push/pull q
 

Cant find that info that pH put up - the push/pull stuff, obviously. Cant remember whether he undervolted them or not.

undervolted push/pull - any different than full speed push pull? (noise/performance wise)

I have my own thoughts, just want them confirmed by someone who has hard DATA on it.

Cheers

IIRC, Bobkoure tested push-pull with L1As. Found a small but significant improvement at 5V, less so at higher V. Bob?

Memory ain't so great - Bob stacked the fans on the pull side. More details here and here. BillA agreed with Bob's analysis in the first link.

Thanks hammer :)

http://phaestus.procooling.com/shrouds-BIX.GIF

Was 100CFM Papst fans @7V; I couldn't ****ing STAND the noise at 12V

woop, missed this pH - much appreciated, cheers!

Was it 1x120mm each side, or two?

Wow, thats one hell of a spreadsheet. Where have you been hiding it!!!???


Did you repeat this with them at 12V just to see how the numbers looked?

1 vote to sticky that.

Etacovda, Tempus and Ruiner would be interested what you deduce from the data.

My analysis gets no further than the "Heat Balance"

http://www.jr001b4751.pwp.blueyonder.co.uk/phRad1.jpg

From which I deduce either the coolant has some interesting heat gains or there is a calibration problem.

The radiator is also dealing with the Hydrothruster pump; easily 50-70W heat added to the loop from that point. Loop was set up as:

res-> pump-> rad-> wb-> flowmeter-> valve-> res

Ignore the calibration comment.
Problem more likely non-equilibrium.
The "bothnoshroud" being nearest equilibrium.

could well be. Was mostly a "proof of concept" round of testing and was deemed a failure due to complexity of the tests.

Looking back at the spreadsheet though there were recorded offsets for all probes at the top of the sheet so it isn't strictly speaking a calibration issue. The full loop wasn't insulated either so I would be surprised if that didnt somewhat affect the heat balances either (the dT rad and dT wb numbers are short distances but some of the other ones are quite long)

Very much accept was a suck-and-see.
Your conclusions appear similar to mine, in that you have purchased a chiller.
Only posted before others drew, possibly, erroneous conclusions from these most interesting experiments.

I can state the conclusions I drew from the testing:

Push/Pull is a horrible choice because the increased noise far surpasses the improvement in cooling. Based on the increase in noise, I can't see any way that undervolted push/pull fans would ever beat a single fan running at the same noise level.

For the BIX there is almost no advantage to adding a shroud for "push" config but a slight advantage to adding a shroud when pulling air through the radiator (this seems sensible based upon how axial fans respond to intake restriction). I'd expect a far more pronounced difference with shrouds on a heatercore for obvious reasons.

wow, nice plot. Data takes on a different representation like that.

I could tell there was something less than intuitive with the push/pull no shroud. I'd make a wild flying guess and say that, IF the numbers are technically sound, then there has to be some other issues - maybe the unshrouded push fan is causing the pull to starve out ?? dead spot issues ??

----------------

Just looking at the raw data, I'm a bit worried about the air exhust temps for the push fans. I think that throws the data off. The temps with the pull and push/pull are lower b/c of the increased air speed at the point of measurement - I know there is a technical term for "it feels colder b/c the air is moving" but I can't seem to remember it.

Regardless, I think the air exhust numbers are crap, which means the deltaT(rad air) isn't worth much either.

It might be more instructive knowing the surface temps of the rad on each side.

I'm actually wondering if a BIX isn't overkill for the heat load you were dealing with. It might be worth trying something like this again and doubling the CPU heat load to see if the data is still linear.

Using only the Rad data
http://www.jr001b4751.pwp.blueyonder.co.uk/phRad2.jpg
The "bothnoshrouds" should give the lowest CPU temperatures.

However, taking to account the Heat Balance, the "bothnoshrouds" is compromised by adding upto 25w extra heat to the coolant.
Something is surely amiss.
Overall I draw no numerical conclusions

Edit:
C/W calculated per Bill Adams :-
"For thermal resistance, or C/W (having the units °C/W):
C = [(Tinlet coolant + Toutlet coolant) ÷ 2] – Tinlet air all in °C, and
W = m * Cg * (Toutlet – Tinlet) "

I'm guessing that those numbers are well within the margins of error for the temp sensors, air temp sensor, pH's eyes (for any none electronic measurements) and the non-uniform distribution of temps.