1000W Discharger Project

[magudaman]

So, I blew up my CBA recently and have found it to be rather undersized for sometime now so I figured it was time to build my own serious discharger.

What I am trying to achieve:

Programmable Constant Current
Programmable Voltage Cutoff
200 Amp Max
75V Max
1000 Watts
Good for single cells all the way up to full packs

The Plan:

I am going to use a Picaxe micro controller to control the entire setup via 4 buttons and a OLED screen. Using the advice of John from Camlight systems I intend on running a bunch of FETs in their linear region basically shorted to the battery. I will be using a current sensor that will feed into the micro-controller to vary the load.

I have been starting to receive some parts in the mail over the last week and hopefully be putting this together.

I am planning on using 6x TO-247 IRFP4110PbF driven directly from the micro-controller's PWM output which will be filtered to DC 0-5V

Early Testing:

I have some IRL2910's from trying to repair my CBA that I have been messing with. I have the gate running to my variable power supply +, Source Battery (-) and variable power supply (-), Drain Battery (+). I getting some really weird effects. Like ramping up from 0-2.2 I end up with 160 watts that just suddenly comes on. But if I connect 2.3 volts with out ramping it up I get like 30 watts that I can controller fairly well as I increase voltage. If while connected I drop below 2.3v I jumps back up to 160w. So that doesn't look all that good.

I did already burn up one of the IRFP4110PbF doing some testing but I think something happened from using 2 ac to dc power supplies sharing a similar ground in the circuit. The IRFP4110PbF now consumes about 80ma @ 5V and is always in a closed or short state.

So I will continue adding photos a progress in this thread!

[Alan B]

You might need more resolution than the PWM has to control the gates. Also to get good current sharing it might be worthwhile to add some current balancing resistors. They can dissipate some of the power as well as reduce the voltage gain of the FETs and make control easier.

[magudaman]

You might need more resolution than the PWM has to control the gates. Also to get good current sharing it might be worthwhile to add some current balancing resistors. They can dissipate some of the power as well as reduce the voltage gain of the FETs and make control easier.

Very True, that is a concern. Looking over the PWM info, it looks like it has 800 steps. Due to the linear voltage zone of the fets I will be dealing with less than 1/5 of that range so maybe 160 steps. I have read about balancing being an issue, but large resisters are the same cost as my FETs. For something like a .05 ohm 50w resistor I'm looking at almost $4 a resistor, but like you said may be a necessary part.

[SamTexas]

Programmable Constant Current

If I were to build a discharger from scratch, I would go for "Programmable Constant POWER" instead. I think it's more useful and it ignores the constantly dropping voltage of the pack during discharge.

[Samson]

Used in the linear region with DC the IRF4110 Maximum Safe Operating Area chart on the data sheet [fig. 8] indicates these devices will not work at more than an amp with 20 volts Vds!

I use Bipolars for my discharge circuit, the secondary breakdown of Bipolars looks similar to the MOSFet Maximum Safe Operating Area chart but there is more headroom typically.

[ZOMGVTEK]

A constant power discharger would be absolutely fantastic.

There would be uses for constant current, and constant resistance as well, but constant power would be my primary pick.

[magudaman]

Programmable Constant Current

If I were to build a discharger from scratch, I would go for "Programmable Constant POWER" instead. I think it's more useful and it ignores the constantly dropping voltage of the pack during discharge.

Well most bikes and motor controllers limit on amps not watts, so I would think that would be more relevant for our application. It would be cool just to have an option to switch between the two.

Used in the linear region with DC the IRF4110 Maximum Safe Operating Area chart on the data sheet [fig. 8] indicates these devices will not work at more than an amp with 20 volts Vds!

I use Bipolars for my discharge circuit, the secondary breakdown of Bipolars looks similar to the MOSFet Maximum Safe Operating Area chart but there is more headroom typically.

I'm looking over the data sheet but not finding what you are referring to. This unit by spec can dissipate 370W but I will only be running it at less than 200w. With its temp junction rating that shouldn't be a problem.

http://www.irf.com/product-info/datasheets/data/irfp4110pbf.pdf

[Samson]

Programmable Constant Current

If I were to build a discharger from scratch, I would go for "Programmable Constant POWER" instead. I think it's more useful and it ignores the constantly dropping voltage of the pack during discharge.

Well most bikes and motor controllers limit on amps not watts, so I would think that would be more relevant for our application. It would be cool just to have an option to switch between the two.

Used in the linear region with DC the IRF4110 Maximum Safe Operating Area chart on the data sheet [fig. 8] indicates these devices will not work at more than an amp with 20 volts Vds!

