DIY led grow

stardustsailor

Well-Known Member
The array's thermal resistance ,as also the TIM's and the heatsink system's therm. resistances ,are the real ' bottle-neck ' in that kind of cooling system ....
(HSF: heatsink-fan )

Imagine this natural analog situation ....

There's a water source up high in a mountain range ...
That water source can provide a water flow rate of 1 million gallons per second !!!

Man goes and builts up a water control facility ....
Now the flow has dropped to 1 gallon per second ....

The pressure underneath the surface ,hundreds of yards ,deep below earth has increased to huge values /figures.

Above the earth ...The surrounding environment still remains vast ..
Under the mountain range ,still the same vast valley is there ...
But without that old ,big river ....

So ...
Instead of trying to alter / fit the vast environment ,in order to 'accomodate' the small river left ...
Better ,say to the staff of the control center ,to increase the flow ,as much as possible ...

Before the earth ,explodes ,under our feet ...
Something like that ...

(And before Pos has to move to S.Pole ....:P...)



-TIM +Heatsink/fan mods ...
Simple,enough ...
 
Last edited:

Positivity

Well-Known Member
First, thanks for the valuable input.

Just a few thoughts on the heatsink.

The cxas on the ends. Three in somewhat of a cluster. Very close to the ends.

I think the leds being close to the edges...versus centered...is one of my biggest problems.

The heat has nowhere to go! If centered...heat would be gradually pulled away. Being close to the edge..the heat has nowhere to go! So it builds up and starts heating up WAY WAY faster than the centered ones.

So in a design where leds are near edges of heatsink...fans should be placed directly over that area.

If my heatsink was a foot longer...it would be easy to keep them farther off the edges...plus have fans directly overhead. Certain it would cut temps in half at least...probably taking me where I'd prefer.

BUT!! I don't want to waste heatsink. The design works wonderful up to about 300w power...very very cool....telling me the design does work quite well. 400w is surely "testing" the limits of its capacity.

So what to do with this information acquired...

I tried arctic silver and didnt notice enough of a difference..minimal..minute. But prolimatech does sound like a good deal better...I'll try that. With doubts though that there is much of a bottleneck there.

Then there's the common sense side of me saying...130w of power right on the ends! Surely the heatsink isnt happy about that! Even if I make a shroud..evening out the airflow...speeding the airflow...STILL..large concentration of power at the ends with nowhere for heat to go. TOO DEPENDANT on fans!!

So...this could all be avoided with a bit larger heatsink I'm quite sure. 2' was minimum and I should of gave myself more real estate.

But I think this is workable...I'll work on it a bit and take everything said into account.
 

Bueno Time

Well-Known Member
I hear you on the clusters of COBs on the ends on the heatsink issue. But I think from seeing SDS couple last posts here that the TIM (and thickness of TIM as well as smoothness of heatsink surface) makes a very large difference in Tc of the COBs. The Arctic Silver has a significantly lower W/mK value than Prolimatech PK2 and PK3 bumps it up 1 more on the W/mK value.

Ive got PK2 on my COBs on the Alpine 11 coolers and just picked up a 5g tube of PK3 for $12. Id check out the PK3 personally and maybe try the idea someone posted about putting tape over the open sections of the heatsink fins between the drivers and whatnot on the top of the unit. No doubt having the clusters of COBs near the heatsink edge is going to limit the cooling capacity of the ends of the unit but it might help more than you think and could only take a matter of 2 minutes to put some tape on and see. That coupled with a better TIM with a higher W/mK value should make a significant improvement. You could also sand and polish the heatsink smoother to help the thermal path between the COBs and heatsink. And if you do all that and its still not enough maybe put a fan over each end cluster?

edit: actually went back and looked again quick and seems the TIM layer thickness makes a bigger difference than the W/mK of the TIM but still redoing the TIM with a higher W/mK value AND a lower thickness than the pads you are using now should help a significant amount if I am computing this correctly. (Its early for me and I am in a hurry to shower and run some errands.)
 
Last edited:

MrFlux

Well-Known Member
edit: actually went back and looked again quick and seems the TIM layer thickness makes a bigger difference than the W/mK of the TIM but still redoing the TIM with a higher W/mK value AND a lower thickness than the pads you are using now should help a significant amount if I am computing this correctly. (Its early for me and I am in a hurry to shower and run some errands.)
If I do the calculation with for example a Vero 29:
dT = P . dx / (k . A), with
dT = temperature gradient,
P = 50 W of heat dissipation,
dx = 5e-5 m average gap width,
k = thermal conductivity,
A = 7e-4 m2 contact area,

then dT = 3.6 / k, so the difference between k = 1 or k = 10 is just a couple of degrees.
 

stardustsailor

Well-Known Member
If I do the calculation with for example a Vero 29:
dT = P . dx / (k . A), with
dT = temperature gradient,
P = 50 W of heat dissipation,
dx = 5e-5 m average gap width,
k = thermal conductivity,
A = 7e-4 m2 contact area,

then dT = 3.6 / k, so the difference between k = 1 or k = 10 is just a couple of degrees.

