Post by thread:[EE]: enclosure cooling.

Hi All, I am looking into the cooling of some enclosures using fans. The enclosures already have fans installed, and due to space constraints they were placed in the top of the door sucking air out of the enclosure. I was going to fit a vent with filter on the bottom of the enclosures (opposite side) to improve air circulation. The question is: Is it better to suck air into the enclosure or suck air out of the enclosure? In my mind it seems better to suck air in, because you can control the air intake and filter it. The only problem I have is that changing it to that configuration will be much more costly and I am not completely sure that it is worth it. Your opinion will be welcomed. Best Regards Luis

> In my mind it seems better to suck air in, because you can control the > air intake and filter it. My take on such things is that if you are sucking the air out, then you are creating a low pressure area inside the cabinet, so you potentially have a lower of air to absorb the heat from local radiation, and as well as this the air will seep in through every gap in the cabinet, potentially bringing dust with it.. If you suck air into the cabinet then you have a higher pressure region, hence higher mass of air, to absorb the radiated heat, and it will then work its way out of every crack in the cabinet, and all the dust attempting to enter is trapped by whatever dust filter you have at the fan.

The quality of the following advice is only as good as the sense it makes. If anything said makes no sense it may be wrong. If it makes lots of sense it still may be wrong, but less probably so :-). If you have equipment in the enclosure that gets significantly hotter than the average equipment then maximising air flow rate over it will be optimal. Adding very small local fans at selected locations inside the cabinet to deal with such hot spots may be practical and easier than a complete change. Getting heat out of equipment into the air is the major task. Getting the air out of the cabinet is easier. If the temperature rise is significant across the cabinet then an exit fan which blows hot air will blow less air mass than an entry fan which blows cold air, BUT the temperature difference would need to be significant for this to matter. Very roughly the effective air mass handled by inlet or output fans working at constant volume is the ratio of the absolute temperatures (eg degrees Kelvin) or in degrees C (Tcold + 273)/(Thot+273). FGor an eg 50C rise fron 20C to 70C (very unlikely for a whole cabinet) then moving fans from output to input will improve flow by about ((273+70)/(273+20) = +17%. As texit will drop when you do this the ratio will adjust slightly. Not liable to be a major factor compared to all the other things you can do. - Maximise flow over hottest surfaces (as above) - Block internal flows which "short circuit"hot areas so air is not wasted. - Consider internal ducting or air when doors are closed to optimise where flow goes. - Add some small local fans to get air flows up in critical areas. Even if you just blow local air over a hot spot and it is not in a mian flow it will help muchly. Once heat is in the air inside the cabinet convectin will help it join other air - if it is in equipment the surface transmission rate will dominate. - Try and ensure that hot areas have good heat sink surfaces and that they are vertical and/or parallel with airflows. - Increasing number of fans at exhaust may help. - Staging exhaust fans if pressure drop is significant or changing fans for units with more pressure capability may help - depending on cabinet pneumatic impedance. R

Quoting "Moreira, Luis A" <spam_OUTLuis.MoreiraTakeThisOuTccfe.ac.uk>: > > Hi All, > I am looking into the cooling of some enclosures using fans. The > ... > Is it better to suck air into the enclosure or suck air out of the > enclosure? > ... (c) *Through* the enclosure...IMO. I'm adding fans to a small embedded PC currently, and have been thinking about this too. From what I've learned from car engines and PC's, the key is to move the cooling fluid (water or air) *through* the hot parts of the engine, with the use of shrouds, channels, etc. I stress "through" because the fluid has to get to the hot parts, and then move away from the hot parts. On my enclosure, I'm adding one fan that pushes, and one that pulls, so it creates a nice flow through the enclosure, but I'm thinking of where I need (and can) add shrouds to route the air exactly over the components generating the most heat. Ideally, I'd make a river of air through the enclosure. I've seen PC enclosures where the fan airflow is blocked by cables, and just tying up those cables out of the way makes a noticeable difference in temps. And I've seen where removing one bay panel far from the fan also helps temps come down. Cheers, -Neil.

And for my 2c worth. Fans removing hot air will run at a slightly higher temperature and so bearings & electronics may fail ealier. And blowing turbulant air onto a heatsink is more effective than sucking laminar flow off it. RP On 29 April 2010 02:43, PICdude <.....picdude3KILLspam@spam@narwani.org> wrote: {Quote hidden}> Quoting "Moreira, Luis A" <Luis.MoreiraKILLspamccfe.ac.uk>: > >> >> Hi All, >> I am looking into the cooling of some enclosures using fans. The >> ... >> Is it better to suck air into the enclosure or suck air out of the >> enclosure? >> ... > > (c) *Through* the enclosure...IMO. > > I'm adding fans to a small embedded PC currently, and have been > thinking about this too. From what I've learned from car engines and > PC's, the key is to move the cooling fluid (water or air) *through* > the hot parts of the engine, with the use of shrouds, channels, etc. > I stress "through" because the fluid has to get to the hot parts, and > then move away from the hot parts. On my enclosure, I'm adding one > fan that pushes, and one that pulls, so it creates a nice flow through > the enclosure, but I'm thinking of where I need (and can) add shrouds > to route the air exactly over the components generating the most heat. > Ideally, I'd make a river of air through the enclosure. > > I've seen PC enclosures where the fan airflow is blocked by cables, > and just tying up those cables out of the way makes a noticeable > difference in temps. And I've seen where removing one bay panel far > from the fan also helps temps come down. > > Cheers, > -Neil. > > > >

