Author Topic: DC load using a CPU cooler  (Read 57259 times)

0 Members and 1 Guest are viewing this topic.

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16385
  • Country: za
Re: DC load using a CPU cooler
« Reply #50 on: September 20, 2014, 02:19:55 pm »
My bad, I was thinking of regular CPU heatsinks, that have a flat bottom that is actually large enough to have the 3 rows of 3 TO247 devices on the bottom. You might have the one edge not in contact on the outside, but the surface would do 9 tightly packed. Those copper rods are going to make poor contact, so you probably want the 3mm copper plate, machine it smooth on the faces and use the thermal epoxy to hold it and the devices together. You will not get good contact onto those flattened pipes without thermal epoxy or a good thermpad.
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #51 on: September 20, 2014, 02:24:14 pm »
Can one get pre-machined copper plates? I don't have access to a CNC mill
 

Offline rob77

  • Super Contributor
  • ***
  • Posts: 2087
  • Country: sk
Re: DC load using a CPU cooler
« Reply #52 on: September 20, 2014, 02:29:58 pm »
Can one get pre-machined copper plates? I don't have access to a CNC mill

you don't need a CNC mill to flatten the surface ;) people were flattening metal surfaces on milling machines before computer control was invented ;) i bet you have a tons of small businesses or even individuals with a milling machine in your area... for few bucks they'll flatten the block for you.
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16385
  • Country: za
Re: DC load using a CPU cooler
« Reply #53 on: September 20, 2014, 02:31:28 pm »
Cut to size, use a file on the edges and make flat with a flat surface and sandpaper. Flat surface can be a surface block, a granite kitchen top or a sheet of thick glass.
 

Offline timb

  • Super Contributor
  • ***
  • Posts: 2536
  • Country: us
  • Pretentiously Posting Polysyllabic Prose
    • timb.us
Re: DC load using a CPU cooler
« Reply #54 on: September 20, 2014, 02:36:51 pm »
As luck would have it, I just got a nice mill! I could make some copper pads for you, it would be good practice!


Sent from my Smartphone
Any sufficiently advanced technology is indistinguishable from magic; e.g., Cheez Whiz, Hot Dogs and RF.
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #55 on: September 20, 2014, 02:43:32 pm »
@timb Sounds great! What's shipping like to the UK?


@rob77, there will be OVP/OCP in hardware as well as a thermal cutoff on the heatsink. I'm using an INA196 for current sensing, so I'll use that to drive a comparator switching the output relay.
I could get access to a CNC mill at my local makespace... but they don't accept members under 18 (I'm 15). I don't know anyone else with access to a mill, but I could probably find someone on the forum if need be.
 

Offline timb

  • Super Contributor
  • ***
  • Posts: 2536
  • Country: us
  • Pretentiously Posting Polysyllabic Prose
    • timb.us
Re: DC load using a CPU cooler
« Reply #56 on: September 20, 2014, 02:48:45 pm »
Shipping's not bad, I'll cover it. Let me find the raw mats locally and you can just chip in a few bucks towards it if you want?

Wow, you're 15? Man I am going to load your package up with some random goodies I've got laying around. I've been waiting for a good opportunity to send them to someone who can use them!


Sent from my Smartphone
Any sufficiently advanced technology is indistinguishable from magic; e.g., Cheez Whiz, Hot Dogs and RF.
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #57 on: September 20, 2014, 03:18:53 pm »
Thanks! PM sent.
 

Offline Jeroen3

  • Super Contributor
  • ***
  • Posts: 4209
  • Country: nl
  • Embedded Engineer
    • jeroen3.nl
Re: DC load using a CPU cooler
« Reply #58 on: September 20, 2014, 03:36:21 pm »
Intrigued by this topic, I went looking for a way to mount the common TO-220 mosfets (or a bts141) in a low profile way such that pc heatsinks (cpu or northbridge) will fit.
I've come up with this:

This also leaves space for isolated TIM material, which is very important.
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #59 on: September 20, 2014, 05:05:44 pm »
Quote
A back brace made from thick insulating material with cutouts for the lead stubs will probably be needed in any case, or a thinner one with an aluminum backing plate to give even pressure.


