This is a virtual sensor I came up with to allow me to run two or more curves on the same profile for a fan or pump header. Within the sensor I can configure triggers so that the fan/pump will switch curves when system conditions are met.



As im not as versed in posting to this forum here is a reddit post linked with the guide.
https://www.reddit.com/r/watercooling/co…s_with_virtual/

CPU/GPU load virtual sensors for more data driven load values.

CPU

GPU

https://pastebin.com/ASTEEz3S
Pump controller V4.1 Virtual Sensor XML code for importing into auquasuite

It's constantly evolving and it has changed already since typing this... mostly just playing with filters lol.
Enjoy ! Hopefully some of you can find a use or share with me what you would change/improve.

It was quite fun. And funny enough I did make a Rube Goldberg machine years ago.

The main reason besides experimentation was to develop a way of increasing the pump speed when it’s needed most and only then. Increasing flow rate will move more heat from components to the radiator. If the fans are maxed the only thing you can do is speed up the pump or lower ambient temp to gain additional cooling. In the case of fans it can be used to shut them off completely when delta T is a value you set for silent operation until load or temp increases. I currently don’t have a flow meter installed but am building a dual pump parallel x flow rad setup that will utilize two flow meter. Both before the manifold to ensure I can tune equal flowrates through the rads, should I decide to split the manifold through waterblocks of different resistance. Something like this allows me to take full advantage of the higher delta T of parallel rads while also minimizing restriction for all the QDC’s. It’s a custom case I’m building in solidworks for minimal footprint and ease of maintenance. It needed an overkill virtual sensor to go alongside.

Creating this sensor also seemed less daunting then setting up two or more profiles to switch back and forth on conditions. But I certainly could use this to drive outputs and switch profiles and achieve the same result. It is much easier to tune it all being integrated into a single virtual sensor, less confusing as well.

But there is a temperature change as a result of increased flow...

And that is the point of this. Not what you should or shouldn't do. Or that its so small it shouldn't matter. This is meant for those who want to squeeze every bit of cooling potential out of their loop, but only when they need it.
If that is your takeaway, you are missing the other configurations this sensor could be used for.
And if you just want a static pump and fan running off Delta T that's fine too.

Its not enough to say flowrate doesn't make a difference because even if that change is 1C or 5C its still there... If you have a target Delta, that 1C might mean something to someone.
Nor is it accurate to say their isn't a hot and cold side off the loop or that "water temps equalize" because my temps sensors all over the loop don't back that up. There most definitely is a hotter and a colder side, sure it might be under 1C most of the time, but if you have fans and a radiator the difference is there.

Too many people just repeat what they hear. Most people don't know that mixed metal loops are totally fine with todays coolants so long as you don't provide a direct metal to metal electrical connection of two metals with different electrical potential or bridge them with a conductor in addition to the coolant.
Its cheaper for manufacturers to say no aluminum than take the risk of someone screwing in a brass fitting to an aluminum rad or having copper touching a conductive surface with aluminum touching the same surface 20" away while both are immersed in a conductive coolant.
Brass fittings are installed in aluminum all the time, like a cars engine, however its typically a very small amount of the noble metal and a very large amount of the reactive metal so the electrical potential isn't great, and corrosion of the anode is very slow.

Rant over lol

This sensor was made for those who want maximin performance and are okay with the increased pump noise that comes with the curve switch.
They're not large values as we are far far from taking advantage of waters thermal properties with the temps we work with, but this is about performance optimization with a given system. Pushing what a homebuilt loop is capable of.


In the build I mentioned before, I am taking steps to optimize coolant paths, using larger tubing and fittings to minimize frictional losses before they split, which halves the flowrate, then flowrate is further divided by the number of coolant paths in the radiator. All tubing lengths for parallel runs will be equal.

If I have a flow of 220lph, that then splits and becomes (65lph split again lol) 110lph entering the rads, then 6.875lph passing through the rads before becoming 110lph again in the end-tank, and 220lph again when all combine in the manifold. (or stay at 110lph if I divide the manifold)
Flow meters and ball valves will allow me to equalize the parallel flow should I split the manifold. There will be water blocks on almost everything and about 8 QDC's, hence dual pumps and the virtual sensor to boost the pump when system conditions dictate its needed.

Its overkill AF and somehow I'm cramming it all in to the volume of a Mini ATX Case with movable maintenance features while keeping an airframe inspiration. Once the coolant is routed aluminum cover panels will be made to look like WW2 fighter aircraft skin complete with aircraft rivets.




2x 420mm rads and 1x 92mm rad, with option for external rad once done
Routing the coolant pipes and planning all the 3D printed cable guides to keep features is proving challenging lol... but I will make it work.
If machined or welded parts need to be made, that is no problem.

