HobbyKing.com New Products Flash Sale
Reply
Thread Tools
Old Sep 22, 2009, 01:03 PM
Registered User
United States, AL
Joined Apr 2007
830 Posts
dpero: I determined with the higher-power mosfets and water cooling that each mosfet can dissipate 250w, so only two are needed for the goal of 500w. If you control each mosfet by itself(one op-amp per mosfet) you eliminate the need for ballast resistors(the large 0.1 ohm resistors), help ensure the load is perfectly balanced between the mosfets, and simplify the device. Additionally, the 100ohm resistors on each mosfet become unecessary.

As for your schematic, at quick glance it is incorrect and won't work as you have it drawn. Check your wiring from pins 1 and 2 on the op-amp up to the shunt and mosfet gate. Don't mean to be rude, but if you can delete or correct it please do so that others won't see it and attempt to build it without reading the thread first.
biskit is offline Find More Posts by biskit
Reply With Quote
Sign up now
to remove ads between posts
Old Sep 22, 2009, 01:20 PM
Registered User
NY
Joined Feb 2007
21 Posts
Thanks for the explanation. Makes sense.

I corrected the schematic. Does it seem correct now? I'll move myself onto your drawing once I get the original straight in my head.

Derrick
dpero is offline Find More Posts by dpero
Reply With Quote
Old Sep 22, 2009, 02:06 PM
Registered User
United States, AL
Joined Apr 2007
830 Posts
At a quick glance it looks correct. Again, the 100ohm and 0.1 ohm resistors are not necessary unless you are paralleling multiple mosfets. The "mv meter" is a bit archaic and I wouldn't even build this thing unless you had at least a wattmeter to ensure you don't damage your batteries.

I didn't get it at first either since I'm stupid, but john explained it. Basically you set a reference voltage at pin 3 of the op-amp via the potentiometer. The op-amp adjusts the mosfet to make the voltage it sees at pin 2(ie. the current) match pin 3. The resistors and capacitors are just there to dampen oscillations and make it all work smoothly. The 9 volt supply needs to be smooth and constant, if it changes the current setting will change. That's why there's a voltage regulator chip to keep it stable. In other words, don't use a battery to power it unless you put a regulator in the device and use a battery that will never drop below 11v.

My scribbly picture is the same thing but re-arranged. The op-amp is an 8 pin chip with the pins as shown in my picture. You can see how easy it would be to make a small pcb to fit all the components and make it very compact and neat.
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 23, 2009, 09:13 AM
Registered User
NY
Joined Feb 2007
21 Posts
Thanks for the explanation, again. I know it was said before but there was a lot of conversation going on and I didn't pick up on it.

Any thoughts about a hall effect sensor for sensing current?

Derrick
dpero is offline Find More Posts by dpero
Reply With Quote
Old Sep 23, 2009, 11:50 AM
Registered User
Dan Baldwin's Avatar
United States, CA, Norwalk
Joined Apr 2004
2,715 Posts
A hall effect sensor for sensing current would work fine, although the schematic would be different since you will be dealing with very different voltages, and the the filter networks (resistors and capacitors) would probably be different too.

Dan
Dan Baldwin is offline Find More Posts by Dan Baldwin
Reply With Quote
Old Sep 23, 2009, 05:21 PM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Quote:
Originally Posted by dpero
Any thoughts about a hall effect sensor for sensing current?

Derrick
Though there are some advantages to using Hall Effect sensors, typically they're a lot more expensive, less linear, and less accurate than sense resistors.

For an electronic load we don't have to worry about any power losses in the sense resistors since all we want to do is turn all the battery power into heat and not deliver it somewhere for use.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 23, 2009, 09:09 PM
Registered User
NY
Joined Feb 2007
21 Posts
Quote:
Originally Posted by JohnMuchow
Though there are some advantages to using Hall Effect sensors, typically they're a lot more expensive, less linear, and less accurate than sense resistors.

For an electronic load we don't have to worry about any power losses in the sense resistors since all we want to do is turn all the battery power into heat and not deliver it somewhere for use.
My first thought was for the nice 0 - 5V output which I can input to a micro controller. But, your absolutely right. Will have to educate myself on shunts.
dpero is offline Find More Posts by dpero
Reply With Quote
Old Sep 23, 2009, 09:56 PM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
You can often size the shunt to give you a very usable voltage range for the current it's handling. Especially for an electronic load where we don't care about power loss in the shunt. For a lot of applications, it's not any harder than trying to find a Hall Effect sensor that has a current range close to the one you're operating with.