I have designed a product with a largish heatsink (~3"x3"x18") an I found that opposing airflow directions on the fans at either end (both blowing into the heatsink) removed heat better than having them both blow the same way. There were other factors.......air then exited through vents on center top and center bottom of heatsink. So I think under certain conditions turbulent air flow can be better than laminar air flow. Mark Richard Prosser wrote: {Quote hidden}> And for my 2c worth. > > Fans removing hot air will run at a slightly higher temperature and so > bearings & electronics may fail ealier. And blowing turbulant air > onto a heatsink is more effective than sucking laminar flow off it. > > RP > > On 29 April 2010 02:43, PICdude <.....picdude3KILLspam.....narwani.org> wrote: > >> Quoting "Moreira, Luis A" <EraseMELuis.Moreiraspam_OUTTakeThisOuTccfe.ac.uk>: >> >> >>> Hi All, >>> I am looking into the cooling of some enclosures using fans. The >>> ... >>> Is it better to suck air into the enclosure or suck air out of the >>> enclosure? >>> ... >>> >> (c) *Through* the enclosure...IMO. >> >> I'm adding fans to a small embedded PC currently, and have been >> thinking about this too. From what I've learned from car engines and >> PC's, the key is to move the cooling fluid (water or air) *through* >> the hot parts of the engine, with the use of shrouds, channels, etc. >> I stress "through" because the fluid has to get to the hot parts, and >> then move away from the hot parts. On my enclosure, I'm adding one >> fan that pushes, and one that pulls, so it creates a nice flow through >> the enclosure, but I'm thinking of where I need (and can) add shrouds >> to route the air exactly over the components generating the most heat. >> Ideally, I'd make a river of air through the enclosure. >> >> I've seen PC enclosures where the fan airflow is blocked by cables, >> and just tying up those cables out of the way makes a noticeable >> difference in temps. And I've seen where removing one bay panel far >> from the fan also helps temps come down. >> >> Cheers, >> -Neil. >> >> >> >> --

What's the theory behind this? Is it really *turbulence* that makes it cool better, or slower airflow (allowing the air more time to grab heat from the components)? Might be interesting to experiment with this using a couple fans and a mass-air sensor. I'll go ahead and add it to my ever-growing project list. :) Cheers, -Neil. Quoting "Mark E. Skeels" <mskeelsspam_OUTcompetitionelectronics.com>: {Quote hidden}> I have designed a product with a largish heatsink (~3"x3"x18") an I > found that opposing airflow directions on the fans at either end (both > blowing into the heatsink) removed heat better than having them both > blow the same way. > > There were other factors.......air then exited through vents on center > top and center bottom of heatsink. > > So I think under certain conditions turbulent air flow can be better > than laminar air flow. > > Mark >

As I am wont to accept every opportunity to speculate about those things I know only a little about, I will do so here :) I think that a turbulent airflow prevents a boundary layer from forming between the heatsink surface and the free airstream. This layer, if it forms, would act somewhat like an insulating layer. Sean On Wed, Apr 28, 2010 at 5:45 PM, PICdude <@spam@picdude3KILLspamnarwani.org> wrote: {Quote hidden}> What's the theory behind this? Is it really *turbulence* that makes > it cool better, or slower airflow (allowing the air more time to grab > heat from the components)? > > Might be interesting to experiment with this using a couple fans and a > mass-air sensor. I'll go ahead and add it to my ever-growing project > list. :) > > Cheers, > -Neil. > > > > Quoting "Mark E. Skeels" <KILLspammskeelsKILLspamcompetitionelectronics.com>: > >> I have designed a product with a largish heatsink (~3"x3"x18") an I >> found that opposing airflow directions on the fans at either end (both >> blowing into the heatsink) removed heat better than having them both >> blow the same way. >> >> There were other factors.......air then exited through vents on center >> top and center bottom of heatsink. >> >> So I think under certain conditions turbulent air flow can be better >> than laminar air flow. >> >> Mark >> > > >

PICdude wrote: > What's the theory behind this? Is it really *turbulence* that makes > it cool better, Yes. Turbulence is a chaotic mixing of the air. Thereby more air molecules than just in the boundary layer of laminar flow come close enough to the object to exchange heat. > or slower airflow (allowing the air more time to grab > heat from the components)? Obviously not. Think about it. More time would allow for more heat transfer to individual molecules, but with faster flow you keep getting more "fresh" molecules. There are always about the same number of air molecules near a piece of the heatsink, since the pressure is the same in both cases. So the extra time for the slowly moving molecules actually hurts since they get to a higher temperature and therefore a lower temperature difference and therefore lower heat flow. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Quoting Olin Lathrop <RemoveMEolin_piclistTakeThisOuTembedinc.com>: > ... More time would allow for more heat > transfer to individual molecules, but with faster flow you keep getting more > "fresh" molecules. There are always about the same number of air molecules > near a piece of the heatsink, since the pressure is the same in both cases. > So the extra time for the slowly moving molecules actually hurts since they > get to a higher temperature and therefore a lower temperature difference and > therefore lower heat flow. But then this contradicts the turbulence theory. Though I'm sure that was not a theory, but one person's observation that may have had other factors involved.

PICdude wrote: > Quoting Olin Lathrop <spamBeGoneolin_piclistspamBeGoneembedinc.com>: > > >> ... More time would allow for more heat >> transfer to individual molecules, but with faster flow you keep getting more >> "fresh" molecules. There are always about the same number of air molecules >> near a piece of the heatsink, since the pressure is the same in both cases. >> So the extra time for the slowly moving molecules actually hurts since they >> get to a higher temperature and therefore a lower temperature difference and >> therefore lower heat flow. >> > > But then this contradicts the turbulence theory. Though I'm sure that > was not a theory, but one person's observation that may have had other > factors involved. > > > > Faster but turbulent flow ? It's not called Wind Chill Factor for nothing!

>> What's the theory behind this? Is it really *turbulence* >> that makes it cool better, > > Yes. Turbulence is a chaotic mixing of the air. Thereby > more air molecules than just in the boundary layer of > laminar flow come close enough to the object to exchange > heat. That's not always true. Turbulence does not always help to cool better. Imagine a computer case with silent radiators inside and a single fan in the case. Non - turbulent air may stream at higher speed. The molecules may get heated by radiation too or through air heat conduction mechanism as well.

PICdude wrote: >> ... More time would allow for more heat >> transfer to individual molecules, but with faster flow you keep >> getting more "fresh" molecules. There are always about the same >> number of air molecules near a piece of the heatsink, since the >> pressure is the same in both cases. So the extra time for the slowly >> moving molecules actually hurts since they get to a higher >> temperature and therefore a lower temperature difference and >> therefore lower heat flow. > > But then this contradicts the turbulence theory. I don't see how. Either way more different molecules moving around close to the heat sink causes more heat flow. Still air is clearly the worst. Moving laminar flow is better, and turbulent flow even better for the same flow speed. I don't see a contradiction. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Marechiare wrote: > That's not always true. Turbulence does not always help to cool > better. Imagine a computer case with silent radiators inside and a > single fan in the case. Non - turbulent air may stream at higher > speed. I was saying turbulence is generally better at the same speed. Certainly fast laminar can be better than slow turbulent. > The molecules may get heated by radiation too Not in any meaningful sense to this discussion. Air is basically transparent to IR for this purpose. > or through air > heat conduction mechanism as well. Meaning convection, apparently, which is exactly what this discussion is about. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

>> That's not always true. Turbulence does not always help >> to cool better. Imagine a computer case with silent >> radiators inside and a single fan in the case. Non - turbulent >> air may stream at higher speed. > > I was saying turbulence is generally better at the same speed. > Certainly fast laminar can be better than slow turbulent. No, "at the same speed" was said by you after my post. The context of the thread was and is "at the same fan". >> The molecules may get heated by radiation too > > Not in any meaningful sense to this discussion. Air is > basically transparent to IR for this purpose. Probably my statement was too terse (hey R, teach me to be as verbose as you :-) , definitely I should have used sort of "indirectly" word, like this "The molecules may get heated by indirect mechanism of radiation too" Just imagine a computer case filled with some stuff, and 1 kWt incandescent VACUUM lamp inside it. Almost all energy consumed by the lamp is radiated by it. The stuff around the lamp would get heated by radiation, and then it would heat air. That's what I meant, - the air is getting heated by lamp radiation indirectly. In this case the main concern would be to stream as much air as possible through the case. The turbulence within it may cause the air flow to slow down. >> or through air >> heat conduction mechanism as well. > > Meaning convection, apparently, which is exactly > what this discussion is about. No, meaning exactly what I said "heat conduction mechanism". Still air does conduct heat quite well. In your neck of the woods building regulations state that the width of the outer wall heat insulation should be at least 50 mm or so if I am not mistaken. Have a look at those huge "silent" (that is fan-less) radiators stuffed dense into a case. With all the cables around them, the convection is seriously limited. The power dissipated is not that big, so the air flow is rather slow. Air may be heated through its mass by just air heat conduction.

Hi All, Thank you very much for all your replies. The conclusion that I got to, is that the all thermal management for this enclosure is very badly done and it would explain why everything is "cooking" in it. This enclosure as a air cooled transformer in it which as losses of about 12KW (optimistic estimation for 90% efficiency), to maintain a feasible 45 degrees C the air as to be changed at least every 0.42 Seconds if ambient is 35C. This is clearly not feasible. I need to start thinking about cooling the air intake. Best Regards Luis {Original Message removed}

Holy frijoles, Batman! 12kW power dissipation?! I missed the beginning of this thread - how big is this enclosure? I would think that you need to take a step back on this design. I would not want to rely on active air-conditioning to take care of that heat load from a transformer, for safety reasons. Does this mean that your total input power is 120kW? (90% efficiency) Sean On Tue, May 4, 2010 at 6:43 AM, Moreira, Luis A <TakeThisOuTLuis.MoreiraEraseMEspam_OUTccfe.ac.uk> wrote: {Quote hidden}> Hi All, > Thank you very much for all your replies. > The conclusion that I got to, is that the all thermal management for > this enclosure is very badly done and it would explain why everything is > "cooking" in it. This enclosure as a air cooled transformer in it which > as losses of about 12KW (optimistic estimation for 90% efficiency), to > maintain a feasible 45 degrees C the air as to be changed at least every > 0.42 Seconds if ambient is 35C. This is clearly not feasible. > I need to start thinking about cooling the air intake. > > Best Regards > Luis > > > {Original Message removed}

Hi, Before all the calculations, I would like to say that are a lot of things I am not considering, like heat dissipation from the sides of the enclosure, colour etc, and I am making a few assumption. I started with the formula: Q = m*c*deltaT Where Q is the energy dissipated on the enclosure, m is the mass of the air in the enclosure, c is the Specific heat capacity of air and deltaT is the difference in temperature required i.e. ambient is 35C and maximum temperature inside the enclosure required is 45C. For the mass I started by measuring the enclosure and calculated its volume, than I added a factor which is a rough estimation of how much of that volume is air, due to the stuff inside the enclosure. In my case I estimate that only 40% is air that I can move. >From the volume calculations in m^3 and the air density per m^3 which roughly is 1.184 Kg/m^3 you get the mass value in Kg. The Specific heat capacity of air is around 1.005 KJ/(Kg K), note the units. This value changes with temperature but for the ranges I am looking into a value of 1 will be fine, if you want the accurate value just Google for the tables. DeltaT is the difference between the maximum required enclosure temperature, in my case 45C and the max ambient temperature, in my case it can be 35C. This calculation will give you Q which is the energy that you can put in the cubicle and still have the enclosure at the required temperature. Now I know by looking at the documentation I have that this transformer has losses of 12Kw, from this value and Q I can calculate how long I have to change the air to keep the enclosure at around 45C, as joules is watt*second. In my case I have 0.82 seconds to change the air. After all this you can then find a fan or any other system to shift the air or lower the input temperature. Fan manufacturers will give you the flow rate in l/s, l/h or m^3/s, hence you will have to convert to get the right units. For my fan I have 275 l/s which is 0.275 m^3/s. By looking at the volume of air in the enclosure, in my case 0.82 m^3, you can see that it will take 2.98 seconds to change the air, hence this fan is not suitable. I have to stress that my calculations may not be correct or I may doing it in a longwinded way, but it seems to tie-up with the current situation. No doubt people will find holes on my calculations and assumptions, please fell free to comment and put me right. Best Regards Luis {Original Message removed}

Moreira, Luis A wrote: > For the mass I started by measuring the enclosure and calculated its > volume, The volume of the enclosure has little to do with it. You're making a simple thing complicated. This is basic freshman year physics. Using the specific heat of air, you calculate what mass of air per second is required for that power for your tolerable temperature rise. Using reasonable pressure and temperature assumptions, you convert the to flow (volume/time). That's what fans are rated at. For a reverse example, let's say you blow in 1 cubic foot per second at 20degC and 1 atm. 1 cubic foot is 28.3l, which is 1.18 moles in this case, which is 34g using 29 as the average molecular weight of air. So, we've got 34g of air per second. The specific heat of air is 1000J/Kg/degC, so this stream carries away 34J per second (=W) per degC rise. If the outflow is 40C, then this air stream is carrying away 683W in heat. Scale from there. Note that this says nothing about how the air swirls around, turbulent or laminar flow, where the vents are, etc. If you've got 1 cubic foot per second of air at 20degC and 1atm going into a black box, and it comes out at 40degC, then 683W of cooling is being performed. You can't cheat or get around this basic physics. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

On Thu, 6 May 2010, Olin Lathrop wrote: > Note that this says nothing about how the air swirls around, turbulent or > laminar flow, where the vents are, etc. If you've got 1 cubic foot per > second of air at 20degC and 1atm going into a black box, and it comes out at > 40degC, then 683W of cooling is being performed. You can't cheat or get > around this basic physics. Excuse me, but this doesn't hold true if the speed and pressure of the air coming out is not the same as that going in. Remember: increasing speed meens decreasing pressure (at same temp) and vica versa. increasing pressure meens increasing temp (at same speed) and vica versa. Regards Sergio Masci

