More Radiator-Sandwich testing

[HESmelaugh]

Radiator-Sandwiches / Stacked Radiators

It's a while back that I posted some rudimentary test results on stacked rads in this thread.

I still have the two magicool triple-rads from that test but now, I also have heaters (300W) which allow for much better testing. So I gave this whole sandwich-thing another go.

All of the tests were done with said Magicool Slim rads, Nanoxia FX-1250, a 300W heater and a MCP355. All of the numbers are indications of the temperature difference between water- and ambient-temperatures achieved after 35 minutes of heating the water.

Here are the different scenarios tested:

Solo Rad with 3 fans



Solo Rad with 6 fans



2 Rads separate with 6 fans (3 each)



Sandwich with 3 fans


Here, the water flows to the rear rad first. This way, the airflow goes from the cooler to the warmer radiator.


Sandwich with 3 fans - Version 2


Here the water goes to the front rad first, so the airflow goes from the warmer to the cooler radiator. It's clear that this is less optimal than the option above but I wanted to test how much of a difference it makes.


Sandwich with 6 fans




Sandwich with 6 fans - Version 2


This is, again, the less optimal flow with the airflow going from warmer to cooler rad.


Rad-Rad-Fan Sandwich


Here, I only tested the optimal flow-version (airflow from cooler to warmer rad).

I also tested with the radiators in parallel flow, but I lost some of the data and got the rest mixed up, so I'll have to redo those tests.


Results 1



The first, shocking realization is this: The stacked rads almost always perform worse than the solo rad with the same number of fans. The rest of the data is as expected: Two separate rads perform best, more fans are always better and the airflow going from warmer to cooler rad is slightly worse than the other way around.

I couldn't believe that the sandwiches performed worse than the solo rad, initially. I retested everything and got identical results, though.

So, I thought it might be a question of air-pressure. The fans need to build more pressure to move the same amount of air through two radiators than one. This would lower performance.
To test this hypothesis, I set up the loop with the solo radiator again. This time, I installed the second radiator as well, just as in a stacked setup, but didn't add the extra radiator to the loop. This way, the second rad acted as an "Airbreak" in front of the fans and I could see how much of an effect this would have on temperatures.


Results 2 - Airbreak



The results affirm my theory in two ways:
1. We see that the "airbreak"-rad has a huge impact on temperatures (much more than I would have guessed).
2. We see that the performance loss is smaller with higher rpm. At 800rpm the loss is 49%, at 1200rpm it shrinks to 23%.

There should be a tipping point where the fan produces more than enough pressure and the stacked rads start outperforming a solo radiator.
Unfortunately, the only high-rpm fans I have here are Yate Loons and since they have closed corners, I can't do a stacked rad setup with these fans.

But I have some Yates here that are 38mm thick and have open corners. The thicker fans should produce more pressure and might be able to overcome the extra resistance in a stacked setup sufficiently.



Results 3 - 38-mm-fans


The thick Yates unfortunately don't go higher than 1250rpm either, so I couldn't test higher rpm. Though from the results it's clear that there is an advantage to the thicker fans. Now, at 1200rpm, the stacked rad's performance catches up with that of the solo rad.

Too bad it wasn't enough for the stacked setup to get ahead, but I already have something else planned.
The thing is, the Magicool rads seem to simply have too high an FPI count to be useful for stacking. At least at the kinds of fan speeds that my ears tell me are reasonable.

So I ordered a second Magicool Slim Elegant rad. This rad has much lower FPI count and did exceedingly well at low rpm. This should make it an ideal candidate for rad-sandwiching.

So, while I once again have to leave you with an "I will do more testing on this later", I think I have come to a useful conclusion here: When stacking rads, the fan pressure requirements increase greatly. Low FPI-rads as well as high-pressure fans should be preffered for such setups.

Hope you liked this report.

Cheers,
Shane

[Jhors2]

Amazing post, thanks for the results!!!

[Eddie3dfx]

Shane, thank you for this report.
I believe the results throw a serious money wrench into the whole sandwhich equation.
Cathar came on here recently and was quite shocked at people sandwhiching their radiators, because of performance decrease. I believe your results have proven his theory, though other radiators and fans would have to be used to get a full answer.

