I think the consensus is that the "water half" is not actually going down. What you are seeing is error in the system.Quote:
Originally Posted by Erasmus354
Printable View
I think the consensus is that the "water half" is not actually going down. What you are seeing is error in the system.Quote:
Originally Posted by Erasmus354
There is all kinds of stuff going on here.. Skinnee nailed one by mentioning that the radiator would have to be 100% efficient, which isnt going to happen. Also mentioned above was using average temperatures, which actually changes the problem around. If you get super technical and were able to take temperature readings from one side of the tube to the other (like the diameter of a circle) you would find that you have a temperature distribution. The temperature of the fluid right next to the wall of the tube is going to be different from the temperature at the center line of the flow. Though the temperature difference might be small, when you start crunching out numbers it will make a difference. Also mentioned above was 'error'.. There is an actual system for calculating the error of experiments; Kline-McClintock. That system goes through all the partial derivatives, resolution error, deviation, and mean to come up with an error. It is really tedious, but will give you a range of where your final result lies. BUT, most modern stuff that we use tells you what the error is (+/- .05% of the reading or whatever). And, then there is the almighty calibration - which I didn't really see mentioned. Calibrating thermocouples is easy..stick it in boiling water and it should read 100C, then make an ice-bath, stick it in that and it should read 0C. When calibrating our thermocouples we just put them in boiling water and found out how much off each one was...which was usually within the error given in the thermocouples description. All equipment should be calibrated before every test..pumps wear out, thermocouples degrade, etc etc.
So then more flow rate is still always better (or at least not worse). However I haven't seen that consensus that the coolingmasters results were erroneous.Quote:
Originally Posted by Bond Number
They were not erroneous, the test was a great test. They just reach the limits of their equipment. Basically it would be difficult and expensive to perform a similar test with less error.Quote:
Originally Posted by Erasmus354
Interesting post.
I was first wonderin if I could get a better cooling performance (in terms of lower CPU temperature) by slowing down the water flow to the radiator using a res. Looks like I will get a disappointing result :D
That is not necessarily true. Again, it is going to be based on the balance equation (basically, your first law of thermodynamics)Quote:
Originally Posted by ci2kla
h*A*(T2-T1) = mdot*Cp*(T2-T1)
h is your convective heat transfer coefficient, which is a function of how fast you the fluid is moving (in the radiator's case the fluid is air). So, if you increase the fan speed your h will go up. A (surface area) will always be the same unless you change radiators or it gets clogged with a lot of dust etc.
mdot is the water mass flow rate, Cp is constant.
So, if you slow down mdot and increase h more than the balance requires the water temperature change has to increase.
The thing with more than 100% efficiency happened to me too. In my case, it is caused by two factors as far as I can tell: Sensor placement and calibration.Quote:
Originally Posted by skinnee
The radiator that I get the weird results with was, let's say a non-standard radiator where I had to change the placement of the sensors. All th sensors are calibrated to match each other in their usual placement, but when I change their positions, the values from the air out sensors get skewed. With the fans blowing horizontally, I can't seem to find a way to get accurate air out measurements... :(
This is all using average water-in and water-out temperatures, by the way.
It probably has no relevance to what you're seeing, Skinnee, but I thought I'd just throw it out there anyway. May be helpful to someone or maybe someone can enlighten me as to what's going on.
That's basically the point to understand and that's why I said that another method is necessary to investigate special zones if necessary, a cross-comparison (two good tests are better than one). The deltaT imposed method can't be very good for each possibility in a very large range (water and air flows), there's always some drawbacks we can't solve. Fortunately, the maximum error is kept in zones where almost no one goes, so it's not a problem in my opinion.Quote:
Originally Posted by Bond Number
Relttem gave another possible source of unavoidable error I didn't want to introduce when we're at high flow rate, but the water temperature in the tube section is not exactly the same everywhere, even if there's insulation and a lot of turbulence at the radiator outlet to mix water near the probe. Every 0.01 °C counts there and even 0.001 °C probes won't solve this kind of problem. Never forget that error is fully part of measurement.
You all should think about shrouding your radiator inlets and outlets (outlet being more important). When the air leaves the backside of the radiator it is going from a section of high velocity to almost no velocity - this will cause eddies and vorticies to occur which will suck in air temperature from the surroundings, which will cause the temperature reading to be inaccurate. A shroud will protect the temperatures from the surrounding air
That's a good idea with the shrouding, relttem. I'd never thought of that.
I think I'll try to build some type of shroud for measuring air out and see how it affects results.
you can build it with some poster-board and tape. I would use the 20 diameter rule - meaning the length of the shroud should be 20x the diameter of your radiator. You'll have to use the hydraulic diameter to get a measurement of the radiator diameter, since it is not a circle. Wikipedia will tell you how to measure that. It is easy. Then, from fluid dynamics, fully developed flow occurs 10 diameters away. So, that is where I would put my thermocouple (as close to the middle of the duct as possible). At that point all mixing will be done and the temperature there will be a good estimate of the average temperature coming out of the backside of the radiator.