Michael Watterson wrote: >> The specific heat of air is 1000J/Kg/degC, so this >> stream carries away 34J per second (=W) per degC rise. > > Just Curious. > > I presume at 1 atm / sea level approx. It doesn't matter much. Think about it. The specific heat is the energy per *mass* per temperature rise. > How much effect has humidity? I don't know. If I needed to know I'd look it up. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

sergio masci wrote: >> Note that this says nothing about how the air swirls around, >> turbulent or laminar flow, where the vents are, etc. If you've got >> 1 cubic foot per second of air at 20degC and 1atm going into a black >> box, and it comes out at 40degC, then 683W of cooling is being >> performed. You can't cheat or get around this basic physics. > > Excuse me, but this doesn't hold true if the speed and pressure of > the air coming out is not the same as that going in. Argh. Why do I sometimes feel like I'm the only one who paid attention in freshman physics class!? The same *mass* of air is coming out as is going in. Note that nothing I said above depends on the volume flow of the air coming out. However since you brought it up, the volume flow will be proportional to temperature, which in this example is 313degK/293degK = 1.07. So the output flow will be 1.07 cubic feet per second. The pressure in both cases is the ambient, which in this example is 1 atmosphere. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

> Note that this says nothing about how the air swirls around, turbulent or > laminar flow, where the vents are, etc. If you've got 1 cubic foot per > second of air at 20degC and 1atm going into a black box, and it comes out at > 40degC, then 683W of cooling is being performed. You can't cheat or get > around this basic physics. However, you can blow that much air in and have it come out with far less temperature rise, initially, while the equipment inside starts to heat up due to poor energy transfer from equipment to air at the delta-T's concerned. After the equipment gets hot enough (if it doesn't expire first) then all the energy will get transferred steady state to the air flow BUT if you want only 10C delta-T you are going to need either MUCH more air or better internal heatsinking. >From information given you have 12 kW power dissipation and 10C ris maximum or 10/12000 =~ 0.008 degrees C/Watt heatsink. [!!!!] While that is the effective C/W of the whole cabinet and as you have a number of radiators in thermal parallel, the actual C/W of any assembly will be higher than that, BUT that's an immensely capable heat transfer system, and not one you are going to easily achieve in practice. I'd seriously suggest looking at liquid cooling. The specific heat of water is about 4.2 J/g/K or about 4 x that of air BUT as it is a liquid you need a much lower flow rate. 12000 Watt = 12 kJ//sec. So you need about 12000/4.2 =~ 3 litres per second of water for a 1 C rise or 300 cc / second for a 10C rise. As with air, you have to get energy transfer into the water with low delta-T, but this is far easier than with air. "A few litres per second" of water and some copper tubing should work wonders. What to do with the removed heat is another issue :-). Heat pipes are also very useful but less easy to adapt to a custom situation. Heat is transferred by vaporisation with return by liquid flow - usually by wicking. Very high energy transfer rates in a small space can be achieved. . Russell

Jonathan Hallameyer wrote: >> However since you brought it up, the volume flow will be >> proportional to temperature, which in this example is >> 313degK/293degK = 1.07. So the output >> flow will be 1.07 cubic feet per second. The pressure in both cases >> is the ambient, which in this example is 1 atmosphere. > > But if there is no pressure difference, how would the air flow? Obviously there has to be something inside the black box propelling the air. However, if this box is sitting in the middle of a room, then the air a little before it enters the box and a little after a basically at the ambient pressure. Yes, the air near the intake has to be ever so slightly below the ambient else the room air would never move towards it, but this pressure difference is very very little and well below meaningful levels of this example. It's the same as when you move your hand thru the air it produces a pressure wave 3 inches in front of it. Yes it does, but not to the point where it even remotely matters for what we're talking about. Even for sub-sonic air flow over a airplane wing, the approximation of air as a incompressible fluid is quite useful and common. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Moreira, Luis A wrote: > I think the volume of the enclosure as everything to do with it, If > you look at the equation for Q you will see that depending on the > volume, for the same conditions you can have a smaller temperature > differential. The volume has nothing to do with the mass of air that needs to flow thru to get a certain level of cooling. It does of course enter in if you want to express air flow as number of air changes per time. I don't know what you think Q is, so I can't comment. However, I've shown my physics in detail. What specifically do you claim is wrong? If you can't find anything wrong with it, then you have to agree with the conclusion. I have left your quote of my post below so that you can refute specific points of my analisys: {Quote hidden}> For a reverse example, let's say you blow in 1 cubic foot per second > at 20degC and 1 atm. 1 cubic foot is 28.3l, which is 1.18 moles in > this case, > which is 34g using 29 as the average molecular weight of air. So, > we've got > 34g of air per second. The specific heat of air is 1000J/Kg/degC, so > this > stream carries away 34J per second (=W) per degC rise. If the outflow > is > 40C, then this air stream is carrying away 683W in heat. Scale from > there. > > Note that this says nothing about how the air swirls around, turbulent > or > laminar flow, where the vents are, etc. If you've got 1 cubic foot > per second of air at 20degC and 1atm going into a black box, and it > comes out at > 40degC, then 683W of cooling is being performed. You can't cheat or > get around this basic physics.

