Best electrically conductive thermal paste ?

Doctorbass

Doctorbass

100 GW
Joined
Apr 8, 2007
Messages
7,496
Location
Quebec, Canada East
I am looking to install the low side FETs of my 36 fets controller direct on the heatsink bar without insulator and i am looking for the best ELECTRICAL and THERMAL conductive compound.

This time i can benefit of the great thermal conductibility of ELECTRICAL conductor material but i just found electical insulated material :(

The Artic Silver 5 is great.. but i think i can get better.. apparently the diamond powder or carbon powder based compound are good..


The Artic silver 5 is 7.5W/degree K / m.

Pure silver is 410.. so i guess there is a trade off between both right?

Any suggestions?

Doc
 
Soldering wins by mile if you need both heat AND electrical conductivity.


I made a really really badly filmed video showing how to do it without needing any special soldering equipment.

If you're bonding them to aluminum rather than copper, than you're going to need zinc solder paste and flux, but otherwise it all works the same.

[youtube]wOXEmoNJgvw[/youtube]
 
Hmm very cool! That looks to be a MUCH better way to do it than using the screws as I was planning... Do you see any issues with doing that if using TO-220's? That won't overheat them?
 
tostino said:
Hmm very cool! That looks to be a MUCH better way to do it than using the screws as I was planning... Do you see any issues with doing that if using TO-220's? That won't overheat them?
Click to expand...

The tabs are the drain, so you can only do what shares a common drain connection.
 
tostino said:
Hmm very cool! That looks to be a MUCH better way to do it than using the screws as I was planning... Do you see any issues with doing that if using TO-220's? That won't overheat them?
Click to expand...


No prob doing that with TO-220, I did that for the Gary BMS and also you see that often on many signalab ( PING) BMS

It work well.. I never noticed FET to fail due to that.
 
liveforphysics said:
Soldering wins by mile if you need both heat AND electrical conductivity.


I made a really really badly filmed video showing how to do it without needing any special soldering equipment.

If you're bonding them to aluminum rather than copper, than you're going to need zinc solder paste and flux, but otherwise it all works the same.

[youtube]wOXEmoNJgvw[/youtube]
Click to expand...


Ah Yes I saw that on this video:[youtube]fWxYK3VFysw[/youtube]

But the Does the FET will overheat due to the higher temp to melt the zinc based solder?

I would not weld them to the heatsink since that would be REALLY a mess to remove if they blow!

That's why i need best thermal paste electrical conductive or not.

Thanks for the exemple Luke.
 
I know you can only do that where the drains are electrically connected, otherwise it will short.

See my progress here if ya want: http://www.endless-sphere.com/forums/viewtopic.php?f=10&t=25063&start=15
 
Doctorbass said:
But the Does the FET will overheat due to the higher temp to melt the zinc based solder?

I would not weld them to the heatsink since that would be REALLY a mess to remove if they blow!

That's why i need best thermal paste electrical conductive or not.

Thanks for the exemple Luke.
Click to expand...


Wouldn't you just drill another bar if you need to replace the FETs?

The zinc solder paste I use, "solder-it aluminum" melts at about the same temp as normal solder, and lower temp than most lead-free crap.

http://www.amazon.com/gp/product/B003YJUT0A/ref=pd_lpo_k2_dp_sr_1?pf_rd_p=486539851&pf_rd_s=lpo-top-stripe-1&pf_rd_t=201&pf_rd_i=B000RB3ZCC&pf_rd_m=ATVPDKIKX0DER&pf_rd_r=01M9HB6RNE8NK1AWXYF0


Also, to avoid the problem, and improve performance by a little bit more, copper bar stock is dirt cheap to buy, and would only improve performance and allow conventional solders.

Then, if you do blow it and need to repair it, just heat the back of the bar with a heatgun or torch until all the burned fets fall off and start over.
 
liveforphysics said:
Doctorbass said:
But the Does the FET will overheat due to the higher temp to melt the zinc based solder?

I would not weld them to the heatsink since that would be REALLY a mess to remove if they blow!