Wow, that's very specific stuff. I would have just gone ahead, and built some type of box and stuck my thermal probes in there. ^^
I'll look into those measurements. One question, though: Is the shroud allowed to be longer than 20x diameter? Because ideally, I would make a shroud that's long enough for the radiator with the largest diameter and use it for all the radiaotors. At least, for all the ones of the same type (i.e. 360s or 480s or whatever).
Actually, now that I think about it: What diameter are we talking about exactly? The total diameter of all the internal tubes of the rad? Or the diameter of the whole radiator? And if so, along what axis?
I'm sorry if I'm asking stupid questions here, but this is all new to me.
I just pulled 20x out of the air. I would make it longer than 10x - 11x would work. It can be as long as you want it as long as it is longer that 10x. I would put the thermocouple at least 10 diameters away. The reason behind this, other than fully developed flow, is that you have different temperatures at different locations coming out the back of the radiator, so you want to get a good average. You can test that to by taking readings at different positions. If you want to get really specific, I would make the duct, get a small diameter (very small) metal rod or wood dowel, put a hole in the duct for the rod, attach my TC to that and position it in the middle of the duct. You can get a compression fitting at the rod's diameters that would let you lock it down in the desired position.
Ok. Well, I'll have to see what size shroud is practical/possible to make, too.
And the problem with different temperatures at different distances from the radiator is very crucial, I think. I'm pretty sure that that's whats messing up my results so far. Especially since, without a shroud, at low fan speeds the warm air will rise up and away from the thermal probes much sooner than with higher airflow.
The shroud will mean only air that has passed thru the radiator is being measured. You can try insulating your duct, but I don't think that will matter much unless your room/ambient temperature is really really cold.
The problem with delta T testing I see is how long it can take to stabilize the run at high deltas and high heat loads with large volumes. I'm seeing some tests taking over 90 minutes to stabilize.
So much for production testing..lol!..:eek:..:D
http://img30.imageshack.us/img30/4817/samplerun.png
I'm also seeing huge variability in the air out numbers. I think every fan draws and spits out air a little differently and any minor movement at all in the sensors gives you a different result, so while it is interesting information, I don't trust air out myself unless you had some sort of flow chamber to collect and mix the air before measuring it. This would probably be a good idea for the air in also, but I'm not changing now, at least my air in sensors are permanently fixed.
I agree there are alot of flaws in any test setup. There's really no way around it even when you try really hard to minimize it. Probably the best we can do is collect more data points to try and average some of it out and last but not least, don't worry about it and just acknowledge that it's there and take the data with a grain of salt..:D
Do the air out temperatures correlate at all to the flow of the water through the rad? Hotter ones being where the water comes in and colder ones where it comes out (slightly hotter water vs colder water and therefore better and worse heat transference).
It might, but I think with my setup and only having 4 air out sensors, that there is error because the sensors are all on one side. It probably depends on both which side I made the inlet on the barbs and it would also depend on the flow rate because in a 2 pass radiator, the inlet side/half of the radiator would be hotter than the outlet, and as flow rate increases, that difference between sides would get smaller.Quote:
Originally Posted by Erasmus354
You'd probably need at least 4 sensors per fan and they would need to be permanently fixed in an outlet shroud of sorts to get any worthwhile air out data.
I have 8 sensors on my air inlet and that's really not that much when you consider how much it changes over the radiator. Fortunately, I'm not really using the air out sensors for any of the c/w or heat dissipated numbers, it's more just for information. Air is just extremely touchy when it comes to measuring temperature, it's pretty amazing how sensitive it is. A small thing like someone walking into a room will throw the sensors for a ride...:eek:
Also any air movement in a room like a fan or ceiling fan, or any little thing like that will give you different results. This all adds up to more error and makes it hard to make consistent comparisons.
Getting towards standardised methods here...Quote:
You'd probably need at least 4 sensors per fan and they would need to be permanently fixed in an outlet shroud of sorts to get any worthwhile air out data.
That was exactly BillA's early testing methodology (iirc)... airflow and temperature measured at each quadrant (minimum) of the fan (if in a duct, using duct traverse method [Link 2] - which is what relttem's on about above - there's an industry standard set of guidelines and methods for doing such things - these should be referred to and adhered to wherever possible to ensure valid data. Bill tried to highlight the necessity to adhere to such guidelines for any airflow based testing, which also points out that you shouldn't test axial fans using a duct - they aren't designed for use in such scenarios - use an airchamber instead - you can see his 'hints' at this in the aftermath relating to Vapor's original fan review threads... comments were also made on folks doing noise testing of fans, which basically meant read this and this). BillA's data was used to produce the coolingmasters article if memory serves, before he got the windtunnel/envirochamber (which he acquired to counter the "small thing like someone walking into a room" effect)... about 6 months prior to doing the PA testing (by which point he DID Have the windtunnel / envirochamber).