Russell McMahon wrote: > However, you can blow that much air in and have it come out with far > less temperature rise, initially, while the equipment inside starts to > heat up due to poor energy transfer from equipment to air at the > delta-T's concerned. After the equipment gets hot enough (if it > doesn't expire first) then all the energy will get transferred steady > state to the air flow BUT if you want only 10C delta-T you are going > to need either MUCH more air or better internal heatsinking. Yes. I was trying to show some examples and how the physics worked. I wasn't pretending to address the OP's problem specifically or decide what the right air flow would be. As you say, that depends on the thermal resistance between the equipment being cooled and the air stream, and on the tolerable temperature rise of the equipment. However, I believe the physics I did present and the example are correct. I agree with your conclusion that in the OP's case air cooling sounds intractable, and water cooling should be seriously considered. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Moreira, Luis A wrote: > I think the volume of the enclosure as everything to do with it, If > you look at the equation for Q you will see that depending on the > volume, for the same conditions you can have a smaller temperature > differential. No (as others already explained :) If you (differently than elsewhere suggested) translate your description into plain math, using the correct equations, you'll see that during the process the case volume will be canceled out and the final equation about the relationship between energy flow and air mass flow under the given constraints will be the same as translating Olin's approach into plain math. Try it... This is not only good for seeing that the two approaches actually lead to the same result, but it is a good thing to do in every case -- because it shows you what, in the end, is really influencing the outcome by canceling out on the way the unnecessary assumptions (like in this case the case volume). And you can calculate the units in the final equation. If they work out right, it doesn't mean that it's necessarily correct -- but if they work out wrong, it definitely isn't correct. (This of course is tremendously simplified by sticking to SI units.) Gerhard

If you look at my first post, you will see the equation there. Luis -----Original Message----- From: piclist-bouncesEraseME.....mit.edu [EraseMEpiclist-bouncesmit.edu] On Behalf Of Marechiare Sent: 06 May 2010 16:51 To: Microcontroller discussion list - Public. Subject: Re: [EE]: enclosure cooling. > Maybe you should pay more attention to the posts. I believe I should not. You are asking for help in this thread, so that's your duty to present your question the way others would be able to get the idea you talking about. If you pretend to be scientific by relying on some super formula, place a reference to where the formula resides. > But I will endeavour to translate into Olin's English for you. ok Regards.

Moreira, Luis A wrote: > If you look at my first post, you will see the equation there. Below is your first post in this thread, where is the equation in it?: On Wed, Apr 28, 2010 Moreira, Luis A wrote: {Quote hidden}> I am looking into the cooling of some enclosures using > fans. The enclosures already have fans installed, and > due to space constraints they were placed in the top of > the door sucking air out of the enclosure. I was going to > fit a vent with filter on the bottom of the enclosures > (opposite side) to improve air circulation. The question is: > Is it better to suck air into the enclosure or suck air out > of the enclosure? > In my mind it seems better to suck air in, because you > can control the air intake and filter it. The only problem > I have is that changing it to that configuration will be > much more costly and I am not completely sure that it > is worth it. > Your opinion will be welcomed.

Is not that part of Olin that you are emulating. In fact everything you posted about this subject and me answering you back added no value whatsoever, hence I suggest we stop posting to each other. Luis -----Original Message----- From: RemoveMEpiclist-bouncesEraseMEEraseMEmit.edu [RemoveMEpiclist-bouncesspam_OUTKILLspammit.edu] On Behalf Of Marechiare Sent: 07 May 2010 10:34 To: Microcontroller discussion list - Public. Subject: Re: [EE]: enclosure cooling. > In my opinion there is only one Olin in this list please > do not try to emulate. Olin's doing great job on this list providing help in EE field. Any hints why don't you allow others "to emulate" that? Best Regards.

> ... I suggest we stop posting to each Sounds like a grand idea :-). 1. *** Suggestion only *** - it would seem useful to limit the content here mainly to the somewhat [EE] related technical subject matter and if people want to discuss the merits of ? hmm ... whatever the bickering is about ... ? to open an appropriate OT thread in parallel. That way people who enjoy the bickering thread can partake thereof while others get on with the technical stuff. 2. ENCLOSURE VOLUME: One value of using case volume to calculate air changes per second or seconds per air change is to get some idea of how "busy" the system is. Dealing entirely with cfm or litres/second or whatever tends to mask the sheer magnitude of what is being attempted relative to the size of the equipment involved. But, when you get enclosure air change times of well under a second it intuitively (to me anyway) indicates that something severely difficult is being attempted. If this was a living room you'd be experiencing a tornado. Tokamac transformers are made of sterner stuff than lounge suites, but that's still an impressive amount of air movement. A water cooled lounge suite sounds more likely to work. Also, increasing air changes per second implies increasing air velocity over equipment surfaces and thus probably also enhanced heat transfer. To get best effect with air cooling it is probably reasonably simple to greatly increase air velocities over the transformer winding surfaces and/or over the laminations (assuming iron core construction). This could be trialled with stuck on stiff cardboard or plastic sheet guides and air blocks. "Hot-melt" glue is a great tool for lashing up such trials and easily removed once trialing is complete - and if it melts in service you know it hasn't worked :-). Water : water copper pipe heat exchangers as used in eg shower heat recovery have ratings in the 1 - 10 kW range and are typically 'smaller than a bread box'. So a liquid cooler for this amount of energy seems quite doable. A few 10's of metres of soft copper tubing and a few hours playing may produce dividends. Start with total loss system from a water tap and consider water cooling once it shows promise. Russell