Yes ...
(Altough you're still using a CXA array ...
mech dim vero 29.JPG
Surface area of Vero 29 is (( 0,0492 / 2 ) ^2 ) * 3,14 = 0,0019 m^2 .....

And ΔΤ then ,would be 1.31 / k .....
As you can understand ,k (thermal conductivity value ),plays a crucial role ,
when the rest provide a high 'sum' to be divided by k ..

For example if P=100 W ..
Then it would have been (for the same vero29 ) ....

ΔΤ= 2,62 / k ...And so on....

And Mr.Flux do not neglect the fact that
" just a couple of degrees " of ΔΤ...
Through a 50 micron layer ..
Is not so 'small' or neligible difference ....

A 1W/mK would show ΔΤ= 1.31 C ,between two sides of less-than-paper thin layer ....
A 10 W/mK would show ΔΤ= 0,131 C ,between the same two sides of the same less-than-paper thin layer ....

Which is not a small figure for the Τc of the led array ,dissipating 50Watts ..
Coming from just 5e-5 m thick layer ...

In this example...
" Rtim with paste = 0.00005 /( 10.2 * 0.000748 ) = 0.006 C/W "

With k=1 Δτ= ~6C ,while with k=10 Δτ= ~0.6C...

When dissipating 100 W of heat ...

~5.4 C difference ,from a super thin layer of paste A vs paste B,
between the CXA and the heatsink ,
is way much.............. (for my standards at least ...)
 
Last edited:

MrFlux

Well-Known Member
When dissipating 100 W of heat ...

~5.4 C difference ,from a super thin layer of paste A vs paste B
Sure if you drive the COB that hard then the TIM becomes important. What I take away from all this is that for a normal running COB the TIM doesn't matter much just keep the gap small.

This all reminds me so much of CPU overclocking, ppl would pay $$$ for a few degrees less. Meanwhile Intel uses a thick layer of cheap goo inside their integrated heat spreaders.
 

stardustsailor

Well-Known Member
Sure if you drive the COB that hard then the TIM becomes important. What I take away from all this is that for a normal running COB the TIM doesn't matter much just keep the gap small.

This all reminds me so much of CPU overclocking, ppl would pay $$$ for a few degrees less. Meanwhile Intel uses a thick layer of cheap goo inside their integrated heat spreaders.
So true ,LOL !!!
 

stardustsailor

Well-Known Member
Sure if you drive the COB that hard then the TIM becomes important. What I take away from all this is that for a normal running COB the TIM doesn't matter much just keep the gap small.

This all reminds me so much of CPU overclocking, ppl would pay $$$ for a few degrees less. Meanwhile Intel uses a thick layer of cheap goo inside their integrated heat spreaders.
Well ,it is not actually fair,of you to say so ....

1) firstly the price differences,amongst thermal pastes , are not three or even two digit figures ...

2)
While ,its true that amongst various thermal pastes,if thickness is maintained in the 50-100 micron range ,
then k doesn't seem so crucial ...
Things change quite a lot ,between a thermal paste and a pad ....
An average quality (and cheap) thermal paste ,is as efficient as a top-quality (and expensive) thermal pad ...
Still ,price difference might ligger in a maximum of two-digit figure ..

3) Even so ...Still a 5C difference does not appeal to me ...
Might be neligible ...But I'd rather have my CXAs operate at 5C less ,
if I'm to spend just few bucks more for it ,to happen ....

5 dt mere.JPG

Why "lose " those extra 83 lumens ,@1.4 A ?
I'd preferred a good quality thermal paste ,over a good quality thermal pad ,
for those 83 lumens ...

(I hope I do not sound like those "overclocking freaks" ..Do I ? )

Cheers.
:peace:
 

churchhaze

Well-Known Member
I think we're actually still on the same page here. I definitely think thermal grease is a must here, as well as active cooling. I agree with everything you suggest to do.

Here's the problem. (I think I need to clarify a bit of what I meant).

I agree that it's a futile effort to bring Ta down from what you normally have it (room temperature).

The problem is that if you don't vent the hot air, Ta will approach Tj (assuming a closed system with no heat losses).

The closer Ta gets to Tj, the lower the heat transfer will be through the heatsink to the air.

This difference (Tj-Ta) is linearly proportional to the heat transferred by the sink. If it is 0, heat transfer will be 0, and Tj and Ta will both rise until catastrophic failure. (obviously there's no such thing as a closed system on earth).

Basically what I was trying to say is that if let Ta rise, your heat sink solution will become irrelevant. You need to make sure Ta is somewhat equalized with your outside environment.



" If your ambient temperature is higher, you should expect the sinks to function poorly. "

Neither if Ta = +35 C ,neither if Ta -35 C ,the heatsink's function is affected ....
...
The air convecting heatsink's (+TIM layer ,+ Fans included ) 'function' ,
is / can be affected by plenty of other facts ,but certainly not from the Ta ...