Ultimately, if testing fans and airflow, either on a radiator or not on a radiator... reference this book
If working out resistance to airflow of a radiator (or anything else) have a read of this one for an idea of correct methods.
Info on temperature fields of radiators (uses domestic household radiators, but gives some good clues to what goes on within w/c radiators) here
And SUPERB info on radiator testing as used in aeronautics - old (as in 1920s) but still very relevant - HERE
I heartily recommend serious testers read all the above links and bookmark them for future reference... the closer folks get to using the standardised methods for everything, the more accurate the data you get... and if all testers follow the same standardised methods, using recommended standardised equipment, then data begins to become comparable across different testers.
Air is a challenging one, unfortunately there's just not incentive for hobby testers having fun to spend that sort of money on equipement. We're all here for fun, not to go broke for the fourth decimal point.Quote:
Originally Posted by Marci
Absolute accuracy we'll never have, but I wouldn't go as far as dismissing the value in relative comparison testing. There is always some value in that even if it's just entertainment which is probably the main reason most of us do water cooling and hang out here..
The only solution to absolute accuracy in testing would be for the manufacturers to adopt some standards for them to test and publish data for. They are the only ones with the incentive and the funding to support that sort of effort. Unfortunately we can't even get simple testing like pressure drop tests out of most manufacturers, so it's wishful thinking that there would be any sort of collaborative effort to create standards.
Expecting hobby testers with shoestring budgets to meet that sort of expectation is completely unrealistic.
As long as people are having fun...it's all good! After all we only have a DTS sensor resolution of 1C and many folks live in homes with 10-15C ambient fluctuations and there's also alot of hardware that's not all that affected by temperature anyway. That's how I look at it. I enjoy the science and strive for accuracy where I realistically can and want to, but you have to stop tripping over pine cones now and then and take a view of the forest or you're missing out...
Martin, from your scatterplot above showing equilibrium times, what heat load was being applied for that run (I'm doing the same on my data sets)?
That was a 590 watt run.Quote:
Originally Posted by skinnee
The wierd thing I'm noticing is the time to equilibrium is not really heat based as much as it is the Delta and the volume of water in the test system.
I also ran an 200 watt load that took just as long because the delta was high from running very low 400RPM fans.
My reservoir for the heaters is a 4" x 24" piece of PVC, so when I add up the volume from my heater reservoir AND the large radiator, it's well over a full gallon of water in my test loop.
I have found that you can speed up the heatup process by estimating the delta from previous runs and heating the loop without fans first....just don't forget about it:eek:
I think ideally for this delta T testing you want the bare minimum of water volume. That and simply alot of time and leave the room and come back in a couple of hours for a really good run. My data points are falling on the trendlines much more accurately with this run. Now instead of stopping the logging run, I simply copy the file over to another and open it, plot the run and check to see if it's good enough to stop before I stop it.
I just wish there was a way we could automatically log and plot the temps on the crystalfontz. I'm sure it could be done if someone had the programming expertise.
I think the only time that the wattage load should make a difference is when the radiator has peaked out. A radiator will only be able to dissipate so much heat, and when you have reached that limit you will start to see stuff like run-away systems, but until then it should be the relatively the same thing.
Again, with the mdot*Cp*(T2-T1)
knowing the wattage, T1 and the flowrate you can solve for T2. That will be neglecting natural convection and radiation losses, but without and fans those two will be small. You can graph that equation for a constant Q and see how T2 will vary with your mdot if T1 is constant.
It would be nice to see a comparison of the equation versus data if you are running a few tests without any fans. That will give you an idea of losses from the radiator itself.
Not entirely true. Higher heat load means that you need a higher delta to dissipate that heat (assuming constant fan speed). That means the water needs to get hotter, which means you need to wait longer for that volume of water to heat up. The biggest thing here though will still be the volume of water in the system.Quote:
Originally Posted by relttem
That is exactly true, but not all radiators can provide that necessary delta, therfore they are limited. All radiators have a specified heat load that they max out at..Quote:
Originally Posted by Erasmus354
it all goes back to
mdot*Cp*(T2r-T1r) = h*A*(T2-T1)
again, h, is related to the air speed, but it has a limit. So, when that has peaked out you only have one option left and that is to lower T1. T1 is going to be your ambient temp, so it can only go so low in practical cases. Sure you can run cooled air thru the radiator, or Nitrogen gas etc, but in practical systems it will be ambient. So, the equation part on the right hand side is maxed out and will equal X Watts.
Now, we move to the left hand side. Cp is constant. T1r, mdot and T2r can/will vary. Mdot will max out due to pressure drop thru the radiator and pump limitations. Now, if we keep pumping heat into the system, T1 goes up and T2 goes up, thus the run away system