Moreira, Luis A wrote: > I get what you guys mean and you are right. The only thing that the > volume of the air in the enclosure makes a difference is on how long > it will take from a state of no power dissipation, for the > temperature in the enclosure to reach steady state temperature. Theoretically yes, because at a fixed flow volume the enclosure volume dictates the time between "air changes". However, in a practical system this will usually be much less than the time for the thing being cooled to heat up. I should point out something that Russell brought up and that I didn't emphasize enough in my original post. The black box view of the system being cooled is valid and useful, and I believe my example was correct. However, that says nothing about how that power is tranferred to the moving air and what temperature the thing being cooled needs to be to tranfer that power, and what temperature different sections of the internal airflow are. In my example, the outflow is at 40degC, but some parts of the internal flow could be much higher with others largely bypassing the hot thing and staying near 20C. The average outflow would still be 40C, or you could envision turbulance so that the outflow air is well mixed and it's all at 40C. Either way you still transfer the 683W, but there can be large differences in what that means to the equipment being cooled. Let's say for example you've got a common off the shelf heatsink for a TO-3 transistor that has a thermal resistance of 3degC/W (a plausible value for such a thing). If such a single heatsink had to dissipate 683W its temperature would have to rise by 2050C. Obviously that's no good. However, if the 683W are spread evenly over 50 such heatsinks, each only needs to rise 41C above the air around it. Designing the air flow system inside the cabinet so that each of these 50 heatsinks gets the proper flow of cool air is no trivial feat. The point is that just because 1 cubic foot per second and 20C rise works from the black box point of view doesn't guarantee that it does what you need inside, at least not without a lot of careful engineering. Also picture the mass of 50 TO-3 heatsinks that is about plausible to dissipate 680W to keep within semiconductor temperatures. That's a lot of heatsinks. If I remember right, the OP wanted over a order or magnitude more cooling. Imagine the very large surface required to exchange that amount of heat with air. That's why Russell suggested water cooling. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Moreira, Luis A wrote: > I have been going over the calculations you sent me and I don't > understand how you got 34g on your example. That's the mass of 1 cubic foot of air at 1 atm and 20degC. I used a somewhat round about way to derive it from numbers I already had. I should have just looked it up I suppose. I just did that and found that 34g per cubic foot is about right. Google "density of air" found a Wikapedia page that said it was 1.204Kg per cubic meter at 20C and 1 atm. Since there are 35.3 cubic feet in a cubic meter, that comes out to 34.1g per cubic foot. There is some variation depending on humidity, so the nice round 34g/foot**3 I used in the example was quite valid. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

>> ... I suggest we stop posting to each > > Sounds like a grand idea :-). > > 1. *** Suggestion only *** > - it would seem useful to limit > the content here mainly to the somewhat [EE] related > technical subject matter That's true, but what I wanted was only to find out what the magic equation they were talking about in their public posts. That's [EE]/[TECH] subject, I am sure. Could you take your admin hat on to help me find the goddam equation out? :-) If not, then, perhaps, it would not be a bad idea at all to increment the number of admins, to make the admin job a bit easy, some said they were willing to :-) Thanks.

On Fri, 7 May 2010 19:28:46 +0300, "Marechiare" said: > That's true, but what I wanted was only to find out what the magic > equation they were talking about in their public posts. That's > [EE]/[TECH] subject, I am sure. Could you take your admin hat on to > help me find the goddam equation out? :-) Oh, I'm more than happy to do the admin on this thread, but I'm not nearly as gentle as Russell when it comes to handling people. If all you were interested in was an equation you certainly have an interesting way of asking. As a suggestion, perhaps scoring hit points everywhere might be the wrong approach. The full archive of this thread is on piclist.com if you want to refamiliarize yourself with the different ways people have suggested modeling enclosure cooling - that might help you put together a reasonable question. But if all you want to be is argumentative, please change the tag to [OT] and you might get one or two more posts in before you get slammed by me and the other admins. Best regards, Bob -- http://www.fastmail.fm - Same, same, but different...

