HobbyKing.com New Products Flash Sale
Reply
Thread Tools
Old Sep 10, 2009, 09:18 PM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
A higher ballast resistor value leads to a more stable multiple-MOSFET setup (when directly paralleling the MOSFETs) as higher values force better current-sharing between the MOSFETs.

Personally, I wouldn't go below 0.1 ohms. The only way you'll be able to know if you can go below that is to parallel a lot of MOSFETs (as many as possible to get better data) and test at different resistor values. You'll need to measure the voltage drop across each MOSFET (source-to-drain) and see how much it varies from MOSFET to MOSFET. If it shows a big difference (i.e., it shows a big current difference between which means lousy current sharing) then you'll need to increase the resistor value and retest.

There are some application notes from the various MOSFET manufacturers that give equations for roughly calculating minimum ballast resistor values, but they're a tough read. IIRC, the voltage drop across the ballast resistors has to be equal to or greater than the possible differences in voltage drops between the MOSFETs. But, I'm not at all sure of this and hopefully someone else can chime in.

Be careful when interpreting MOSFET power level specs!
On the data sheet you'll see that the 300W rating is only applicable if the case temperature (Tc) is held to 25C or lower. This is impossible with a real world heat sink and is only used as a spec for very, very short term pulses or to compare MOSFETs. As soon as the case of the MOSFET starts to heat up, that 300W rating goes out the window. In industry, TO-220 cases are typically used at the 50W level for reliability. I've used them to 100W with better cooling. Perhaps 150W for your setup?

The key to getting higher power levels without overheating the MOSFET (assuming you don't exceed any other specs) is the MOSFET's thermal resistance. Both the junction-to-case and case-to-sink resistances are added together to give you an idea of how efficiently the MOSFETs can move the heat from the junction to the heat sink.

Small case sizes like TO-220 aren't very good at moving out the heat (with combined junction-to-sink resistances of 0.9C/W to 1.5C/W). Larger MOSFETs in TO-247 and TO-264 cases have much lower thermal resistances (down to 0.33C/W) but are typically more expensive. Some times the higher-power, but more expensive, MOSFET is a better choice though as you require fewer ballast resistors, fewer mounting holes, less wiring, etc. You'll have to compare/contrast the different MOSFETs to see what it's better to stick with, more TO-220's or fewer TO-247/TO-264's.

Another warning!
For increased reliability, be sure that the MOSFET you select has a DC plot line in the Safe Operating Area (SOA) graph! This plot line means the MOSFET is rated for use as a load as long as you don't go past any part of that load line. If the MOSFET doesn't have a DC load line (just lines for shorter on-times like 10mS, 100uS, 10uSec, etc), then it's susceptible to "thermal runaway". This means it can burn out even when used within all of its other ratings, particularly at higher temperatures and battery voltages. And when it burns out, it burns out as a short-circuit. Very bad for the battery.

But, there are lots of people out there using MOSFETs not rated for use as a load. Some have had MOSFETs burn out, some haven't. You'll have to weigh the risk.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Sign up now
to remove ads between posts
Old Sep 10, 2009, 10:43 PM
Registered User
United States, AL
Joined Apr 2007
832 Posts
It sure does get complicated fast.

One more thing(I sound like Steve Jobs, don't I!): How about a HUF75652G3 mosfet? It's rated for 515w. Could I run one on each of the op-amps on a dual op-amp chip? Feed the same shunt voltage and same potentiometer signal to each op-amp and the only imbalance should be from the tolerances of the parts, right? No ballast resistors so $10 instead of $20+ and far fewer parts. I figure I'm getting over 2/3 of rated power from the to220 case, so 2/3 of 515 is over 300w each and easily meets our goals with only two. Any comments? I couldn't find the dc line you mentioned in any of the graphs for it. Failing that one anybody got a better choice for $10 or less each? I really like the idea of getting rid of the ballasts and having so few parts.

Edit: actually I think you should have one shunt per mosfet in this case? That would require a change in the resistor that determines the total adjustment range also right? If so then this could be a 100a/50v/600w discharger, maybe even 750w. And total parts cost would be $30!
biskit is offline Find More Posts by biskit
Last edited by biskit; Sep 10, 2009 at 11:08 PM.
Reply With Quote
Old Sep 11, 2009, 12:12 AM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Quote:
Originally Posted by biskit
It sure does get complicated fast.

One more thing(I sound like Steve Jobs, don't I!): How about a HUF75652G3 mosfet? It's rated for 515w. Could I run one on each of the op-amps on a dual op-amp chip? Feed the same shunt voltage and same potentiometer signal to each op-amp and the only imbalance should be from the tolerances of the parts, right? No ballast resistors so $10 instead of $20+ and far fewer parts.
Exactly!
Running each MOSFET individually can often seem to be a more expensive solution but sometimes is isn't.

But, you'll need a sense resistor for each MOSFET since you'll be controlling each MOSFET with its own op-amp. While this adds a bit of cost to the load you gain the benefit of individual control of each MOSFET (if desired), no worries about current sharing (this is a HUGE benefit!), and, you're correct, the only imbalances are due to tolerances. In fact, since you have individual control, you can use another pot to balance the two "channels" to be the same. IMHO, that's not needed though,

Quote:
Originally Posted by biskit
I figure I'm getting over 2/3 of rated power from the to220 case, so 2/3 of 515 is over 300w each and easily meets our goals with only two. Any comments? I couldn't find the dc line you mentioned in any of the graphs for it. Failing that one anybody got a better choice for $10 or less each? I really like the idea of getting rid of the ballasts and having so few parts.
Forget about power ratings for the MOSFETs. It's a useless spec for this application as it's only for a theoretical infinite heat sink or very short pulse applications. The important specs are the power the MOSFET has to dissipate and the thermal resistances. This will tell you how hot any MOSFET will get for the power you want it to handle. Then you only need to worry about the MOSFET's max temperature spec.

The HUF75652G3 is pretty nice. But, as you noted, it's not rated for use as an electronic load. Also, its max temp rating is 150C. Many MOSFETs are rated for 175C and this is a big jump. You can either run those 175C MOSFETs at higher power levels or at the same power level (as the 150C ones) and have a safety margin. If the thermal resistances are the same, that is.

Most TO-247 cased FETs have superior thermal resistances to TO-220 case FETs. Run the equations for the IPP048N06L and the HUF75652G3 and see what temperatures you get for the possible power levels you want to run at. Compare and contrast the possibility of using more, typically less expensive, TO-220 devices vs. fewer TO-247 devices. Be sure to take into account the max temperature ratings and whether each device is rated for use as a load.

Here are the equations:

Max Allowed Junction Temp Rise = (Max Junction Temp Rating - Derating) - (Max Ambient Temp)

Derating = 0% - 50%, depending on how close to the maximum you want to operate and how much unknown variance there might be in the cooling capabilities for each user or the ambient temperature. Some folks have "cold" water that's a lot warmer/cooler than others.

Junction-to-Sink Thermal Resistance (theta-js) = (Junction-to-Case Thermal Resistance) + (Case-to-Sink Thermal Resistance)

Power in Watts = (Max Allowed Junction Temp Rise) / (theta-js)

Actual Max Operating Junction Temp = (Ambient Temp) + (Max Allowed Junction Temp Rise)

For my loads, with a 175C MOSFET max temp rating, I use a 20% derating. This means that my max junction temp drops to 140C. Assuming a 30C ambient temp, I can have a 140C - 30C = 110C rise in the MOSFET's junction temp before it goes beyond my desired max temp.

To calculate the max power I can dissipate in that MOSFET, I do the following:

Power = (110C) / (0.5C/W), where theta-js = 0.5 for the MOSFET
Power = 220W, max I can have that MOSFET handle before it reaches 140C.

All this assumes that you have an "ideal" heat sink. This means that LOTS of water needs to be user. As soon as the heat sink goes above room temp (actually, as soon as the MOSFET case goes above room temp), then you can't have that much power flowing through the MOSFET.

There is a way to easily calculate the actual MOSFET junction temp when using a heat sink but we need to know the sink-to-ambient thermal resistance (theta-sa). This is a manufacturer's rating for commercial heat sinks. But, I can't even begin the guess the theta-sa for your setup other than "very, very low" so we can't calculate junction temps.

A type-K thermocouple or thermistor (buried in epoxy) and touching the back of the MOSFET (directly) can give us enough info to calculate the actual junction temp at the power level you want to operate at. It's worth doing it to make sure all is OK. This is only needed if the MOSFET tab is hot while running at that power level. If it's cool or warm, it's not worth doing the equations. The MOSFET is fine.

Quote:
Originally Posted by biskit
Edit: actually I think you should have one shunt per mosfet in this case? That would require a change in the resistor that determines the total adjustment range also right? If so then this could be a 100a/50v/600w discharger, maybe even 750w. And total parts cost would be $30!
Yup, one per MOSFET.
Each MOSFET can receive the same voltage to set the current but that voltage now needs to be based off the current going through one current sense resistor. Then you know that two MOSFETs will give you 2x that current for that voltage level. This makes it easy to expand to any number of modules provided that your voltage source doesn't get its voltage pulled down because you're drawing too much current from it with all those op-amps (if you have lots of "channels").
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 11, 2009, 12:22 AM
Registered User
United States, AL
Joined Apr 2007
832 Posts
So how do you find a mosfet with a dc line on a graph? Just look at the specs for all that meet your basic criteria?

Two of those big ones separately controlled is definitely the way to go, now I just need to find one that works well as a load I guess...
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 11, 2009, 01:18 AM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
I found out the hard way when searching for MOSFETs for my own loads that no manufacturer allows you to search for a MOSFET based on that spec. You'll have to do it the hard way by looking at every MOSFET from several companies that fit your general price range and specs. A quick check of the SOA graph will show a DC plot line, or not.

I ended up looking at the data sheets for several hundred MOSFETs over a 2 month period. It should go a lot faster for you.

Manufacturers to consider:
Infineon
IXYS
International Rectifier
Fairchild
ST Microelectronics

Keep the part numbers for the MOSFETs that meet all your other specs except for the DC plot line, especially if they're really inexpensive. That way, if you can't find one rated for DC use as a load you can at least fall back to the ones that are less expensive. And, as I mentioned, others are successfully using ones not rated for use as a load. But, there are commercial loads out there that are burning out at power levels lower than their ratings and, IMHO, that's not always due to over-optimistic power ratings. Some of those MOSFETs are blowing out from hotspotting and thermal runaway.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 11, 2009, 01:29 AM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Another option is to use a "linear" or "planar" MOSFET. These are either designed to be biased in their linear region (i.e., as a load) or don't hotspot and burn out when doing so.

They're more expensive and typically have much lower max current ratings and higher min. on-state resistance but might fit your requirements.

Manufacturers to consider:

MagnaTEC
ST Microelectronics
IXYS
Infineon
APT

If you end up wanting to use a switching MOSFET (those typically without a DC SOA plot line), designed for on/off use as in a speed controller, then get the highest voltage one you can that meets all your other specs. The higher the voltage rating, the better it will be able to handle being operated as a load at higher pack voltages. For up to 48V packs, I'd try at least 150V MOSFETs, easily up to 250V.
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 11, 2009, 08:13 AM
Registered User
United States, AL
Joined Apr 2007
832 Posts
You see anything wrong with a FDA75N28 ?
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 12, 2009, 01:19 PM
Registered User
United States, AL
Joined Apr 2007
832 Posts
Attached is a test run on a zippy pack. 3s/800mAh discharged at 8a(10c). I wrote down the voltage every 10 seconds and graphed it in excel. A bit cumbersome, but not a big deal to do it to test a pack here and there. An eagle tree or similar to catch the data would be awesome to use with this discharger.

The little bump in the graph is when I tried to adjust the current - it dipped down to about 7.5a for some reason and I nudged it back to 8a. With a 0-50a adjustment range the knob is too sensitive, so a multi-turn or limited adjustment range is a must to get the best possible results. This was the 68th cycle for this pack and it has had a very, very hard life(multiple charges have put in over 800mAh and it's run with lots of time around 25c). Pack started off room temperature and ended up medium-warm.
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 12, 2009, 01:36 PM
Southern Pride
everydayflyer's Avatar
Haralson County GA. USA
Joined Oct 2004
34,839 Posts
One possible LVC / alarm

http://www.rcgroups.com/forums/showp...92&postcount=1

voltage is set for 3.5 (lowest cell)however with alarm triggering (buzzer) at 3.5 3.3 or even 3.0 is close behind at high rate discharges.

You could also build a Micro Screamer (thread in this DIY Forum) and use it as an alert to stop discharge.

http://www.rcgroups.com/forums/showt...ight=LVC+alarm

Copper sheets (strips) available at sheet metal shops that make copper gutters,flashing, vent hoods etc. Square brass tubing available at hobby shops (KS brand I believe) and would provide better contact area for water channels.

Water cooled CPU heat sinks are available but expensive unless you got lucky on E Bay.


Charles


Charles
everydayflyer is offline Find More Posts by everydayflyer
Last edited by everydayflyer; Sep 12, 2009 at 02:54 PM.
Reply With Quote
Old Sep 14, 2009, 03:40 AM
CamLight Systems
New York City, USA
Joined Oct 2003
1,172 Posts
Quote:
Originally Posted by biskit
You see anything wrong with a FDA75N28 ?
Hi biskit,
Sorry, had very limited internet access this past weekend.

