Evaluating HSF fans that are subjected to backpressure

Felger Carbon · Post by **Felger Carbon** » Sat May 31, 2008 12:37 pm

from an earlier posting, repeated as reference:

YL SM 1041RPM 560prop 48.2dBA
GW NCB 970RPM 560prop 44.9dBA (NCB sample #2)
SC PWM 910RPM 560prop 48.6dBA Kama PWM 100%
AC PWM 902RPM 560prop 46.1dBA Arctic Cooling PWM 100%
UC12EB 880RPM 560prop 41.2dBA (a sweet-sounding fan)
SilenX 866RPM 560prop 45.4dBA "11dBA" version
SFlexE 852RPM 560prop 42.3dBA (quieter than NCB!)
9blade 686RPM 560prop 39.6dBA Scythe SlipStream

The above measurements were made under zero backpressure conditions: free air! All of the fan measurements us SPCR folks encounter are made under zero backpressure, whether by the staff of SPCR or in anechoic chambers by big business. Since we commonly use fans exhaust hot air from cases or to cool HSFs, both of which involve backpressure, these measurements aren't very helpful. But they're the best data we've had available.

Here's how that can be fixed: Use a single HS such as the Thermalright U120E (TRUE). Use a single airflow sensor propeller to measure the airflow provided by various fans (when mounted on the TRUE) and adjust the fan voltages so all the airflows are the same (equal airflow); see above for the free-air example. Measure the fan noise under equal airflow conditions through the HSF fins (again, see above example). The lowest noise level wins.

Two possible configurations, upstream (air source) to downstream (air destination):
1) Sensor prop, fan under test mounted on the TRUE, TRUE with sensor prop as close as possible to the fan.
2) Fan under test mounted on the TRUE with the sensor as close as possible to the backside of TRUE.

This test should NOT be performed in a PC case or on a mobo. The TRUE has no reason to be mounted on a CPU socket; all we're measuring is the effect of the HSF fins on airflow. My present equal airflow fixtures (for 120, 92, 80, 60mm fans) are standalone devices and need no PC.

PC systems with HSFs like the TRUE always have exhaust fans to overcome the resistances associated with the case.

The above is an exceedingly simple (and obvious) extension of my existing equal-airflow test jigs (which work fine).

Bluefront · Post by **Bluefront** » Sat May 31, 2008 1:04 pm

The idea sounds reasonable, and I guess the use of a relatively common heatsink, could enable the tests to be duplicated by those interested, using a standard resistance.

I'd be interested if the comparative results were the same if the fan were blowing or sucking through the resistance. I suspect the airflow would be greater if the fan were sucking, but the comparative results would be similar.

Maybe you could furnish some conclusive evidence that certain fans dealt with resistance better than others. I still hear people doubting that Slip-Stream fans handle resistance well.....

FartingBob · Post by **FartingBob** » Sat May 31, 2008 1:37 pm

This would definitiely be an improvement and make the numbers more in line with actual results. I dislike how many sites still do not test cooling options in a case, rather they just use an open testbed, which makes them far from realistic, although it can be used to say X HSF is 3c cooler than Y HSF. But the actual numbers reported are irrelevant.

Felger Carbon · Post by **Felger Carbon** » Sat May 31, 2008 3:32 pm

FartingBob wrote:This would definitiely be an improvement and make the numbers more in line with actual results.

Um. I decided not to include cooling as part of the measurement because:
1. Variable quality of HS mounting (775 push-pins etc).
2. IHSs and HS bottoms that aren't flat.
3. variable quality of thermal compounds.
4. variable quality of thermal compound application.
5. Case and mobo interference with airflow and the sensor prop.

So: I would expect substantial cooling variance for the above reasons.

But I am keenly interested to see how Slipstream, 9-blade Noctua, 7-blade NCB and SFlex would compare noisewise while pushing the same airflow thru HSF fins. I predict NCB whips SFlex. I'm ecumenical on Slipstream vs. 9-blade Noctua.

DeMoB · Post by **DeMoB** » Tue Jun 10, 2008 12:55 pm

I'm supprised no one has tried to do this before. If you ever manage to get some testing done, would love to see the results.

Dan

Modo · Post by **Modo** » Tue Jun 10, 2008 1:34 pm

Since it was reported by many users that conditions on a heatsink and as a case exhaust are different (proximity of the obstacle might be an issue), I think that there should be a separate test for this. Basically use one case, closed, empty, forcing one exhaust to be used (maybe block nearby vents, especially for the PSU?) with different fans.

Felger Carbon · Post by **Felger Carbon** » Tue Jun 10, 2008 6:13 pm

I've been gathering parts. For a heatsink impedance, I need a rectangular fin block. I've already received an SI-128SE for the high-Z (Z for impedance in electronics, so I'll use it here) HS - the TRUE isn't rectangular, and an Xigmatek HDT-1283-alike (GlobalWin Revolution RVT-12025) for a mid-range HS Z (about the same as an XP-90, which isn't rectangular). I'll have to truncate the 128SE's heatpipes etc to make a test jig, but the RVT can be used as-is.

I'm gonna test 2 7-blade fans and 3 9-blade fans initially. GW NCB low-pitch 7-blade, SFlex high-pitch 7-blade. 9 blades: high pitch Slipstream, mid-pitch Enermax Magma, low-pitch Noctua. The new NCB fans aren't in yet; the others all are.

I agree that the HSF-Z fixture should be separate from the exhaust-Z fixture. So my exhaust-Z fixture will simply be a funnel-type fan adaptor, 120mm to 92 or 80mm, with slight mods to the 80mm to get added Z (plus my existing equal-airflow test jig). This can be as simple as tape across the 80mm end to reduce the effective open area (and so increasing Z).

If this stuff works out, the Slipstream that easily wins the free-air (no Z) tests may be challenged or surpassed as the Z increases, either HSF or exhaust fixture.

I'm amazed that most people think the S-Flex fan is a low-pitch fan. In fact, it's the highest-pitch 7-blade 120x25mm fan I know of.

Felger Carbon · Post by **Felger Carbon** » Wed Jun 11, 2008 12:40 pm

Here are the two HSFs, an RVT-12025 and an SI-128SE:

First we cut the four SI-128 heat pipes. Crimp the tube flat with the right-hand tool, cut the flat tube with the center tool, and bend the tube ends with the left-hand tool so the heatpipe ends clear the fin stack:

The CPU plate assembly now easily slides out of the fin stack, where it was held by two steel rods:

All that's needed to complete an equal-airflow test jig for HSFs is to add a sensor propeller and some sort of framework to hold things in place.

thejamppa · Post by **thejamppa** » Wed Jun 11, 2008 1:52 pm

It actually horrifies me to see perfectly good heatsinks to be... mutilated like that. But I understant reason for it. I am waiting the results. May the spirit of SI-128 SE rest peacefully, for its sacrifice was not invain... I hope...

Felger Carbon · Post by **Felger Carbon** » Thu Jun 12, 2008 8:49 am

The fan adaptor on the fan is a Nexus 80/92mm/120mm from Coolerguys, as is the blue adaptors shown. The hole openings are 79 and 85mm, respectively. Jab-Tech carries the blue adaptor (called a "funnel" adaptor) in both 80mm and 92mm sizes. The screws came with the Nexus.

Another view:

If I use tape instead of screws to mount the adaptor on the fan, then I can make a fixture almost identical to the free-air 120mm fixture. If I carve some plastic off the bottom two corners of the adaptor, I *can use the existing 120mm fixture*!

The Xigmatek/Revolution has neat rubber shock mounts for the fan that aren't too much trouble to mount once or to clean the fan once a year. Constantly switching fans, though, would be a real pain. If I cut off the part of the rubber that pulls all the way through the fan hole, I can use tape to hold the fan on the remains of the rubber shock mounts and still switch fans quickly and easily.

The SI-128SE will require some thinking about. I tend to want to stay with the wire clips...

I hope it's obvious that an 80mm adaptor corresponds to choking the intake area of the fan by 50%, pretty typical for a well-arranged exhaust fan in a computer case with the grille removed. I'm sure 50% will be about right to simulate your computer case airflow Z, but some of the other SPCR types are probably choked more than that - especially if they haven't removed the grille by cutting or dremeling!

[brag mode on]
I have a computer case that lets the exhaust fan run at 75% of its free-air CFM! Honest! In a regular case/PSU, there are two 120mm fans and the PSU eats ~33% of the case intake total area, leaving only 67% for the exhaust. I removed the fan from the PSU, which not only allows the exhaust to use 100% of the original intake area, but turns the PSU into another 33% intake area!

That simple removal of the PSU fan increased the effective intake area for the exhaust fan from ~67% to ~133%! Recommended up to 160W wall power.
[/brag mode off]

Felger Carbon · Post by **Felger Carbon** » Mon Jun 16, 2008 4:10 pm

The hi-Z test fixture, based on the SI-128SE, ready to go:

A view from the rear, showing the sensor prop reasonably close to the fin cooling stack:

This shows the two angle braces on the white base that establish the distance from the sensor to the fins (same every time), and less clearly shows the ~.22 inch rubber base that allows the fan to be mounted with the standard fan clip wires without modification. Made from 3/32 and 1/8 inch sticky-back rubber from McMaster. The same rubber has a high friction coefficient with the styrene base, so once a position is established against the braces, the fin stack/fan assembly won't move unless you firmly want it to!

The Medium-Z fixture based on the HDT-S1283 will be very similar when I get around to building it. Me, I'm kinda curious how the Slipstream will compare to the new Noctua under what I believe is hi-Z backpressure. S-Flex and GW NCB, same thing.

There's also a rubber pad covering all of the styrene base bottom, to keep the fixture from sliding around in use. More 3/32" sticky-back. FYI, the base is 124mm square.

jhhoffma · Post by **jhhoffma** » Tue Jun 17, 2008 5:57 am

Absolutely awesome, Felger!

Can't wait to start getting some numbers!!

Felger Carbon · Post by **Felger Carbon** » Tue Jun 17, 2008 9:54 am

jhhoffma wrote:Absolutely awesome, Felger! Can't wait to start getting some numbers!!

Alas... the purpose of experiments is to learn. I learned a lot this morning when I gave the fixture its trial run. I just posted the results in this forum, using a separate thread.

Well, one Z fixture down, two to go... I predict that at least one of the two will work as advertised (fingers crossed).