> If all you were interested in was an equation you certainly > have an interesting way of asking. So the question remains the same [EE]/[TECH]: I'd like to see the equation the OP talked in his post. I asked him and he gave me wrong direction pointing to his first post in the thread. ... > The full archive of this thread is on piclist.com if you want > to refamiliarize yourself with the different ways people have > suggested modeling enclosure cooling - that might help > you put together a reasonable question. That is not relevant to my question. > But if all you want to be is argumentative, please change > the tag to [OT] and you might get one or two more posts > in before you get slammed by me and the other admins. Please confirm that you meant offense by using word "slammed"; "slammed" for what - for asking to provide the equation and the definition of "Q" the OP was talking about? Thank you.

> Could you take your admin hat on to As noted, it was >> 1. *** Suggestion only *** And > That's true, but what I wanted was only to find out what the magic > equation they were talking about in their public posts. That's > [EE]/[TECH] subject, I am sure. I absolutely agree. Anything that relates technically is fine. But there was a very significant to and fro " ... your mother wears army boots ..." & "did so-did not-did so - ... " type exchange developing. Soon we all leap in in such things and the original material gets buried. Please do keep asking questions re technical matters which are unclear. Olin just posted something which will lead me to go an reassess a constant that has long been stored in my brain. To 2 significant figures I've used 1.3 kg/m^3 as air density for many years. He notes that Wikipedia says 1.2 kg/m^3 at STP. Time for me to check again . Hmmm - http://en.wikipedia.org/wiki/Density 1.204 at STP 1.293 at 0 C I've been using a 0 C figure mentally while probably thinking of it as a 20 C figure. Lets see 293/273 x 1.204 = 1.29... As expected. An extra 7% due to temperature alone. This is wrong (apparently) - probably a dyslexic typo 1.239 http://physics.kenyon.edu/EarlyApparatus/Pneumatics/Density_of_Air/Density_of_Air.html > Could you take your admin hat on to Presumably meant to read "admin hat off"..." ? No admin hat intended. I always mention said hat when it's worn. And "suggestion only" was meant to convey that it was only a suggestion :-). > help me find the goddam equation out? :-) Others are better able to comment on what they meant. My contribution so far has mainly been the, so far mainly uncommented on except by Olin, note that water cooling is overwhelmingly indicated here and various other bits noting that the difference between theory and practice may be greater in practice than in theory - especially so when the theory is incompletely applied. > If not, then, perhaps, it > would not be a bad idea at all to increment the number of admins, to > make the admin job a bit easy, some said they were willing to :-) Not that there was any admining happening. I need to be able to have a life of my own apart from the social stigma of being one of the despised ones. Comment allowed and entirely acceptable. But a smiley is not accepted in that context :-). Complaint about deficient admin action noted :-). > Thanks. In lieu of me answering for others - let me instead summarise my input so far: Just water cool it. Or, as noted, direct air tightly and at maximum achievable velocity over hot surfaces, block alternate unproductive air paths, add finning. Then, water cool it. Far easier long term. Russell

Hi Olin, Just to let you know that I manage to go trough your calculations, I understand where you got the 34g. My calculations were not wrong but the way I was thinking about it was. Thanks for making that clear. In the end if you are looking at the "black box" what matters is the volume of air you are shifting. Thank you also to Russell and all the other that helped. This was a very useful discussion as it actually made me see a few other things a bit more clearly. In relation to the problem I was trying to solve, when I looked at the setup further, I noticed that transformer as spacers in between the windings and that the only air intake into the enclosure is right under the transformer. The air actually flows through the middle of the windings to cool it down. Ideally I would like to cool the air coming in with a heat exchanger under the transformer, but I do not have enough space under it or budget to do so. Another thing I notice was that two of the legs of the transformer are right in the path of the fan with no restrictions but one of the legs had the air flow restricted due to a plate that someone fitted a few years back to limit access to the transformer and incoming supply. I have removed it. I decided at this point to fit a more powerful fan where the original was (double the flow rate), and fitted another fan on the other side to make sure the all transformer as a balanced airflow through it. When this PSU is back operational I will do a bit of temperature monitoring and then see if further cooling is required. Thank you all for the help. Best Regards Luis {Original Message removed}

> In relation to the problem I was trying to solve, when I looked at the > setup further, I noticed that transformer as spacers in between the > windings and that the only air intake into the enclosure is right under > the transformer. > The air actually flows through the middle of the windings to cool it > down. Ideally I would like to cool the air coming in with a heat > exchanger under the transformer, but I do not have enough space under it > or budget to do so. I had previously said: > To get best effect with air cooling it is probably reasonably simple > to greatly increase air velocities over the transformer winding > surfaces and/or over the laminations (assuming iron core > construction). This could be trialled with stuck on stiff cardboard or > plastic sheet guides and air blocks. "Hot-melt" glue is a great tool > for lashing up such trials and easily removed once trialing is > complete - and if it melts in service you know it hasn't worked :-). So, it sounds like the original designers took my advice :-). If their design still proves less than ideal there's still water cooling to consider. It sounds from the above like this may be quite easy to achieve. Thermally conductive tubes through spacers at a loose sliding fit with thermal grease of some kind and ... . For any tubes that cut flux lines you probably want to avoid closed electrical conducting paths in the cooling system plumbing or coolant. :-). Russell