But, having said that, I use Allegro Hall Effect sensors quite often. Especially when I need the isolation.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 25, 2009, 06:45 PM
Registered User
United States, AL
Joined Apr 2007
830 Posts
I got the heat sink block from Dan today and tested it out. I ran a small mosfet (the "IPP" ones) at 200w for a minute or so steady state no problem. The block was cool, but the top of the mosfet was too hot to touch. It was obviously being stressed, but this is a stress test so I turned up the power and the mosfet melted immediately. You can see the block in the picture.

I attribute it to the longer heat path on this large block. The shortest path from the mosfet to the water is about 1cm. The water was 24c out of the tap and flowing fast enough that there was no measurable temperature increase in the water. The tiny plate was only about 1mm thick so the heat had much less distance to travel and thus the tiny plate kept the mosfet cooler.

Solution? Can't make the heat sink from copper(too expensive) and I can't make the tap water colder, so in order to keep the mosfet cooler we need to reduce the distance the heat has to travel. I'm going to try to find a place to shave 1/2 inch or so off the top of the block to put the mosfet much closer to the water and try again. The larger "fda" mosfets are significantly better at conducting heat into the heat sink, so they can probably get 250w each with no changes. Hopefully shaving the block will give them enough wiggle room to be reliable. If you make your own block, get the water passages as close to the heated surface as possible!
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 25, 2009, 07:57 PM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Congratulations on the test! It's cool to see this coming together.

That IPP048N06L FET was being severely overstressed at 200W and would have blown soon enough. With a theta-js (junction to sink thermal resistance) of 1.0C/W, at 200W it was running at (200W) x (1.0C/W) = 200C rise above ambient, or about 225C!!!

Using the FDA75N28 FET, at 250W, you'll have a (250W) x (0.48C/W theta-js) = 120C temp rise. At 30C ambient the junction will be at 30C + 120C = 150C. This is the max temp for the FET.

That all sounds good but it depends entirely on the block directly behind the MOSFET being held at 30C or lower. At those power levels, no matter how much water is flowing through the block, I think it's going to rise above 30C (directly behind the FET). I recommend a 200W limit (perhaps 150W), all depending on temp measurements of the block. Using copper won't really help as we still have to deal with the thermal resistances between the junction of the FET and the heat sink.

If you can get a good reading inside a hole drilled diagonally into the block so it's behind the FET, that would be an OK way to measure the temp. Fill the hole with thermal compound and wait a few minutes for it to heat up (with the thermocouple inside the hole so it heats up too) before measuring. We can then easily calculate the junction temp for different power levels, water volume, water temperature, etc.

Otherwise, just hook up a FET for 24 hours and see what happens. If it doesn't blow up, raise the power another 25W-50W and wait another 24 hours. When it blows, back off 50W and you'll have your rating. The temperature will still be wayyyyyyy too high for long-term reliable operation but I think most users will gladly trade 50W-100W of power for an occasional FET replacement.

Be sure to fuse the unit though as MOSFETs fail as short circuits and if the sense resistor doesn't blow out, that battery pack is toast.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 25, 2009, 09:22 PM
Registered User
United States, AL
Joined Apr 2007
830 Posts
The IPP spec sheet only said 0.5 for "junction-case" while the FDA had "junction-case" and "case-heatsink" which both added up to 0.5... is the second value just not listed for the IPP? Are the two values typically similar and that's why you use 1.0 for the IPP?

The copper would help as it has more than double the thermal conductivity of aluminum, so it would conduct more heat away to other parts of the block and keep the mosfet slightly cooler. Moot point anyway since a similar block of copper would probably be $50 commercially. The block surface within a finger's width of the mosfet was too hot to touch as well, so the mosfet was obviously waaaay past what it was rated for. The water has to get closer to the mosfet!

The mosfet did melt into almost a short. it was glowing for a few seconds and a tiny flame started before I unplugged the battery. Thankfully it was an a123 pack so it didn't even know the difference! It was only a few dozen amps, but it could be bad if you were testing a tiny battery or the short was.. shorter. I think the small-ish wire I used helped keep the short-circuit current down too. All that red wire is tefzel, 18g I think.

So it sounds like as-is the two FDA mosfets will be on the verge of self-destruction. Shaving the block to get the water closer will help quite a bit. I think the tap water here is also unusually warm(unless you live in FL or CA maybe?) at 25c. There's room for huge improvement in the block, but it would quickly turn into a serious machining project rather than be a block with two holes drilled in it. Multiple small diameter passages only a couple mm away from the surface to keep the distance down and water velocity up. But then you need lots of tubing or some sort of manifold passage and it gets impractical fast. I thought about a pump and ice water, but that's not very practical(or cheap once you buy the pump).