[HESmelaugh]

Thanks, guys.

Right, I think it's important to see that sandwiched rads aren't automatically better than a solo rad. But I don't want people to come to the conclusion that stacking rads is generally bad just because I didn't have the right rads and fans for the job.
The point of this post is to show that selecting the right kind of components for this thing is crucial. I'm still hoping that I can set up a good sandwich for low-rpm fans with the new rads I got. We'll see.

I'll post an update as soo as I finished the next round of testing.

[Vapor]

Nice testing...good to show sandwiching isn't always a free lunch.

Would like to see more higher RPM testing though for a couple of reasons:

1) air saturation levels of most radiators decline as airflow increases (meaning the air leaving the first radiator still has a decent potential to absorb more air)

2) static pressure of a fan scales with the square of the RPM. Airflow scales linearly with RPM. At higher RPMs, fans should have little problem moving air through the sandwich The increase in static pressure from increasing RPM is way greater than switching from a 25mm fan to a 38mm fan. Though the right low-speed 38mm fan could be a decent middle-ground (aren't many out there though )

Of course though, showing a lack of efficiency at lower RPMs is critical to knowing how well it performs too though

p.s., I like the idea of the airbrake testing!

[Zehnsucht]

Wow these results are truly surprising. I'm just about to pull the trigger on a MCR-220 combo. Maybe I should order two 38mm fans as well? I was planning to use petras YL low speed.

How does an undervolted Scythe Kaze 2000 sound?

[Eddie3dfx]

What I would love to see is swiftech the line tested, but more so a dual mcr-120 sandwich. For some of these cases that have very little room, but maybe enough for a few 120's, it would be perfect, especially for lan boxes.

[SNiiPE_DoGG]

merely a result of using 1250 rpm fans with weak static pressure capability, sanaces are a different story

[lennox]

what if you stack 3 rads? lol

[Jhors2]

Originally Posted by lennox

what if you stack 3 rads? lol

The space time continuum collapses with the amount of static pressure you need.

[faster3200]

Nice tests, that is some pretty good info. For ver.1 vs. ver 2 where is water going in and out. I assume it goes in the red arrow and come out the blue one. BTW as a Web/Graphic Designer I have to say... nice diagrams. Good idea with the air break too. My only criticism... Results 3 is misspelled .

I am really surprised just how big a gap there is between 800rpm and 1200rpm. At 1200rpm there isn't much of a difference between stacked and unstacked and I bet at 1500rpm+ with higher pressure fans there will be no measruable difference between stacked an unstacked, in this setup.

Originally Posted by SNiiPE_DoGG

merely a result of using 1250 rpm fans with weak static pressure capability, sanaces are a different story

Yeah, I wonder how my Ultra Kaze at 3000rpm change the story.

[SNiiPE_DoGG]

Indeed I have forgotten my compliments, thank you for testing my theory or nverted flow patterns I am glad to see it has ended up successfully better than the opposite direction

and the testing is definitely useful, it at the very least lets people know that they cant use their favorite low speed fan with the rad sandwich

[Zehnsucht]

Maybe it's me being very tired, but in your other sandwich thread you presented these results:



where the sandwich was better than normal?

[soulesschild]

I'm very curious to see the results of MCR320 stacking. As I'm constrained for space and want to put my 4870x2 on water.

@Sniipe
It makes me sad to see your selling PermaFrost. Such a great build

[oRIDDLERo]

Killer testing... thanks for the input!

[faster3200]

Originally Posted by SNiiPE_DoGG

Indeed I have forgotten my compliments, thank you for testing my theory or nverted flow patterns I am glad to see it has ended up successfully better than the opposite direction

Me too. I setup stacked rads about 6 months before all this excitement about them and I followed your advice. Right now I have a .8C delta on idle and a 1C delta on load between air and water temp. We'll see how things change with a more accurate probe and more heat to the loop .

[faster3200]

Originally Posted by Zehnsucht

Maybe it's me being very tired, but in your other sandwich thread you presented these results:

http://www.abload.de/img/graph1wasserluftsandwii69l.jpg

where the sandwich was better than normal?

That was comparing stacked vs. a single rad.