If that sounds crazy I asked chat GTP to write me a script to actuate USB relays to power solenoids valves, so I could have an aquasuite output trigger the process and divert coolant to an external rad only when needed lol. Obviously its kinda pointless when I could just switch profiles to turn on external radiator fans, but cool nonetheless.

Sorry that was confusing wording. Flow rate isn't reduced. A flow of 220lph split through two tubes will still have a total flow off 220lph, but the flow through each of the two tubes after the split will be 110lph. And yes the velocity does decrease with larger less restrictive tubing compared to the same flow through smaller tubing.


But how does increasing flow rate not contribute to additional cooling?

Technically there is a Delta T for the waterblocks-coolant and a Delta T for the coolant-ambient air.


Q=mcDT
The rate at which heat is transferred is directly related to the mass flow rate, by increasing flow you increase heat transfer. Flow rate is directly related to the total amount of heat transfer occurring. Its non proportional, as Delta T is decreasing, doubling flow isn't gonna double Q.

At a constant CPU load there is a constant amount of heat energy being dumped in to the loop (Q)
So by increasing flow, Delta T does decrease in regards to the waterblock-coolant to maintain that constant (Q). Which is good! Cooler water in the block can absorb more heat energy and has a higher flow to move that additional absorbed heat.
Its confusing because that formula represents the mass flowrate, given that we have a cap to the total liquid volume of coolant in the system, higher flowrate = faster mass of water circulating.


In regards to the radiator.

Q=mcDT

We are decreasing delta T from the moment water enters to when it exits in order to cool it, so yes the higher this number, the more efficient our rad can be.
But with a constant flow, the fact Delta T is decreasing does not mean flow is increasing to maintain Q or that the specific heat of water is changing significantly enough to matter. In this case Q is changing as the amount energy is leaving the system is related to how fast the mass of coolant is moving and how fast delta T is decreasing. Which is where fans come in. (all of this is assuming a constant airflow)

So, by increasing the mass flow of water though the radiator, we are physically moving more heat through the rad, meaning more energy is removed compared to a lower flowrate.
It by no means cancels each other out. Even if delta T is .0001C, the more we increase flow, the more heat energy can be transferred up until our pump maxes out.


Among a myriad of other factors affecting water cooling that aren't accounted for.

• Increased pump speed means more turbulence which leads to better absorption and shedding of heat as opposed to laminar flow which has a shield of water on the edges of the coolant path blocking most of the water from heat. This is a big factor in heat transfer and good blocks are designed to induce turbulence, but higher flow usually means more turbulence.
• Increased pump speeds or bigger pumps will add more of their own heat energy to the loop (maybe 5-20W depending on size and speed)
• Technically as the fluid heats up its viscosity decreases leading to better flow characteristics.
• What speed your fans are running, and how much they vary with changes in Delta T.
• Specific heat capacity of the waterblocks themselves
• Restrictive components

Overall it isn't a lot in our systems, because the amount of heat energy and temp deltas aren't large numbers.
It is also completely dependent on the individual setup and ambient temperature, you might have .5C temp changes you write off as insignificant, but someone else might see 5C+ even with a good flowrate to begin with.




In my case, when the fans are running at 100% with a high system load, and the pump speed increases up to 100%, the rate at which Delta T is increasing is dramatically slowed as the pump speeds up. I know I currently need more radiator SA, but increasing flow is definitely increasing the amount of heat removed from the loop.


Actually typing this is making me wonder if I should be running my pump controller off the first Delta T, under stress test load that Delta goes almost immediately to 35C while the Radiator Delta T climbs about 2C over a slightly longer period as the water absorbs thermal energy. Typically is sits ~.5C-2C at idle compared to Radiator Delta T.
By bumping pump speed, the rate at which that 2C change occurs will increase. While that might mean lower fan speeds initially, they are much quieter so I prefer pump speeds increasing secondary to fans when noise is factored in. But it might be worth a test...




But anyway this is my understanding of how it works. Please let me know if that is incorrect and why.

Zitat von »PvF«

What you need is the Formular for the "heat flow" = "Power".
This Formular is similar to Q=mcDT but with a dot above the "Q" and the "m" to indicate this is per "time" - e.g. per second (1J/s = 1W)

Sorry that was my error in typing.It’s been a while since I took a physics or chemistry class lol, It’s the formula for cooling capacity.
But yes it is supposed to be
heat = mass flow * specific heat capacity * temperature differencekW = (kg/s)x(kJ/kg/°C)x( °C)
As you said we can convert units as needed just so they are ones we are more familiar with for use with watercooled computer systems.

W=(lph) x (4.18KJ/Kg K@35C) x (DT) I use an aquasuite sensor that someone else created to calculate my cooling capacity with a virtual sensor.https://www.reddit.com/r/watercooling/s/Cp1UYwU24D

Zitat von »Speedy-VI«

I am by no means an expert on this, but I typed this formula into ChatGPT and got back the following.