If it works, I'm all for it.
« Last Edit: September 20, 2014, 05:07:58 pm by microbug »
 

Offline Christopher

  • Frequent Contributor
  • **
  • Posts: 429
  • Country: gb
Re: DC load using a CPU cooler
« Reply #60 on: September 20, 2014, 07:44:39 pm »
Why not attach the fet to the heatsink, bend the pins up thru the board?


Some tall-ish spacers on the heatsink for some thermal insulation?
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #61 on: September 21, 2014, 08:35:41 am »
OK, I've found my MOSFETs: IXTH40N50L2. They're designed for linear operation so they're ideal for this. Two of them will be fine for 300W operation, I think I'll have a max voltage limit of 100V. They're TO-247, so they'll fit nicely under the cooler as previously discussed. There were a couple of alternatives from the same company, but that's the cheapest one I can get from Farnell UK which has a minimum order cost of £20 rather than £50 like everyone else.


On the subject of cooling, as SeanB said, the copper heat pipes won't make great contact without a copper plate underneath them (thermal epoxy again). What about the back side of the MOSFETs (between them and the board)? Options as I see them:
- Cutout on the board, glue the MOSFETs to the copper plate;
- No cutout on the board, glue upside-down MOSFETs to the copper plate;
- No cutout on the board, MOSFETs sandwiched between back brace made from insulating material (e.g., FR4) and copper plate.


I'm currently investigating hardware over-voltage / over-current protection (watching Dave's video on multimeter input protection seems like a start).
 

Offline Spikee

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: nl
Re: DC load using a CPU cooler
« Reply #62 on: September 21, 2014, 08:56:54 am »
OK, I've found my MOSFETs: IXTH40N50L2. They're designed for linear operation so they're ideal for this. Two of them will be fine for 300W operation, I think I'll have a max voltage limit of 100V. They're TO-247, so they'll fit nicely under the cooler as previously discussed. There were a couple of alternatives from the same company, but that's the cheapest one I can get from Farnell UK which has a minimum order cost of £20 rather than £50 like everyone else.


On the subject of cooling, as SeanB said, the copper heat pipes won't make great contact without a copper plate underneath them (thermal epoxy again). What about the back side of the MOSFETs (between them and the board)? Options as I see them:
- Cutout on the board, glue the MOSFETs to the copper plate;
- No cutout on the board, glue upside-down MOSFETs to the copper plate;
- No cutout on the board, MOSFETs sandwiched between back brace made from insulating material (e.g., FR4) and copper plate.


I'm currently investigating hardware over-voltage / over-current protection (watching Dave's video on multimeter input protection seems like a start).

get the 75v version and contact IXYS. They like these kind of projects and should have no problem sampling 2-4 mosfets for free.
I have used the arctic silver adhesive before and it works reasonably. To get it off again just soak it in ipa and you can twist the mosfets off the heatsink.
Freelance electronics design service, Small batch assembly, Firmware / WEB / APP development. In Shenzhen China
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #63 on: September 21, 2014, 09:06:38 am »
I assume you mean IXTP80N075L2 (TO220)?


EDIT: Yes, they look suitable. They have a significantly lower Rds(on). Given the extended SOA, two would be fine for 300W. Should I get two in the TO247 package or 3/4 in TO220?


EDIT 2: So far for input protection, I'm thinking of a fuse for each MOSFET, reverse protection diode, MOV(s) and a thermal cutoff on the heatsink. There will be some protection in software too (shut off output relay in overload condition).
« Last Edit: September 21, 2014, 09:18:52 am by microbug »
 

Offline timb

  • Super Contributor
  • ***
  • Posts: 2536
  • Country: us
  • Pretentiously Posting Polysyllabic Prose
    • timb.us
Re: DC load using a CPU cooler
« Reply #64 on: September 21, 2014, 10:25:48 am »
I ordered some 5mm copper today. You're using the A30, right microbug? I'll grab one off Amazon and make something very slick for mounting. I've got a really good idea. (I'll draw it up tonight and post it here.)


