Make a cool case out of PCB only

[eneuro]

so you can see that the temperature delta between the heatsink and copper trace and components is really small at most a few single digit °C above heatsink temperature.

teflon has ONLY 0.2W/mK, while Cu 400W/mK, Al 250W/mK, so I'm, still sceptical about better performance of thin a few oz copper layer with 0.1mm teflon insulation to aluminium heatsink than 8-10 fets in parallel while those fets has huge drens surface to help disipate heat-copper heatsink can be used soldered to.... copper pipes filled with oil like in home thermal heating systems, but ok maybe for mass production it is not the cheapiest solution to ship, but messing with teflon...no 

[electrodacus]

so you can see that the temperature delta between the heatsink and copper trace and components is really small at most a few single digit °C above heatsink temperature.

teflon has ONLY 0.2W/mK, while Cu 400W/mK, Al 250W/mK, so I'm, still sceptical about better performance of thin a few oz copper layer with 0.1mm teflon insulation to aluminium heatsink than 8-10 fets in parallel while those fets has huge drens surface to help disipate heat-copper heatsink can be used soldered to.... copper pipes filled with oil like in home thermal heating systems, but ok maybe for mass production it is not the cheapiest solution to ship, but messing with teflon...no 

Like I mentioned manufacturer spec for that PCB is at most 0.5°C cm^2/W and based on my observation seems to be real. It really is a big difference between a SO8 package with heatsink on top and a IPT007N06N mounted on a metal core PCB.
They use this metal core PCB for Power LED's up to hundred watt LED's that need to have the die at no more than 60C for long life.
Thermal resistance of the IPT007N06N is just 0.2K/W for junction to case.
The metal core PCB is also a really inexpensive solution compared to multilayer FR4 and heatsink (I have no invested interest in promoting this metal core PCB) but may point sharing the case design was to show what is possible with this solution.
Hope more people will take a closer look at the metal core PCB for their high power applications. 
Not sure what you want to say with "messing with Teflon".

[macboy]

Interesting, out of curiosity is pin that dissipates heat. Are they electrically connected to the aluminum or does it go though some kinda insulator in the pcb..?

Hi the way metal core PCB is build is like this. A 2.5mm Aluminium plate in my case then a 0.1mm layer of Teflon and then your copper layer so no electrical contact between your pads and aluminium plate but a really good thermal transfer do to that thin 100um layer of Teflon.
I hope more people to find about this and work with this.

PFOA is toxic if its heated above certain temperatures. Its also (I think) an endocrine disruptor

http://www.ncbi.nlm.nih.gov/pubmed/?term=pfoa

Sure but Teflon (PTFE) doesn't contain PFOA. It is used in the manufacture, but is almost completely absent in the final product.

[tszaboo]

To be honest I find interesting, that you would use such big mosfet in your design. As far as I understand you are not switching it at high frequency, because it works like a chopper, so gate capacitance and switching losses are less significant. In my designs, where I have to do similar tasks, I usually use multiple smaller fet in parallel. For that 6 dollar Infineon, you could get 8-10 Power SO8 FET, which can switch much higher current, you just have to make sure that you know how to parallel FETs. Not to mention, because you use higher quantities, you get a lot better discounts at the suppliers.

What is the RDSon on those 8 or 10 SO8 also there will be longer copper traces to connect them all and if you look at my replay above that is quite significant part of the power loss.
I use the FET as a switch mostly so I'm interested in the smallest loss so small RDSon.
I did had in my initial design (just prototype) 4 smaller FET that cost almost the same while they where just 40V not 60V
Yes that discount for larger number is a good thing but I do this in low volume anyway so saving even 1$ at 100 units will save me 100$ not even worth the manual placement of all those SO8 Not to mention I have no space for more than 4x SO8 in my design.

Say, you have 40mm space for the FETs, you can put 7 of those SO8 packages there, plus some spacing. IRFH7085 (conveniently selected for you) has about 2,3Ohm Rdson if you drive it with 15V. A pcb trace 20mm wide 40mm long with 35um plating has half a milliohm resistance, lets calculate with half of this, because the current distribution. That is 0,57mOhm DC resistance. But yes, I get that manual placement is not fun.
I see that you used the open hardware symbol. Is there schematic somewhere? This stuff is quite interesting, even that I dont have a solar system.