So next I'm going to try to get the block shaved down and melt another mosfet. I have a 350w power supply so it'll rack up the hours fast! Honestly finding a place here to cut the block down will be the hard part.

I'm thinking two ways: two mosfet/two ap-amp or three paralleled mosfets on a single op-amp. The 1-mosfet-per-op-amp obviously will work no problem, maybe I should test the paralleled version and people can build whichever best suits their needs? If there was a pcb available to just drop the little components onto it would be trivial to make a 4-mosfet/4-opamp and have 600w very safely! The tiny components would be an extra $1 instead of $$$ for ballasts and other extra parts for paralleling. PCB making is far beyond my knowledge and time available limits though.
biskit is offline Find More Posts by biskit
Last edited by biskit; Sep 25, 2009 at 09:34 PM.
Reply With Quote
Old Sep 26, 2009, 12:17 AM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Quote:
Originally Posted by biskit
The IPP spec sheet only said 0.5 for "junction-case" while the FDA had "junction-case" and "case-heatsink" which both added up to 0.5... is the second value just not listed for the IPP? Are the two values typically similar and that's why you use 1.0 for the IPP?
A typical theta-cs (case-to-sink thermal res.) for a TO-220 FET is about 0.5C/W. Some manufacturers don't list it as it's almost a standard. You can look at other manufacturer's data sheets or even other data sheets for the same manufacturer for typical TO-220 theta-cs values.

With the spec'd theta-js of 0.5 and the "standard" theta-cs, you get about 1.0C/W for theta-js for the IPP FET.

Quote:
Originally Posted by biskit
The copper would help as it has more than double the thermal conductivity of aluminum, so it would conduct more heat away to other parts of the block and keep the mosfet slightly cooler. Moot point anyway since a similar block of copper would probably be $50 commercially. The block surface within a finger's width of the mosfet was too hot to touch as well, so the mosfet was obviously waaaay past what it was rated for. The water has to get closer to the mosfet!
I agree.
But, the copper won't help keep the FET cooler than aluminum.

Since the theta-js is the same no matter what material you use for the sink, if the "ambient" (or in this case, heat sink) temperature is above 30C, then the FET has to be derated as it will rise to the same temp no matter what the material is. You have a very small surface area for the water to "work on" so the ability of the copper to move the heat won't be taken advantage of. The heat will just "sit there" waiting to be taken away by the water, letting the heat sink temperature rise. Copper won't hurt, but its higher cost (as you mentioned), extra weight, and how hard it is to machine (compared to aluminum) makes aluminum a better solution IMHO. A copper heat spreader can help improve the heat sink's effectiveness by having the heat more equally spread over the entire block, but only by a few percent, at most. Especially if you're already using more than one FET.

Also, copper does conduct heat better than aluminum but aluminum's specific heat is three times that of copper. This will take some of copper's conductivity advantage away in this application as we can't remove the heat as quickly as it's generated and the heat sink will heat up.

Thermal resistance, conductivity, specific heat, and all that stuff is often very non-intuitive...to say the least! And, unfortunately, there's a stunning amount of useless info on the web from heat sink and cooler manufacturers.

Quote:
Originally Posted by biskit
The mosfet did melt into almost a short. it was glowing for a few seconds and a tiny flame started before I unplugged the battery. Thankfully it was an a123 pack so it didn't even know the difference! It was only a few dozen amps, but it could be bad if you were testing a tiny battery or the short was.. shorter. I think the small-ish wire I used helped keep the short-circuit current down too. All that red wire is tefzel, 18g I think.

So it sounds like as-is the two FDA mosfets will be on the verge of self-destruction. Shaving the block to get the water closer will help quite a bit. I think the tap water here is also unusually warm(unless you live in FL or CA maybe?) at 25c. There's room for huge improvement in the block, but it would quickly turn into a serious machining project rather than be a block with two holes drilled in it. Multiple small diameter passages only a couple mm away from the surface to keep the distance down and water velocity up. But then you need lots of tubing or some sort of manifold passage and it gets impractical fast. I thought about a pump and ice water, but that's not very practical(or cheap once you buy the pump).

So next I'm going to try to get the block shaved down and melt another mosfet. I have a 350w power supply so it'll rack up the hours fast! Honestly finding a place here to cut the block down will be the hard part.
I agree, try shaving down the block first. But, you're going to need a way to measure the MOSFET's temperature as the finger method just isn't going to help much once you're past the "ouch" point. And we definitely want to operate these FETs at above that point to get the most out of them.