Sent from my Smartphone
Any sufficiently advanced technology is indistinguishable from magic; e.g., Cheez Whiz, Hot Dogs and RF.
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #65 on: September 21, 2014, 10:40:20 am »
Yes, it's the A30. I'll probably have two TO-247 packages under it (the linear-optimised MOSFETs from the previous post are available in TO-247 and I think two TO-247s will be better than 3 TO-220s. Thanks timb!
 

Offline Spikee

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: nl
Re: DC load using a CPU cooler
« Reply #66 on: September 21, 2014, 11:27:59 am »
I have soldered the initial prototype but i soon found out there is a short somewhere on the board.
I was unable to locate the problem  :-\ .  I have started populating a second board and assembling and testing between each step to find the culprit. The power segment seems to function without any issues.

Edit:
This one IXTH75N10L2
http://ixdev.ixys.com/DataSheet/DS100200(IXTH-TT75N10L2).pdf
« Last Edit: September 21, 2014, 11:29:45 am by Spikee »
Freelance electronics design service, Small batch assembly, Firmware / WEB / APP development. In Shenzhen China
 

Offline timb

  • Super Contributor
  • ***
  • Posts: 2536
  • Country: us
  • Pretentiously Posting Polysyllabic Prose
    • timb.us
DC load using a CPU cooler
« Reply #67 on: September 21, 2014, 11:30:18 am »
Micro, what sort of driver are you looking at?

Edit: I mean OpAmps to drive the MOSFETs. Sorry, have been working on a SMPS tonight and have MOSFETs drivers on the brain!

Sent from my Smartphone
« Last Edit: September 21, 2014, 11:35:10 am by timb »
Any sufficiently advanced technology is indistinguishable from magic; e.g., Cheez Whiz, Hot Dogs and RF.
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #68 on: September 21, 2014, 04:15:14 pm »
Tim, it's an OPA4188 in a typical constant current configuration (100mOhm sense resistor). The current implementation is based on this thread, but I will probably change it to a more 'standard' configuration (input to the non-inverting terminal, shunt voltage to inverting) to make it easier to do a hardware constant voltage mode. I'm working on the schematic at the moment and I'll post a screenshot of the power stage soon.

Thanks for pointing out the right FET, Spikee. That one looks like it could (in theory) do 300W at the right current on its own, but I'll use two of them in the design (separate op-amp for each) to account for all the graphs other than the SOA in the datasheet that I didn't bother to read  ;) . I'll send an email to IXYS to ask for 3 (in case I blow one) and post the result.


EDIT: Spikee, that FET looks 100V rated to me ;)


EDIT 2: Uh Oh! I just looked up stock of IXTH75N10L2 on Octopart having just sent a grovelling email to the IXYS sales rep office in the UK, and it turns out neither Mouser nor Digikey have the part in stock (it's a 'factory direct' part)! I hope that IXYS has some internal stock or something... I suppose it wouldn't be too hard to find another suitable MOSFET.
« Last Edit: September 21, 2014, 05:42:22 pm by microbug »
 

Offline Spikee

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: nl
Re: DC load using a CPU cooler
« Reply #69 on: September 21, 2014, 07:10:18 pm »
IXYS has it and will send you a few free because it is not in stock at their distributors.
Freelance electronics design service, Small batch assembly, Firmware / WEB / APP development. In Shenzhen China
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
Re: DC load using a CPU cooler
« Reply #70 on: September 21, 2014, 07:11:07 pm »
Oh, good!
 