[electrodacus]

To be honest I find interesting, that you would use such big mosfet in your design. As far as I understand you are not switching it at high frequency, because it works like a chopper, so gate capacitance and switching losses are less significant. In my designs, where I have to do similar tasks, I usually use multiple smaller fet in parallel. For that 6 dollar Infineon, you could get 8-10 Power SO8 FET, which can switch much higher current, you just have to make sure that you know how to parallel FETs. Not to mention, because you use higher quantities, you get a lot better discounts at the suppliers.

What is the RDSon on those 8 or 10 SO8 also there will be longer copper traces to connect them all and if you look at my replay above that is quite significant part of the power loss.
I use the FET as a switch mostly so I'm interested in the smallest loss so small RDSon.
I did had in my initial design (just prototype) 4 smaller FET that cost almost the same while they where just 40V not 60V
Yes that discount for larger number is a good thing but I do this in low volume anyway so saving even 1$ at 100 units will save me 100$ not even worth the manual placement of all those SO8 Not to mention I have no space for more than 4x SO8 in my design.

Say, you have 40mm space for the FETs, you can put 7 of those SO8 packages there, plus some spacing. IRFH7085 (conveniently selected for you) has about 2,3Ohm Rdson if you drive it with 15V. A pcb trace 20mm wide 40mm long with 35um plating has half a milliohm resistance, lets calculate with half of this, because the current distribution. That is 0,57mOhm DC resistance. But yes, I get that manual placement is not fun.
I see that you used the open hardware symbol. Is there schematic somewhere? This stuff is quite interesting, even that I dont have a solar system.

All the space I have there is at most 30mm for FETs. I looked at your recommended FET http://www.irf.com/product-info/datasheets/data/irfh7085pbf.pdf that has 2.6mohm typical so say I can somehow fit 6 but probably realistic just 5 is possible to have some space between them.
So 2.6/5 is 0.52 mohm that is not big improvement over 0.66 mohm typical on the IPT007N06N and where is the cost gain that is still over 1$ each in quantities so it will be more expensive than IPT007N06N or at least there will be no gain. Sorry I looked at CAD $ still you need exactly 4 to replace one IPT007N06N that costs $4 US at 500pcs where IRFH7085 cost 0.76 X 4 = 3$ but more parts so you need a larger PCB to fit them that will add to cost also longer PCB trace that will negate the use of 4 or even 5 smaller FETs to replace the IPT007N06N
For the new design I have even less space is just enough to fit two parallel IPT007N06 on the output and of course much higher current 120A.
The entire SBMS100 will be using 6 of those IPT007N06N so if I need to replace with the smaller ones I will need at least 24 of them.
The SBMS 25 and SBMS40 will actually use something similar a smaller FET but that will need to be in a DPak case so that it can fit on the same footprint of the IPT007N06N  probably IPD034N06N3 or similar I will decide when I get closer to the components order.
Yes is open source the schematic for the current model is available you can find the link on youtube in the description of the SBMS4080 HW video or in the latest updates on the first kickstarter project just search for Solar BMS on Kickstarter.
Just noticed you are from Romania Salut.

[3roomlab]

i face similar problem when trying to squeeze alot of things into 5cmx5cm to make a PSU controller ... good Tjc component vs more smaller cheaper component. in the end the real estate problem is something so hard to get over when it comes to thermal dissipation  (that was where i started to explore mounting alot of SMD NMOS on heatsink and not on PCB ... https://www.eevblog.com/forum/beginners/smd-on-normal-heatsinks/)

i did see 1 of the tear down i think is shahriah on the keithley SMU parrallel SO8. but @ 80A ... i mean SO8 is so easy to overheat  . ... but at hobbyist level it would be a dream if its a it would never breakdown design

[electrodacus]

i face similar problem when trying to squeeze alot of things into 5cmx5cm to make a PSU controller ... good Tjc component vs more smaller cheaper component. in the end the real estate problem is something so hard to get over when it comes to thermal dissipation  (that was where i started to explore mounting alot of SMD NMOS on heatsink and not on PCB ... https://www.eevblog.com/forum/beginners/smd-on-normal-heatsinks/)

i did see 1 of the tear down i think is shahriah on the keithley SMU parrallel SO8. but @ 80A ... i mean SO8 is so easy to overheat  . ... but at hobbyist level it would be a dream if its a it would never breakdown design

metal core PCB is a really great use in this sort of applications but even my large 121 x 87mm x 2.5mm pcb can only handle around 5 to 6W as passive without getting over 50 to 60C (I never design things that go over 60C at full load)
But the nice thing on metal core PCB is that you can mount your device on a larger heatsink as large as you need to dissipate the power. In fact any flat metallic surface will do.
My new SBMS will be 110mm x 80mm a bit smaller than first and needs to handle 2x the power (TDP will be the same max 22W).
Is nice for shipping also new SBMS at around 340g so I can keep the parcel under 500g probably one of the smallest devices able to handle 2x 120A simultaneously for unlimited amount of time. Trick is that it needs an external heatsink or a a piece of flat aluminium plate. Is somehow similar to a modern CPU at 22W TDP will be one of those low power mobile versions.   