Interesting, I myself have used chat GPT to try to understand flow/pressure/volume and their relationships in coolant pathways
I believe the diminishing returns it is referring too, is due mostly to pump power, and oversizing your pump. Too powerful a pump will dump excess unwanted heat that takes more energy to cool. Ideally we don’t want the pump adding significantly to what it’s working to counter so I agree there is a balance. I don’t believe the pumps we are running are capable of tipping that balance however. Maybe some of the top line eheim pumps but even that might be a stretch.



I suppose aquasuite has a great logging feature
Time to do some tests and log.

• Delta T waterblock-coolant
• Delta T coolant-ambient
• System Wattage
• Pump speed
• Flow rate
• Fans set at a constant 100% so my radiator will be performing at its max given ambient temp

Since its so heavily dependent on your radiator surface area and fan's CFM(CMM).


What Delta T are you getting at your radiator?
edit* stupid question lol


What Delta T are you getting at system Idle? and under system Load?
Since the temp plays such a large factor for efficiency i'm curious what number you are seeing.

Zitat von »Remayz«

only the D5 is watercooled and still, it dumps such minute amounts of heat in the water that you can basically disregard it.
The temperature sensors are so imprecise anyway you'll probably have bigger measurement errors than pump heat "offsets".
If you use DDC pumps for space savings, they are air cooled so it won't trouble the measures during your tests.

What ChatGPT told you about flow rates influencing heat transfer is correct but in our case with computer watercooling, the energies involved are so low it makes virtually no difference as you saw.

I use D5's so good to know.

I also use 4 temp sensors and will be upping that to 8. I'm less worried about reporting inaccuracies with that many.

I also found that the Koolance ones seem to be the most accurate as all of their probes, including the inline ones, have temp probes protruding into the coolant path, as opposed to just buried in the wall.

Zitat von »Taubenhaucher«

My water temperature under load is ~2.5 °C above room temperature.
The only significant changes I see with this type of load are when the room temperature changes. Then the water temperature fluctuates.
However, when I increase the 3 fans of the upper 420 radiator from 490 to 740 RPM, the CPU temperature drops from 37.75 °C to 37.38 °C.

Yes because the formula Q=mCpDT, air flow is treated like a fluid and has a mass flow rate like the coolant, the only difference is the specific heat capacity of air is like 1 and water is around 4x that. Water is also much denser, and non compressible.
Which, like flow rate for your pump, is why the faster you move air at a temp delta the more energy is passed on. A larger volume of air would have to move significantly faster than water to carry the same energy at the same delta T. Which makes sense, and is why Radiators have huge SA and waterblocks just have a fraction of the SA. The mass of each is relatively similar provided they are the same material like copper, just have different densities.
But that's also why I don't think there is a benefit to increasing pump speed without first having to increase your fan speed, and any testing I will do is with the fans at 100%. Its also how my controller is set up, the pump speed doesn't increase until the delta T is high enough that the fans are close to their max or at their max.
It's probably also why i'm seeing such different results. My radiator SA is much smaller it seems than most people, so i'm working with larger temp Deltas. That large delta, means that increasing flowrate is quite noticeable. For example i'm working with a Delta of 5C at idle and more than double that under any significant load. Time and future tests on the new system with like 3x the rad space will tell, and ill make the necessary config changes to the controller.

Zitat von »Remayz«

The problem is those 8 are probably not in the same places in the loop, so you never know which ones read correct :p
NTC thermistors are notoriously inacurrate, so you'll have to cheat a bit with offsets in Aquasuite, and even then, the response will not always match between sensors. for a same temperature change, some will be reading higher, some will be lower.

Normally yes, and in my case 4 will be random, but because the other 4 will be responsible for inlet and outlet temps on a unconventional parallel radiator setup, essentially they are symmetrical. The temps they see should be the same, or enough that an average of each inlet and outlet sensor should be close to actual temps unless both sensors run hotter or colder.
But that's good to know, i'll have to pick up a thermocouple for my multimeter so I can calibrate, label, and adjust sensor offsets in AS.

Zitat von »Speedy-VI«

Are these the models you are referring to? Koolance does have spec sheets for these thermistors that list max temp but they do not accuracy. The thermocouple specs list a temp range and an accuracy of +/-1.5°C which is not that great.

Yes those are the ones I was referenceing, I dont have any on hand and can't remember if i ever viewed the spec sheet myself. At the time i was scrounging for CAD models so i wouldn't have to model everything myself. The rads and GPU block were fun enough lol. But when i opened the .step files and saw the probes inside the coolant path i was sold. If they aren't accurate, they will at least be more responsive, as the fitting plays less of a role.