I use Bipolars for my discharge circuit, the secondary breakdown of Bipolars looks similar to the MOSFet Maximum Safe Operating Area chart but there is more headroom typically.

I'm looking over the data sheet but not finding what you are referring to. This unit by spec can dissipate 370W but I will only be running it at less than 200w. With its temp junction rating that shouldn't be a problem.

http://www.irf.com/product-info/datasheets/data/irfp4110pbf.pdf

Page 4 top right corner [fig 8] of the data sheet. You only get the big amps with fast pulses not DC operation.

[Alan B]

For this linear application there is no advantage to a low RDSon FET unless you are going to test very low voltage sources (like a single cell). More important is the thermal resistance and power dissipation and low cost. Paralleling a bunch of cheap FETs might be the best choice. The balancing resistor, if chosen right, will dissipate a useful amount of power which relieves the FET of that part of the load.

Software can be used to get constant power, or current, or whatever. That's not hard to do. If you have good current and voltage measurements, and control over the variable load you can do many different algorithms.

[magudaman]

Used in the linear region with DC the IRF4110 Maximum Safe Operating Area chart on the data sheet [fig. 8] indicates these devices will not work at more than an amp with 20 volts Vds!

I use Bipolars for my discharge circuit, the secondary breakdown of Bipolars looks similar to the MOSFet Maximum Safe Operating Area chart but there is more headroom typically.

I'm looking over the data sheet but not finding what you are referring to. This unit by spec can dissipate 370W but I will only be running it at less than 200w. With its temp junction rating that shouldn't be a problem.

http://www.irf.com/product-info/datasheets/data/irfp4110pbf.pdf

Page 4 top right corner [fig 8] of the data sheet. You only get the big amps with fast pulses not DC operation.

Arg yah I see that, that is very interesting! Hmm, well that may be a problem. Yah when I contrast it to the Irl2910 which is the device they use in the CBA, the unit I chose does better at low voltage with current (only marginally) but doesn't make it to the upper voltage at all (comparing the 10ms). Damn. Well I got 9 of them left, I guess I'll see if I can sell those off and consider a different FET and a larger quantity. Nice spot though thank you. Less headache in the future then!

For this linear application there is no advantage to a low RDSon FET unless you are going to test very low voltage sources (like a single cell). More important is the thermal resistance and power dissipation and low cost. Paralleling a bunch of cheap FETs might be the best choice. The balancing resistor, if chosen right, will dissipate a useful amount of power which relieves the FET of that part of the load.

Software can be used to get constant power, or current, or whatever. That's not hard to do. If you have good current and voltage measurements, and control over the variable load you can do many different algorithms.

Well I wasn't looking at low RDS on. This had the best thermal resistance by far of almost any device I could find with a crazy 370 watts and when I ordered my 10 units of them it was only $2.80 a piece. It satisfied my requirements and I though it was in every way beyond the unit used in the CBA device but I was wrong!

[amberwolf]

If you really want an easy way to create a high-power controllable load, then depending on the voltage needed, you may be able to use the carcas of an old late-1970s FET-based audio amplifier. Some of them used very large arrays of paralleled FETs already mounted on very nice heatsinks, the whole thing designed to be run in linear mode. Newer ones could also be used, too, but you may find some of these old things easily enough by advertising a want ad for them on Craigslist and the like. Maybe find newer types, too.

All you'd need to do is set up the gate-control-voltage source to drive the gate control of the amp, severing the connection from the audio input side of things.

If you find the FETs in it arent' suitable for your voltage range, you could probably replace them with modern versions that are.

[magudaman]

If you really want an easy way to create a high-power controllable load, then depending on the voltage needed, you may be able to use the carcas of an old late-1970s FET-based audio amplifier. Some of them used very large arrays of paralleled FETs already mounted on very nice heatsinks, the whole thing designed to be run in linear mode. Newer ones could also be used, too, but you may find some of these old things easily enough by advertising a want ad for them on Craigslist and the like. Maybe find newer types, too.

All you'd need to do is set up the gate-control-voltage source to drive the gate control of the amp, severing the connection from the audio input side of things.

If you find the FETs in it arent' suitable for your voltage range, you could probably replace them with modern versions that are.

Well I jumped the gun on a lot of my stuff but I already have a heat sink that should be arriving this week sometime. I'm poring through spec sheets now that I know what is actually important


So I am looking in to a FET let me know what you guys all think. I guess I will run 10 of them:

http://www.nxp.com/documents/data_sheet/BUK95_9610_100B.pdf

They actually rate these ones to be able to do 2 amps @ 100v DC which is better than the IRL CBA units which don't even list the DC. I'm sure there is something I over looked in their spec

.5c/w, should be good for around 100W a unit, has a 175C junction


I am also not exactly sure what I should be looking for with the resistors. They basically go in series with my connection to my FET right?

Battery negative -> Resistor leg -> other resistor leg -> fet source -> fet drain -> battery positive

There are some 35W .05 ohm units that would be dissipating 20 watts at 20 amps (required when doing 200a tests). They $2.60 a piece and are in a T0-220 package too!

[nieles]

They actually rate these ones to be able to do 2 amps @ 100v DC which is better than the IRL CBA units which don't even list the DC. I'm sure there is something I over looked in their spec

the baseline is 1. so you could do 3A at 100v with DC

[Samson]

Nichrome wire in distilled water - 5 ohms

Nichrome.jpg (186.69 KiB) Viewed 1136 times

The BUK95_9610 looks like a better choice but be aware the 2-3 amps at 100 VDC assumes the case is at 25 deg C which it will only be for a few seconds at high power levels.

Using an external load resistor of large wattage will reduce the voltage drop across the FETS allowing them to run more amps. My dis charger has output terminals for an external load. The terminals can be shorted for dissipating internally at reduced power levels or externally into the resistor load for higher power.

You could use a high wattage resistor or nichrome wire with a fan to get dissipation in the 500-1000 watt range.

For load balancing resistors I use nichrome wire cut to length mounted on terminal blocks and Nichrome coiled up in distilled water for a 400 watt external load.

Silicon can only work up to 200 deg C max, Nichrome melts at 1400 deg C and its cheap!

[img]Nichrome[/img]

[Samson]

You might check out this MOSfet, designed for linear mode APL502B2. Very expensive though!

http://www2.microsemi.com/datasheets/APL502B2&L.pdf

[CamLight]

You might need more resolution than the PWM has to control the gates. Also to get good current sharing it might be worthwhile to add some current balancing resistors. They can dissipate some of the power as well as reduce the voltage gain of the FETs and make control easier.

Very True, that is a concern. Looking over the PWM info, it looks like it has 800 steps. Due to the linear voltage zone of the fets I will be dealing with less than 1/5 of that range so maybe 160 steps. I have read about balancing being an issue, but large resisters are the same cost as my FETs. For something like a .05 ohm 50w resistor I'm looking at almost $4 a resistor, but like you said may be a necessary part.

No need for current-balancing resistors if you drive each FET individually...which I highly recommend. Driving each individually also gives you higher resolution for whatever driving method you use as you can adjust each FET on its own. But, Sampson has a great load using resistors like this so it can be done. Either way has its advantages and disadvantages.

If you try to drive a paralleled-array of FETs that are being driven in their linear region, you're going to have lots of problems IMHO. Driving each one separately is so easy to do and gives you the max flexibility for the setup. Attached is the circuit I used for my first load prototype years ago. It was very stable and worked beautifully. You can substitute op-amps (keep the rail-to-rail input feature!) and FETs (make sure it has a DC load line in its FBSOA graph in the datasheet and that you won't go past it). Just drive the op-amp with a filtered PWM signal (series resistor and cap to ground on the PWM output) or a digital pot.

[Edit] Or, if adjusting by hand is OK, just a regular pot.

National Semi CC Dummy Load.pdf

(68.93 KiB) Downloaded 134 times

[CamLight]

Arg yah I see that, that is very interesting! Hmm, well that may be a problem. Yah when I contrast it to the Irl2910 which is the device they use in the CBA, the unit I chose does better at low voltage with current (only marginally) but doesn't make it to the upper voltage at all (comparing the 10ms). Damn. Well I got 9 of them left, I guess I'll see if I can sell those off and consider a different FET and a larger quantity. Nice spot though thank you. Less headache in the future then!

Yea, the IRL2910 is a lousy FET to use in a electronic load and is, IMHO, one of the big reasons why CBAs were burning out at wattage levels below the CBA's rating. I did a thermal analysis of the CBA and that FET was at a 140C junction temperature (max recommended for reliable operation) at 65W! Add on the susceptibility to hotspotting and thermal runaway (especially at higher pack voltages) because it's not rated for use in its linear region with DC current and its amazing that the FETs last at all

[Samson]

http://endless-sphere.com/forums/download/file.php?id=75624

The PDF file above could not possibly dissipate 50 V and 10 amps (500 Watts) as the write up suggests. The Max diss of the the IRF540 is 130 watts at 25C. I think they have misplaced a decimal!