[tszaboo]

All the space I have there is at most 30mm for FETs. I looked at your recommended FET http://www.irf.com/product-info/datasheets/data/irfh7085pbf.pdf that has 2.6mohm typical so say I can somehow fit 6 but probably realistic just 5 is possible to have some space between them.
So 2.6/5 is 0.52 mohm that is not big improvement over 0.66 mohm typical on the IPT007N06N and where is the cost gain that is still over 1$ each in quantities so it will be more expensive than IPT007N06N or at least there will be no gain. Sorry I looked at CAD $ still you need exactly 4 to replace one IPT007N06N that costs $4 US at 500pcs where IRFH7085 cost 0.76 X 4 = 3$ but more parts so you need a larger PCB to fit them that will add to cost also longer PCB trace that will negate the use of 4 or even 5 smaller FETs to replace the IPT007N06N
For the new design I have even less space is just enough to fit two parallel IPT007N06 on the output and of course much higher current 120A.
The entire SBMS100 will be using 6 of those IPT007N06N so if I need to replace with the smaller ones I will need at least 24 of them.
The SBMS 25 and SBMS40 will actually use something similar a smaller FET but that will need to be in a DPak case so that it can fit on the same footprint of the IPT007N06N  probably IPD034N06N3 or similar I will decide when I get closer to the components order.
Yes is open source the schematic for the current model is available you can find the link on youtube in the description of the SBMS4080 HW video or in the latest updates on the first kickstarter project just search for Solar BMS on Kickstarter.
Just noticed you are from Romania Salut.

That seems to be justified in that case. Interestingly, I was working with smaller voltages, and I came to a different conclusion when I was making my design. I think the main reson behind this is the huge market of the DC-DC converters inside PCs. If you want to make a design which works at 12V, you have FETs with 3mOhm Rdson for 30 cents and you can parallel as many as you want, but if you increase the voltage, not so much anymore.
The MOSFETs on metal core is a very interesting design concept. I have designs which dissipate over 300W, with water cooled heatsinks, I always toyed with the idea of the IMS PCB but never got the boss to agree to launch a prototype.
There is very little talk on the web about high current FETs unless it is in SMPS, so I take any chance to talk about it.
BTW I'm living in Belgium, but I'm originally from Hungary, not Romania.

[3roomlab]

have you considered just using vias? lots of it? if say the thermal resistance could work ... so the PCB could be thin? maybe 0.4mm? i saw a page selling bare 0.3mm thick PCB, i didnt think there is such thin boards 

[eneuro]

...
Hope more people will take a closer look at the metal core PCB for their high power applications. 
...
Not sure what you want to say with "messing with Teflon".

For high  power apps experimenting rather with direct connecting TO-220  case to aluminium/copper heatsink by http://en.wikipedia.org/wiki/Brazing using short high energy pulses (spot welding) to melt filler metal 
Brazing is a metal-joining process whereby a filler metal is heated above melting point and distributed between two or more close-fitting parts by capillary action.

So, actually teflon might be really good in my method to seperate electrically coper case and aluminium/copper heatsink which will allow easy heat filler metal which has much higher resistance than copper or aluminum.
Additionally teflon will help limit brazing bondary to element case.

It looks like that even thin teflon tape 0.075mm~0.1mm should be  fine while there is no electrical contact between case and heatsink elements, also tried this tefon tape  in fire - it is not flammable 

Probably teflon used on those metal core PCBs is very different  than cheap teflon tape for pipes, but teflon might be this what I needed for spot welding brazing 

[electrodacus]

That seems to be justified in that case. Interestingly, I was working with smaller voltages, and I came to a different conclusion when I was making my design. I think the main reson behind this is the huge market of the DC-DC converters inside PCs. If you want to make a design which works at 12V, you have FETs with 3mOhm Rdson for 30 cents and you can parallel as many as you want, but if you increase the voltage, not so much anymore.
The MOSFETs on metal core is a very interesting design concept. I have designs which dissipate over 300W, with water cooled heatsinks, I always toyed with the idea of the IMS PCB but never got the boss to agree to launch a prototype.
There is very little talk on the web about high current FETs unless it is in SMPS, so I take any chance to talk about it.
BTW I'm living in Belgium, but I'm originally from Hungary, not Romania.