That's why i need best thermal paste electrical conductive or not.

Thanks for the exemple Luke.
Click to expand...


Wouldn't you just drill another bar if you need to replace the FETs?

The zinc solder paste I use, "solder-it aluminum" melts at about the same temp as normal solder, and lower temp than most lead-free crap.

http://www.amazon.com/gp/product/B003YJUT0A/ref=pd_lpo_k2_dp_sr_1?pf_rd_p=486539851&pf_rd_s=lpo-top-stripe-1&pf_rd_t=201&pf_rd_i=B000RB3ZCC&pf_rd_m=ATVPDKIKX0DER&pf_rd_r=01M9HB6RNE8NK1AWXYF0


Also, to avoid the problem, and improve performance by a little bit more, copper bar stock is dirt cheap to buy, and would only improve performance and allow conventional solders.

Then, if you do blow it and need to repair it, just heat the back of the bar with a heatgun or torch until all the burned fets fall off and start over.
Click to expand...

Thanks for the advice Luke, But this time i thing you just dont imagine how complicated it would be to replace fets and resolder 18 fets on the bar with properly aligned legs!!

Usually you put the fet legs in their holes for each fet locations on the pcb.. than you fix them on the pcb...

Trying to solder 18 fet a time would take alot of patience and luck to not overheat something..

I'm not optimistic here, i just figure how it could be done easy!

Copper bar are alot heavier than aluminum for the ratio of added thermal conductibility.. and 36 fets fixed on 2 foot of copper bar would make it heavy. Unless i use it for motorcycle, or small car, i think that these kg saved are precious on an ebike. That's why i choosed aluminum and copper combo.

I might also make the bar smaller, it would make sense.. maybe on the second 36 fets controller i'll assemble ( i have two i plan to assemble. The second one will be a 75V version with more power.. probably using these IRF3006 fets

Doc
 
Well I got one of my messed up pieces of copper stock, and an old crappy fet that I tore out of a controller, my butane soldering iron (turned into a torch) and got to work.

I just wanted to see how easy it was. I just used my normal radioshack solder that was sitting in a pool of flux on the copper bar, heated it with the torch, dabbed the back of the fet in a little flux, and when I saw the solder wet to the bar, I placed the fet in place, pressed down, and removed the heat. It looks pretty crappy I know, but the fet was that bad looking from the start.

Was extremely easy to do.

I think i'll pre-drill all the holes for screws, and just use solder paste on the back of the fets, put them in place with the screws, line up the legs with the pcb, then get to soldering. Do ya think that will work alright?
 

Attachments

  • IMG_20110228_222710.jpg
    IMG_20110228_222710.jpg
    193.3 KB · Views: 2,745
I bookmarked one product from this forum a while back = http://www.mgchemicals.com/products/8331.html
 
I know these silver conductive glue or paste.. they are like 7.5W/mK

Compare with silver 429W/m k or copper 401W/mk

But i need it electricaly conductive too!.. I mean not partially..

Doc
 
I just bought some of this: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=230590572344#ht_1995wt_905

So I think that I need to get some higher temp solder for the tubing, so that when I heat up the bars with the tubing already soldered to it for the fets, it doesn't detach the tubing. Any recommendations?
 
Doctorbass said:
I know these silver conductive glue or paste.. they are like 7.5W/mK

Compare with silver 429W/m k or copper 401W/mk

But i need it electricaly conductive too!.. I mean not partially..

Doc
Click to expand...

I think that you will find that conductive glues are pretty crappy conductors of electricity. They will not handle the current. Your only choice is direct clamping (and taking the small thermal conductivity hit) or a proper solder job.

The thermal difference between a good clamp job and one with thermal compound is rather small (probably less than 4C). Also the difference between the worse thermal compound and the best (as has been measured in numerous tests) is maybe a couple of degrees C!
 
Doc,
You won't find a paste/compound that is a decent electrical conductor. It's best to just use thermal compound behind the FETs for the best cooling and take the current in/out the pins or off the front of the tab (if TO-220). Even the PGS graphite sheets and SMA-TIM alloys (soft metal alloy thermal interface material) don't have very good electrical conductivity through them (compared to Al or Cu).