[Zehnsucht]

Originally Posted by faster3200

That was comparing stacked vs. a single rad.

Exactly my point. But in the graph in the first post also tests single versus stacked, and here the single clearly is in favor.

EDIT:

it. I'll bite the bullet and buy an MCR 220 combo and Ultra Kaze 2000. If it works it works. If not I'll just disconnect one radiator.

[faster3200]

Originally Posted by Zehnsucht

Exactly my point. But in the graph in the first post also tests single versus stacked, and here the single clearly is in favor.

Oh wow, how did I miss that. I am going to check his first review and see if it was setup differently.

[HESmelaugh]

Originally Posted by lennox

what if you stack 3 rads? lol

Yeah, you'd need ridiculous fan speed/pressure.
The point Vapor made about saturation is a very important one here. The air would need to move through the rads so quickly, that it's still well above the water-temperature in the third rad once it gets there. Otherwise, you're just blowing hot air at a radiator...

Originally Posted by faster3200

Nice tests, that is some pretty good info. For ver.1 vs. ver 2 where is water going in and out. I assume it goes in the red arrow and come out the blue one.

The flow is shown in a very simplified manner. The arrows are just there to indicate which radiator is getting the cooler water (blue) and which the warmer water (red). But yes, it's red in, blue out. I always set the rads up so that both of them had water inflow at the lower thread. That makes bleeding them a lot easier.

Here's an example of what a setup looks like:


This is for the Rad-Rad-Fan setup.

BTW as a Web/Graphic Designer I have to say... nice diagrams. Good idea with the air break too. My only criticism... Results 3 is misspelled .

Wow, thanks a lot! I'm glad you like the diagrams. And I think I can live with that criticism, too.

Yeah, I wonder how my Ultra Kaze at 3000rpm change the story.

I might get some of those to try it out, too... don't know yet.

Originally Posted by SNiiPE_DoGG

Indeed I have forgotten my compliments, thank you for testing my theory or nverted flow patterns I am glad to see it has ended up successfully better than the opposite direction

and the testing is definitely useful, it at the very least lets people know that they cant use their favorite low speed fan with the rad sandwich

Thanks. Yeah, the flow pattern makes quite a difference. And of course, the lower the flowrate, the bigger the difference this makes.

[HESmelaugh]

About my first review: Yes, there I got different results.
Sorry, I forgot mentioning this.

I tried to figure out what the went wrong with the first set of tests (or the second, for that matter) but couldn't find the problem. The only thing I can think of is error of measurement.
I had a low heat load during the first tests, and since it consisted of actual hardware, the heat load was fluctuating. Additionally, during the first tests I only had one probe for water temperatures. Now I have four, all calibrated.

The only way I can explain the first results to myself is that the method was simply not good enough and due to the low heat load, it didn't take much error to throw the results off completely.

[Metasheep]

Does it make a difference if water flow through the rads is parallel or in series?

[DeathWalking]

Originally Posted by Vapor

2) static pressure of a fan scales with the square of the RPM. Airflow scales linearly with RPM. At higher RPMs, fans should have little problem moving air through the sandwich The increase in static pressure from increasing RPM is way greater than switching from a 25mm fan to a 38mm fan. Though the right low-speed 38mm fan could be a decent middle-ground (aren't many out there though )

Is this with an "ideal" fan, or something? Because I know a lot of fans perform less efficiently at higher RPMs, when you do a direct scaling from lower RPMs.

[NaeKuh]

HES great test, however i noticed you made a HUGH goof in your methodology.

You CAN NEVER ASSUME THE AQUARIUM HEATER IS CALIBRATED PERFECTLY WITH ON ANOTHER.

The heater YOU have can have a + or - 10-15% Wattage heat.

meaning if your

First heater is -15% off 300W = 255W
Second heater is +15% or 345W

Your first and second heater can have a delta of 90W which means your testing is BOO BOO, fail.

You need to use the same heater thoughout ALL your comparison test or your testing and comparision is FLAWED.

[HESmelaugh]

Yes, NaeKuh, the same identical heater was used for every test. It has an actual heat output of 289.9W, monitored per Watt-meter.

Don't worry.

Anyway, as I stated, practically all of these tests were repeated and I got identical results, +/- .1K