Quote:
Originally Posted by biskit
I'm thinking two ways: two mosfet/two ap-amp or three paralleled mosfets on a single op-amp. The 1-mosfet-per-op-amp obviously will work no problem, maybe I should test the paralleled version and people can build whichever best suits their needs? If there was a pcb available to just drop the little components onto it would be trivial to make a 4-mosfet/4-opamp and have 600w very safely! The tiny components would be an extra $1 instead of $$$ for ballasts and other extra parts for paralleling. PCB making is far beyond my knowledge and time available limits though.
For a few of the PCBs, I'd be happy to lay out one and arrange to get it manufactured.
I'll need specific model numbers for every single component you're using though as the PCB needs to have the correct solder pad spacing. I have recommendations for the capacitors and low power resistors but the other stuff we'll need to talk about when the design is finalized.

I think you have the right ideas about the circuit's configuration. Your 4-FET/4-amp solution is the best solution by far. Having 4 FETs helps spread out the heat to all parts of the heat sinks so it's more efficient. Up to 15% more than having one hot spot in the center (from a single FET or group of FETs). And one amp per FET is completely stable. Having two paralleled FETs per op-amp, four FETs total might work well but FETs hate being paralleled when used as a load and lots of testing would have to be done. And, IMHO, it's not worth the dollar or two in savings.

For any lurkers....
Each time you double the number of FETs, you halve the theta-js resistance. If you use the IPP FETs, four FETs gives you a theta-js total of 0.25C/W. Assuming a max allowable 150C temp rise (150C + 25C ambient = 175C = max temp for that FET), you get (150C) / (0.25) = 600W rating. This assumes that the heat sink DIRECTLY under the FETs stays at 25C. It won't, but you can see how much of a difference there is when paralleling FETs to reduce thermal resistance.

Using the FDA FETs will give you even better numbers and I strongly feel is worth the extra cost.

To save money, the user could only install 2 FETs and the dual op-amp chip but the cost difference is so low, and the work involved is so little between the different versions, that each person might as well do a 4-FET version as it will make the best of whatever heat sink and cooling they select.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 26, 2009, 12:27 AM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Quote:
Originally Posted by biskit
So it sounds like as-is the two FDA mosfets will be on the verge of self-destruction. Shaving the block to get the water closer will help quite a bit. I think the tap water here is also unusually warm(unless you live in FL or CA maybe?) at 25c. There's room for huge improvement in the block, but it would quickly turn into a serious machining project rather than be a block with two holes drilled in it. Multiple small diameter passages only a couple mm away from the surface to keep the distance down and water velocity up. But then you need lots of tubing or some sort of manifold passage and it gets impractical fast. I thought about a pump and ice water, but that's not very practical(or cheap once you buy the pump).
I agree.

I think this was mentioned earlier....
Have you looked into using a small water-tight aluminum utility box, perhaps 3"L x 5"W x 2"H? They're 5-sided boxes with a gasketed cover made from 0.1" or so thick aluminum. If you mount the FETs to the bottom of the box get to use the 5 sides of aluminum for heat-spreading and the interior space of the box gives you a large surface area to remove lots of heat with the water. You might have to epoxy a hunk of aluminum beneath the FETs for heat spreading in really high power applications but that's easy to do for those users who need max power levels.

With an inlet hose on one side of the box and an outlet hose on the other end, offset from each other, the water will swirl around very well inside the box an help to remove more heat. Epoxy just about anything to the inside middle of the cover and now the water has to swirl around that too.

The boxes are only a few dollars, are widely available, and the user only needs to drill holes in 0.1" aluminum. No machining needed. Gasketed nuts will be needed on the interior of the box for each FET being mounted, but that's easy. Or, just use thermal epoxy to mount the FETs (with LOTS of pressure until the epoxy cures) and you don't even need to drill for the screws! Just drill the inlet/outlet holes.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 26, 2009, 04:23 AM
Registered User
United States, AL
Joined Apr 2007
830 Posts
Quote:
Originally Posted by JohnMuchow
But, the copper won't help keep the FET cooler than aluminum..............Also, copper does conduct heat better than aluminum but aluminum's specific heat is three times that of copper. This will take some of copper's conductivity advantage away in this application as we can't remove the heat as quickly as it's generated and the heat sink will heat up.
Once you're operating steady-state the specific heat doesn't matter because the temperature isn't changing. Think of the heat as pressure and thermal conductivity as the size of a pipe. More conductivity is a bigger pipe which has less pressure drop and so lower pressure for the same flow, or lower temperature for the same heat flux.

Quote:
Originally Posted by JohnMuchow
For a few of the PCBs, I'd be happy to lay out one and arrange to get it manufactured.
I'll need specific model numbers for every single component you're using though as the PCB needs to have the correct solder pad spacing. I have recommendations for the capacitors and low power resistors but the other stuff we'll need to talk about when the design is finalized.
So does the cost of a pcb depend only on the size? Or does the complexity also effect it? What would a pcb cost in such small quantities? I have to admit I'm a bit disappointed at the lack of interest in this thread! I was hoping at least a few people would build this silly thing right away.

Quote:
Originally Posted by JohnMuchow
I think you have the right ideas about the circuit's configuration. Your 4-FET/4-amp solution is the best solution by far........... To save money, the user could only install 2 FETs......
If a pcb can be made for a couple $ then it's easy too. If not then it does get a bit messy trying to piece it together on a generic radio shack board. If a pcb can be made then it'd be easy to build it with 1, 2, 3, or 4 fets for whatever power level you want.
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 26, 2009, 06:21 AM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Quote:
Originally Posted by biskit
Once you're operating steady-state the specific heat doesn't matter because the temperature isn't changing. Think of the heat as pressure and thermal conductivity as the size of a pipe. More conductivity is a bigger pipe which has less pressure drop and so lower pressure for the same flow, or lower temperature for the same heat flux.
Good point. I agree that the specific heat differences between Al and Cu won't matter once the system comes to thermal equilibrium.

But, I'm not sure if the better conductivity of the copper will matter here as there would be so little of it touching the water.

Quote:
Originally Posted by biskit
So does the cost of a pcb depend only on the size? Or does the complexity also effect it? What would a pcb cost in such small quantities? I have to admit I'm a bit disappointed at the lack of interest in this thread! I was hoping at least a few people would build this silly thing right away.
Not to be a downer, but...
Being a developer of resistor and electronic loads, I can tell you that the market for them is very, very, VERY small. And, you're competing against the CBA and its great ability to both make measurements and plot the results. Most users won't trade that away for more power.

Having the unit act as a module to increase the power of a CBA might be worth considering (what I do with my CC-400 load). But, you'll have to decide how you want it to work with the CBA and how the user's data logging and display will be affected by the addition of your load.

The quantity of boards ordered is the biggest determinant of cost. After that, the size and number of layers are probably the next biggest. After that, surface plating, PCB material, number of holes, and special solder mask colors or silkscreen colors have an effect too.

For example, let's assume the boards are standard 2-layer, 2oz. copper, 2 week delivery from the company I've used for my products...
100pcs of 1" x 2" = $375
100pcs of 2" x 2" = $400
100pcs of 4" x 2" = $460

Not much difference based on size.

But, here's what happens when we vary quantity...
100pcs of 2" x 2" = $400
50pcs of 2" x 2" = $370
25pcs of 2" x 2" = $350
10pcs of 2" x 2" = $330
5pcs of 2" x 2" = $315

A huge difference!
For quantities below 50-100pcs, I recommend getting a quote from one of the companies that batches prototype (i.e., low quantity) orders together and sends them to China for manufacture. It can often take 3 weeks or more to get the boards back but it will be a LOT less expensive than using a regular board house for the work. They just have too much overhead involved in setting up the job, tooling, etc., for each order to lower the price.

Quote:
Originally Posted by biskit
If a pcb can be made for a couple $ then it's easy too. If not then it does get a bit messy trying to piece it together on a generic radio shack board. If a pcb can be made then it'd be easy to build it with 1, 2, 3, or 4 fets for whatever power level you want.
I agree. This thing screams out for a nice PCB to allow it to be put together in a few minutes.

But, can you sell enough boards to make a 100pc., or larger, order worthwhile? Let's say, at least 50pcs?
Or would it be best to just write up a great document for any other DIY'ers to use if they want to roll their own? You can include your prototype development experience, the thermal data, parts list, parts sources, schematic and PCB files (PDF and Gerber format).

Personally, I don't know.

But, if you need a few boards, and can find a few other users who might need a few boards, this can work.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Reply


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Discussion PIC PWM and FETs as constant-current discharger? meteor DIY Electronics 21 Jan 09, 2010 10:17 AM
Mini-HowTo How to build a constant current battery discharger biskit Batteries and Chargers 25 Sep 06, 2009 07:53 PM
how to build a vaccum forming box? rayc Scale Kit/Scratch Built 14 Feb 28, 2008 10:46 AM
Discussion DIY analog constant-current discharger? meteor DIY Electronics 7 Jan 06, 2006 10:09 AM
Anbody know how to build a prop? Pork Tornado Parkflyers 4 Aug 03, 2001 04:10 PM