Where you use normal FR4 multiple parallel FETs make more sense since all that heat is better distributed. I prefer to keep the power dissipation as low as possible even if involves more expensive FETs since the additional heatsink or water cooling will probably be even more expensive.
Those metal core PCBs dropped in price significantly thanks to modern high power LED's.
Sorry for some reason that black stripe on the Belgium flag looked blue on my screen.   

[electrodacus]

have you considered just using vias? lots of it? if say the thermal resistance could work ... so the PCB could be thin? maybe 0.4mm? i saw a page selling bare 0.3mm thick PCB, i didnt think there is such thin boards 

Then you need some electrical isolator between the vias on the other side and a heatsink that will be nowhere near as good as that 0.1mm Teflon. It will probably also cost more. 

[electrodacus]

...
Hope more people will take a closer look at the metal core PCB for their high power applications. 
...
Not sure what you want to say with "messing with Teflon".

For high  power apps experimenting rather with direct connecting TO-220  case to aluminium/copper heatsink by http://en.wikipedia.org/wiki/Brazing using short high energy pulses (spot welding) to melt filler metal 
Brazing is a metal-joining process whereby a filler metal is heated above melting point and distributed between two or more close-fitting parts by capillary action.

So, actually teflon might be really good in my method to seperate electrically coper case and aluminium/copper heatsink which will allow easy heat filler metal which has much higher resistance than copper or aluminum.
Additionally teflon will help limit brazing bondary to element case.

It looks like that even thin teflon tape 0.075mm~0.1mm should be  fine while there is no electrical contact between case and heatsink elements, also tried this tefon tape  in fire - it is not flammable 

Probably teflon used on those metal core PCBs is very different  than cheap teflon tape for pipes, but teflon might be this what I needed for spot welding brazing 

I do not think that Teflon used on metal core PCB is any different. I just have no idea what sort of glue they use to stick that Teflon to copper and aluminium. You can also order metal core PCB with copper instead of aluminium but of course is more expensive.

[eneuro]

You can also order metal core PCB with copper instead of aluminium but of course is more expensive.

Those MCPCBs are copper on white surface due to this teflon layer on aluminium base like in this video below?

Probably need review this thread to see estimated pricing per square inch of such aluminium MCPCB, but maybe I will be able using spot welder brazing join a few mm aluminium/copper plates cut on CNC to classic cheap 1oz double layer PCB with help of teflon tape between PCB copper<->teflon contour<->aluminium/copper heatsink<->teflon contour<->element case and  additional 3mm in diameter rivets througth vias and TO-220 case to ensure mechanical stability at accidential high temperatures 


"May the Force be with us" 

Note: Teflon countour I mean there is no 0.1mm teflon isolation over whole joint area of course, but only contours required to keep electrical isolation between brazed elements in spot welder pulses, but let brazing fill metal most of contact area between element case and heatsink-in the case of copper to copper solder itself, copper to aluminium brazing with help of brazing past etc

[electrodacus]

Those MCPCBs are copper on white surface due to this teflon layer on aluminium base like in this video below?

That is a metal core PCB but is white instead of black do to paint colour of the solder mask you can not actually see the Teflon.

If you look on eBay you can get 50 pcs of those 16mm diameter metal core PCB's for LED with $12 so is like 25cent/pcs you can build your custom PCB's and if they are in sufficient volume you will get at about same price.
My board is quite a bit larger 121x87mm and the aluminium is thicker 2.5mm vs those usually 1.6mm or less for LED and mine in relatively low volume 100pcs was about $6.
I will say is a decent price and is hard to find a better or more cost effective solution for high power stuff.

That LED in the video is probably a 10W LED so TDP around 8W or over while keeping the LED die under 60C
 

[cdev]

Please be careful with teflon/PTFE, when heated beyond a few hundred degrees it produces really toxic fumes.

https://www.ncbi.nlm.nih.gov/pubmed/?term=ptfe