Offline Kevin.D

  • Frequent Contributor
  • **
  • Posts: 290
  • Country: england
Re: DC load using a CPU cooler
« Reply #71 on: September 21, 2014, 08:02:19 pm »
You will get about max 150W from 2 of those if your Heatsink has a Rth of about .25 K/W (allowing NFet junction temp to reach ~130 C) .
150W is plenty  anyway  to test most things you will come across ( a typical bench supply 30V@5 A =150W) .Why do you need 300W ? thats alot of power .

p.s I have made a 150 Watt one and if I ever needed to extend the power range then it's easy enough to make a power range extender .(Say 4 * 50W 1 ohm resistors fixed on to the side of a can you can fill with water ) .
You could then put these resisitors  in series with your eload and the DUT so that they dissipate 50% of the total power and your eload drops the other half  .(i.e the resitors drop ~ half the total voltage ,( thats why youd have say 4* 1 ohm resistors permanetly fixed fixed to your water can so then you could wire them to have the exact required  resistance you needed to drop 50% of the total dut voltage . (i.e with 4 *1 ohm resistors  you can have 1/4 ,1/2,1 ,2,4 ohm  values by wiring them in various ser/par combinations.)
 
« Last Edit: September 21, 2014, 08:10:12 pm by Kevin.D »
 

Offline microbugTopic starter

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: gb
  • Electronics Enthusiast
DC load using a CPU cooler
« Reply #72 on: September 21, 2014, 08:35:45 pm »
The heatsink is rated at 320W power dissipation. Unfortunately, it doesn't specify efficiency in K/W, but I'd expect better performance than that! The SOA covers 300W operation easily; I'd expect some thermal derating, but that's why there will be two. I chose a 300W limit because the extra 20W of heatsink capacity gives a little headroom should things go wrong.

Anyone else care to weigh in?
 

Offline Kevin.D

  • Frequent Contributor
  • **
  • Posts: 290
  • Country: england
Re: DC load using a CPU cooler
« Reply #73 on: September 21, 2014, 09:36:40 pm »
Well I dont know about the heatsink you have chosen (does it's spec give a max temp at 320W dissipation ? if so then to get it's Rth is just   (temp@320W-25)/320.
I used 2 standard amd K6 sinks+fans  on mine they each had a Rth of .45 K/W each when I measured them .

Forget about these soa graphs they dont contain much usefull info for dc operation(they simply reflect the  Rth j-c if you look),manufacturers use them to produce nice banner spec headlines on there product's like '300W dissipation ' .They  produce them by emersing the whole transistors in a cooling fluid to hold the case temp at 25 and then measure junction temp rise wiht power dissipation .
 You can't do that of course  you have to cool your transitors in a conventional fasion .
So you have to take into account the extra thermal impedance of the Case to Sink(Rth c-s) an also + any insulating material you have between case tab and sink (which can add another ~ .2 C/W to thermal resistance ).

So  looking  at your data sheet for to264 case style we  have
Rth j-c   .28 C/W      (notice if we only had this one Rth we could get the SOA graph)
Rth c-s  .2 C/W + another .2 C/W if we use tab insulation  .
total Rth j-s =  ~ .7 C/W
now your thermal calculation is easy .lets say you have 150W power into that ,so that would be
150 * .7 = 105 C temp rise at fet junction above whatever your heatsink temp is at 150W dissipation .
since I dont know what the  sink you using is rated at , lets guess and say it's rated at ~ .25 C/W
so at 150W this will reach a temp of  = 150*.25 =37 + 25 ambient temp =62 C

So with a single fet at 150W your junction temp will reach 62+ 105 = 167 C (to HOT)
with two fets the j-s temp rise would be 105/2 = 52 C
so now your junction temp at 150W is 62+52 =114  C (so at least it's now it'sunder your 150 C absolute max for junction temp)  .

DONT take to much notice of manufacturers headline grabbing specs and soa graphs done under  liquid cooling conditions ,  Just gather your various Rth's and do the simple calculations yourself.
 

Offline Spikee

  • Frequent Contributor
  • **
  • Posts: 568
  • Country: nl
Re: DC load using a CPU cooler
« Reply #74 on: September 21, 2014, 11:06:24 pm »
I tested two ixys75L2 mosfets on a arctic freezer 7 pro (150w max rated). and i was able to get the full wattage. The A30 heatsink should be twice as good. Around 0,01 - 0,005deg c/w should be possible.
Freelance electronics design service, Small batch assembly, Firmware / WEB / APP development. In Shenzhen China
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf