemble it, put a short tubing, fill it with distilled water, plug it in to my RPM/PWM tester, and voila, the RPM signal is back! I quickly drain it and reassemble back to my system, fill it with EK cryofuel, let it fill, turn my system on, and there you go, now I have my RPM signal back!

RPM is back on, hooray!

Summary:

It is quite possible that I fried that transistor by plugging the connector on an offset. That small transistor is only capable of sinking several hundred milliamps of peak current, while modern MB fan header can source current for around 2A or more.

I don't know if it's a brand-specific or not, but EK might have a habit of putting wrong connectors to their products. My local EK distributor told me many times to be careful when plugging that Velocity block RGB header, as EK use 4 pin female connector while in fact it uses 3 pin addressable RGB. Well, I think the same goes for their D5 connectors, although I can't really blame EK for this, since I saw some D5 from the other brands have this same connector.

As for the warranty, well, I'm sure I have voided my own D5 because I've been disassembled it. But I'm all in with self-repair movement! 😁 Anyway, if you happen to blow the tach/RPM circuit by plugging in the wrong way like this, I'm pretty sure you should be covered by warranty. Especially if they used incorrect connector like this.

However, I think this is still a rather serious issue. If the MB's fan header don't have any current limiting capabilities, it is also possible that not only you'll fry your pump's tach circuit, but also your MB's fan circuit.

Edit:
After digging deeper into image search, I saw some EK D5 pumps that already have a proper PWM connector. I don't know if this is a batch-related or not, but takeaway on this issue is do plug your D5's PWM connector very carefully, especially if your pump comes with this kind of PWM connector. Also if you want to be safer for future uses, just replace the connector with a proper one (steal it from PWM fan or something 😁).

Links:

- Noctua PWM Specifications White Paper:
https://noctua.at/media/wysiwyg/Noctua_PWM_specifications_white_paper.pdf

- Intel 4-Wire Pulse Width Modulation (PWM) Controlled Fans:
https://docplayer.net/23812689-4-wire-pulse-width-modulation-pwm-controlled-fans.html