Let's check the SOA graph DC plot line...
- Can handle 10A@50V (high voltage pack), about 500W, assuming an awesome heat sink.
- Can handle a bit over 75A at <7V, about 525W, assuming an awesome heat sink.

This is good. Power handling doesn't go down much as the voltage increases.

Its thermal resistance is low, 0.48C/W theta-js. This allows it to handle more power before hitting the same temp. as the IPP048N06L. But, it's lower max temp, 150C, means that it won't be able to handle as much as a 175C rated FET. Run the equations to see what power level each FET can handle before reaching its max temp.

The FDA75N28 is more expensive than the IPP048N06L, but not by too much and needing only two FETs means that the total extra cost is minimal.
The FDA75N28 was one of my choices for an electronic load and my #1 choice for a replacement CBA II FET until I realized that it would be just too hard to fit the pins of a TO-3PN cased FET on the CBA's circuit board.

Good choice!
JohnMuchow is offline Find More Posts by JohnMuchow
Reply With Quote
Old Sep 14, 2009, 07:32 AM
Registered User
United States, AL
Joined Apr 2007
832 Posts
Great, so two FDA75N28 it is. One per op-amp, two shunts, and no ballast resistors. Total parts cost will be probably be under $15 now!
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 14, 2009, 09:51 AM
Registered User
Dan Baldwin's Avatar
United States, CA, Norwalk
Joined Apr 2004
2,724 Posts
Biskit;
Did you ever find someone to make a water cooled heat sink for you? I could probably make an aluminum heat sink for you, but not 'till this coming weekend. I'm out of town (Twin Falls Idaho) until Wednesday.

Dan
Dan Baldwin is offline Find More Posts by Dan Baldwin
Reply With Quote
Old Sep 14, 2009, 12:12 PM
Registered User
MoFl's Avatar
Espaņa, AL, Sevilla
Joined Apr 2007
515 Posts
Maybe one of the 2 first products listed here can be used to cool your FETs
MoFl is offline Find More Posts by MoFl
Reply With Quote
Old Sep 14, 2009, 01:04 PM
Registered User
United States, AL
Joined Apr 2007
832 Posts
Dan: No. If you're willing to make something that'd be a huge help! I'd be more than happy to reimburse you for your effort as well.

Based on my little experiments in the sink I want 5 in^2 of water-metal contact per mosfet. I'm thinking a block around 3x5 inches and 1 inch thick with 2-3 ~0.5 inch diameter holes drilled the long axis for the water flow. NPT threads on the holes to screw in hose barbs. And two small tapped holes on the large face between the cooling paths to attach the mosfets.

It doesn't have to be exact, pretty, or perfect. A scrap chunk of aluminum and 30 minutes with a drill press and a tap is all it needs. I think it could be made smaller, but I'd rather err on the side of caution and have it a bit bigger than needed instead of too small and be limited in power.

MoFl: I was looking for things like that, but those are waaay too small to get so much heat out of them. I found lots of "water blocks" for cooling desktop computer CPUs that would be perfect but they start around $50 and go up, waay out of our budget.

I added up the parts, $14 with a crappy single-turn potentiometer and no on-off switch. I'm thinking maybe a $10 10 turn potentiometer and fixed current range is the answer to accurate adjustments?
biskit is offline Find More Posts by biskit
Reply With Quote
Old Sep 14, 2009, 01:22 PM
Registered User
MoFl's Avatar
Espaņa, AL, Sevilla
Joined Apr 2007
515 Posts
Another (crazy?) idea: What about making the thing watertight and submerging it in water?

Thr 25 kcal you calculated at post#9 would raise the temperature of 5 l of water (small bucket) by only 5ēC.

The load could be built inside a square shaped section alluminum tube (let's say 2x2x5 inches, more or less the coke can you were thinking about, but square section instead) and both sides sealed with plastic caps, taking the cables through sealed holes.
MoFl is offline Find More Posts by MoFl
Last edited by MoFl; Sep 14, 2009 at 04:09 PM.
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 11:17 AM
Mini-HowTo How to build a constant current battery discharger biskit Batteries and Chargers 25 Sep 06, 2009 08:53 PM
how to build a vaccum forming box? rayc Scale Kit/Scratch Built 14 Feb 28, 2008 11:46 AM
Discussion DIY analog constant-current discharger? meteor DIY Electronics 7 Jan 06, 2006 11:09 AM
Anbody know how to build a prop? Pork Tornado Parkflyers 4 Aug 03, 2001 05:10 PM