Soldering is a great solution though...just use a hot plate. By using solder paste under the FETs and metal jig (just a bar/plate with FET-shaped holes) to hold the FETs in place on the bar, you can easily raise the temperature of the bar until the paste flows and solders the FETs on. A temperature controlled hot plate works best but an electrical "griddle" or even just putting the bar in a toaster oven or on a piece of metal on the stove can work too. An IR gun or temperature chalk stick works good here for measuring temperature. Or, just watch very, very carefully. :mrgreen:

Hmm...something to consider though is that a properly clamped FET will have no compound between the two metal surfaces, just filling the air spaces in between. Since the conductivity of the interface between the FET and your metal bar is only determined by the amount of metal that's touching, not the air spaces, the conductivity should not be affected by using compound. If, and only if, the clamping is perfect and stays that way though.

How much power is each FET going to be dissipating as heat? If only a few watts, or less, you can often just skip using a compound as texaspyro mentioned. For something like an electronic load, with up to a couple hundred watts of power being dissipated in each FET, this won't work. But, for your controller, it's worth considering.
 
CamLight said:
Doc,
You won't find a paste/compound that is a decent electrical conductor. It's best to just use thermal compound behind the FETs for the best cooling and take the current in/out the pins or off the front of the tab (if TO-220). Even the PGS graphite sheets and SMA-TIM alloys (soft metal alloy thermal interface material) don't have very good electrical conductivity through them (compared to Al or Cu).

Soldering is a great solution though...just use a hot plate. By using solder paste under the FETs and metal jig (just a bar/plate with FET-shaped holes) to hold the FETs in place on the bar, you can easily raise the temperature of the bar until the paste flows and solders the FETs on. A temperature controlled hot plate works best but an electrical "griddle" or even just putting the bar in a toaster oven or on a piece of metal on the stove can work too. An IR gun or temperature chalk stick works good here for measuring temperature. Or, just watch very, very carefully. :mrgreen:

Hmm...something to consider though is that a properly clamped FET will have no compound between the two metal surfaces, just filling the air spaces in between. Since the conductivity of the interface between the FET and your metal bar is only determined by the amount of metal that's touching, not the air spaces, the conductivity should not be affected by using compound. If, and only if, the clamping is perfect and stays that way though.

How much power is each FET going to be dissipating as heat? If only a few watts, or less, you can often just skip using a compound as texaspyro mentioned. For something like an electronic load, with up to a couple hundred watts of power being dissipated in each FET, this won't work. But, for your controller, it's worth considering.
Click to expand...


Thanks John.

What i want to do is use thermal compound behind the FETs for the best cooling and take the current in/out the pins.

But i know that good electrical conductor are also good thermal conductor... and that is why i was searching for that kind of paste. generally the most common thermal paste need to be non conductive.. so..

To solder the fet i would need copper bar.. and i thought about that too but this is really heavier than aluminum and for the volume i want to use i must consider that too.

I allready soldered fet on copper bar or copper on pcb and it worked well.

So if i would go anyway with copper bar and solder the fet on it and also use it as electrical bus for the (low side only) could it ameliorate the current sharing if i use FEt that are from a same batch?

Doc
 
Doc, if you look at how I am planning to do my controller, you can use the heat sink on the high side for current sharing too. You just then need to insulate the three high side heat sinks from each other electrically.

Not very hard to do.
 
Doc,
Ahhh, that's right, I forgot that you'd need to use copper for soldering and that you had already decided not to use it (if possible) due to its weight.

But, I'm confused.
If you're just taking the current in/out via the FET pins, you don't need any electrical conductivity in the thermal interface compound.

If soldering to a copper bar, and using it for current, then you can always make it quite thin (3 to 4mm?) to keep it lightweight and and just connect it (with a sheet of insulating silicone) to the thicker liquid-cooled aluminum bar for heat removal. You get the benefits of the copper's great heat-spreading and current handling and it's also easier to solder to a thinner piece of copper.