Cheers.
:peace:
 
Last edited:

stardustsailor

Well-Known Member
I think we're actually still on the same page here. I definitely think thermal grease is a must here, as well as active cooling. I agree with everything you suggest to do.

Here's the problem. (I think I need to clarify a bit of what I meant).

I agree that it's a futile effort to bring Ta down from what you normally have it (room temperature).

The problem is that if you don't vent the hot air, Ta will approach Tj (assuming a closed system with no heat losses).

The closer Ta gets to Tj, the lower the heat transfer will be through the heatsink to the air.

This difference (Tj-Ta) is linearly proportional to the heat transferred by the sink. If it is 0, heat transfer will be 0, and Tj and Ta will both rise until catastrophic failure. (obviously there's no such thing as a closed system on earth).

Basically what I was trying to say is that if let Ta rise, your heat sink solution will become irrelevant. You need to make sure Ta is somewhat equalized with your outside environment.
Yes,you are totally right..
We're on the same page and your thoughts are correct and based on solid facts basis .

Still ,mj has a temperature range ,in which it can grow/thrive ..( 15- 35 C )
As long as Ta is maintained between the "limits" ,
dealing directly with TIM,heatsink,fans and array placement ,
it remains as the most efficient way to
keep those CXAs running relatively 'cool' ....

Yes...The way you've put it ,yes ,then Ta can "stop" totally the "convection" of a HSF cooling system ...
But ,in order for such thing to happen , the Ta will be way beyond the mj's growing temp range ...:P

Cheers.
:peace:(:
 

churchhaze

Well-Known Member
I think I just had that stoner moment... as if the first problem noticed wouldn't be wilting plants...
:eyesmoke:

And that's why you don't run your grow room at 200C folks! Your lights will burn out quicker!

I almost forgot what we were doing with the leds.

This project is for lighting your backyard deck, right?

Yes,you are totally right..
We're on the same page and your thoughts are correct and based on solid facts basis .

Still ,mj has a temperature range ,in which it can grow/thrive ..( 15- 35 C )
As long as Ta is maintained between the "limits" ,
dealing directly with TIM,heatsink,fans and array placement ,
it remains as the most efficient way to
keep those CXAs running relatively 'cool' ....

Yes...The way you've put it ,yes ,then Ta can "stop" totally the "convection" of a HSF cooling system ...
But ,in order for such thing to happen , the Ta will be way beyond the mj's growing temp range ...:P

Cheers.
:peace:(:
 

PetFlora

Well-Known Member
My BML bars get pretty hot.

I positioned a 12" household fan half way between the bars and the canopy

The bars are now just warm and the canopy benefits from the circulation
 

only1realhigh

Well-Known Member
Not sure about the C conversion, but I have had temps. down as low as 35 degrees F and no plant damage, yet I have had temps up to 95 F with no damage as long as air was moving around well and I provided the needed water that they require when that hot.
I have understood that LED's will last longer and perform better if kept in a specific temp range, same with the driver's, so that is what I worry about as I am sure the plants will thrive like always for me.

I get lost in all SDS's abbreviations, so I had to commit about something I know ( if no one minds)
 

Positivity

Well-Known Member
Alrighty then

Haven't had much time to work on things..pretty much rebuilt everything. Still will be a bit before I use it..taking a break for now.

Will do more thermal testing to see what the changes I did achieved. Made a top plate...added two fans..centered middle cxas...drilled/tapped screws for the small stars..applied prolimatec pk3 to everything. Seems to run a whole lot cooler but I'll wait till I hang it and take some measurements before I say success.

Also took two leds off of my driver string that was up to 204w..so now it's at about 199-200w. My other driver string is at 194w. May add one more uv led to that string to even out the uv exposure in a 2 x 4 area. That should put both drivers right at the recommended ratings.

I kinda like my naked design...especially for simple setups. But this one got more involved at 400w on 2' of heatsink. So I'll be looking at cases..found a few potential ones. There's a few features I'd like to add in the future and since I like going higher power on one fixture the extra space for everything will be important if not essential

image.jpg image.jpg image.jpg image.jpg image.jpg image.jpg
 

Positivity

Well-Known Member
Thinking of another way to optimize my setup. Flower room is cxa..veg room is xml/luxeon...now I think I'll put my cree globe bulb in the house somewhere.

Make room for a cob testing micro area. May have to venture into scrog, never done it before. Thinking 1 3590 2700k at 1.05a. But to be run at 350ma using a resistor and pot. Passive cooled..maybe a computer fan off to the side blowing at the sink for circulation/heatsink cooling.

So 142 l/w passive cooled dissipating 24w. Something like that..

Just a fun side project. Will be helpful for testing different options that become available without a huge setup and cost.

image.jpg image.jpg
 
Top