[CamLight]

So I am looking in to a FET let me know what you guys all think. I guess I will run 10 of them:

http://www.nxp.com/documents/data_sheet/BUK95_9610_100B.pdf

They actually rate these ones to be able to do 2 amps @ 100v DC which is better than the IRL CBA units which don't even list the DC. I'm sure there is something I over looked in their spec

.5c/w, should be good for around 100W a unit, has a 175C junction

Be careful, that 0.5C/W thermal resistance is only the junction-to-case (theta-jc) thermal resistance value.
You also have to add on the case-to-sink value (theta-c-s) which is typically 0.5C/W for a TO-220 case. This makes the total thermal resistance for each FET about 1.0C/W. You also need to add on the resistance of any insulator you use between the FET and the heat sink, anywhere from 0.2C/W to 2.0C/W. Then add on the sink-to-ambient resistance of the heat sink (theta-sa) to get the total system thermal resistance you need for power calculations.

IMHO, it's best to design to about 80% of the max temperature spec to ensure long life and a buffer against uneven mounting, dusty fans, thermal fatigue of the FET's internal connections, etc. That's 140C for a 175C-max-rated FET.

As an example, using a two of those FETs on a good CPU cooler, here are the calculations for the BUK9510:
Theta-jc = 0.5C/W
Theta-cs = 0.5C/W
Theta-pad = 2.0C/W
Theta-sa = 0.35C/W

For each FET you have a total thermal resistance of 0.5 + 0.5 + 2.0 = 3.0C/W
Since you have two FETs on the one heat sink the effective thermal resistance is 3.0 / 2 = 1.5C/W
Add on the heat sink thermal resistance and you get the system total 1.5 + 0.35 = 1.85C/W

Assuming a 110C acceptable max temperature rise (30C ambient + 130C = 140C) for the junction and you get 110 / 1.85 = 59.5W
Assuming a 145C max temp rise (to run the FETs at the max 175C) and you get 145 / 1.85 = 78.4W

Using a huge heat sink with 1/2 the thermal resistance (0.175C/W) and you only get about 66W and 87W for the two power numbers. As you can see, getting the FET thermal resistance down as much as possible is a LOT more effective than using larger heat sinks!

If you can accept an electrically live heat sink, you can eliminate the insulating pads and then the power numbers jump to much higher levels (assuming you're using thermal grease on each FET), about 129W and 171W for a two-FET design using a good CPU cooler heat sink and fan. A huge difference!

[CamLight]

You might check out this MOSfet, designed for linear mode APL502B2. Very expensive though!

http://www2.microsemi.com/datasheets/APL502B2&L.pdf

Yea, those linear FETs are fantastic for loads! I did some calculations a while back comparing them against standard (but DC-rated) switching FETs and the cost, as you mentioned, was just too high to be worth it for any of the loads I've designed. Even with all the extra components need to drive several switching FETs. Verrrrrrry tempting though!

[CamLight]

http://endless-sphere.com/forums/download/file.php?id=75624

The PDF file above could not possibly dissipate 50 V and 10 amps (500 Watts) as the write up suggests. The Max diss of the the IRF540 is 130 watts at 25C. I think they have misplaced a decimal!

I believe the specs only refer to the voltage (50V) and the current (10A) as separate, independent, maximums with the power limited by the cooling used for the FET. That document only mentions small loads so they weren't worried about the total power. We, of course, definitely are!

[CamLight]

I am also not exactly sure what I should be looking for with the resistors. They basically go in series with my connection to my FET right?

Battery negative -> Resistor leg -> other resistor leg -> fet source -> fet drain -> battery positive

There are some 35W .05 ohm units that would be dissipating 20 watts at 20 amps (required when doing 200a tests). They $2.60 a piece and are in a T0-220 package too!

If you do decide to use resistors, I wouldn't recommend using ones that mount on the same heat sink as the FETs because then you lose the advantage of having the resistors mounted separately. The heat sink still has to get rid of all the heat if the resistors and FETs are mounted together. Also, those resistors need a LOT of heat sink to get those kind of power levels. Better to use the big tubular resistors (50W and up) with a fan on them.

[CamLight]

So, I blew up my CBA recently and have found it to be rather undersized for sometime now so I figured it was time to build my own serious discharger.

I should have mentioned this earlier but I am very interested in seeing how this turns out. Great idea!

[Alan B]

Another approach is to use some big resistors, a switching FET (or several) and an inductor. Make a switcher into the load. The load can be nichrome wire in water, or the big high power resistors.

[frodus]

always overdesign That's my motto!



Here's my BAD (Big Ass Discharger)..... working on the gate driver and controll circuitry right now, but the mechanical side of it is done....

Designing for ~200A per module, and there's 6 of em..... still need to find some decent fans, these get hot! Each pair of modules protected with a contactor and fuse and each unit has it's own shunt. I know it's overkill, but we plan on testing some serious loads.

and BTW, thanks for the help John. We were considering load resistors too, but ended up going with driving them in their dissipative mode.