Using FETs from the same batch will help to keep the current sharing fairly equal but any temperature differences between the FETs will try to unbalance the current. And no matter how good the cooling is, the FETs in the center of the bar will run hotter than the ones on the ends, maybe a lot. This will only affect your total power handling a bit though as you have lots of FETs sharing the current and can always just measure the temperatures of the inner FETs to determine the controller's max power limits.
 
CamLight said:
Using FETs from the same batch will help to keep the current sharing fairly equal
Click to expand...

I have found that FETs from the same batch tend to have very well matched Rdson values, but their gate threshold voltages can still be quite different. For the 100 IRFP2907 FETs that I bought for some welders, the Rds value ranged from 0.2 to 200K ohms at the same gate drive voltage. At full on the Rds values were within around 10 micro ohms.

I always match my paralleled FETs for threshold voltage. I usually do this by driving the gate with a fixed voltage that is in the FETs linear region (say 3-4V on a 10 Vg FET) and matching the FETs for their Rds value. Not quite as good as putting them on a curve tracer, but I have found it works about as well.
 
Well Alot of good solution you give to me guys :wink:

John, I love the idea of copper first for conductibility and aluminum for heat with insulation between both.
That's a similar ides than tostino is making.

Tex, The matching of the fet with threshold voltage seem a good idea.. i'll see how i can do that in my lab.

Well I already spend 74$ of material on my actual copper and aluminum combo and hours of hand made aluminum pieces..

I'll see how i can patch my actual version to get these advantages without too complicated job..

Otherwise i'll go direct with two big fat copper bar and will forget about the Weight watcher program for it.

My friend with the CNC showed me some spare copper part of 1/4" thick and also 1/8" thick he have.. but he really hate to machine copper with his CNC.. i'll have to convince him.
Also.. I dont like making thread on copper! that remind me the nightmare job i had 4 years ago about making 32x 7mm deep thread for 0-80 !!!

Do you think that if i solder the fet on copper bar i'll still need to fix them with screw or holder?.. I would be surprised to see them unstick because of excessive heat since they would be water cooled...

Doc
 
I forgot to mention that i'll assemble probably two of these 36 fets controllers so i might try both methods.

Doc
 
Doctorbass said:
Do you think that if i solder the fet on copper bar i'll still need to fix them with screw or holder?.. I would be surprised to see them unstick because of excessive heat since they would be water cooled...

Doc
Click to expand...
No, IMHO you won't need to also mount them via screws.
They can come unsoldered at excessive power levels though as the thermal resistance from the FET case to the water cooling system is high enough that no matter how much water you pump, or how cold the water is, the system won't be able to remove enough heat if the power levels are high enough.

But, if they come unsoldered you've exceed their max operating temperature by so much that the controller requires a complete redesign in order to operate at those power levels. Or, at least, a much, MUCH lower power level. :mrgreen:
 
Doc, I really wouldn't worry too much about the weight on these things... if copper even allows you an extra few hundred watts because of better cooling, it would probably be worth it over the 1/2lb you may save with just aluminium. It's not like the controllers are a major source of the total weight to begin with!

But anyways, I am going to put screws through the bars pretty much solely to hold them in place while I solder them... unless I can think of an easier way. I don't think they will do anything at all once they are soldered.
 
You must log in or register to reply here.

Similar threads

Ianhill
Tab cooling prismatics.
2
Replies
31
Views
2,123
bobwind
B
CUDAcores89
Thermal paste on hub motor windings. Dumb idea or dumb idea?
Replies
10
Views
2,288
MadRhino
MadRhino
methods
Sevcon Gen 4 - Heat Sink Insulator Material?
Replies
2
Views
850
methods
methods
L
How far can I push 6xTO220 controller
Replies
10
Views
1,734
LuboN
L
LawrenceEaden
Single Gear Stealth Build - Battery in Frame Tubes
2 3 4
Replies
80
Views
14,070
memberseven
M
Back
Top