100WV 50A max 500W Active Load

[ZapPat]

Camlight is right, since we already see this SOA problem with the FETs in the original CBA. If you really want something that'll eat up lots of current without a problem (like hotspots developping in your linear biased FET), just use the FET as a switcher as it is designed for.

Of course, by using the FETs as a switch, it obviously won't be dissipating much heat anymore. You would have to find yourself some real chunky heatsink-mountable resistors to put on that huge aluminum heatsink, plus an appropriate inductor (coil) and capacitors to filter the FET output going to the resistors. There would have to be some kind of current feedback from a shunt to control the FET's duty cycle. This may sound complicated, but even the very simple brushed DC motor controllers work this way once they hit their current limited operating zone.

For this load solution one would basicaly need:
1 FET + 1 diode + small heatsink (or multiples for really high currents)
1 Shunt (for current feedback)
PWM controller (could be a specialized IC or use a microcontroller)
Inductor + capacitors (to filter FET switcher output)
Large heatsink such as Rick or Doc have posted here
Heatsink mountable load resistors such as these:
View attachment 50W power resistors - small.jpg

I'm more of a digital-type guy, so this is my non-linear way of seeing this solution. I'm sure there are simple to use IC's out there that could be the brains of this circuit, but one could also write a fairly simple and small MCU program to do this in a customizable way. Another advantage of not using the FETs themselves as the load is that we can change the power resistor hookup quite easily since it is external to the PWM regulator circuit.

 

[rkosiorek]

okay, so which FET would you recommend to use in a linear fashion? is there a suitable fet out there?

rick

ps - i tried a breadboard using a single 4110 and alrady discovered that it only works well at low volts before letting out the magic smoke.

 

[ZapPat]

okay, so which FET would you recommend to use in a linear fashion? is there a suitable fet out there?
rick
ps - i tried a breadboard using a single 4110 and alrady discovered that it only works well at low volts before letting out the magic smoke.

Modern FETs are designed for switching, not linear operation. Maybe find some older scrap equipment and pull the FETs from it to try them out, but I have no idea if this would be any better. I also just checked some lower voltage FETs SOA curves and they don't seem to be much help either.

A switching circuit would not be any more complex than that schematic you posted a while back, BTW, Rick. But I guess the non-linear (PWM) approach scares many people if they are used to linear designs (I'm the opposite). Hey, I would design such a circuit and build it for someone in exchange for some cool ebike stuff! Personnally, I would use a PIC microcontroller since I'm used to working with them, so many possible features could be added in the firmware later if need/want be (serial data out, LCD display, internal resistance measurement, constant power or current modes, selectable low voltage cutoff, user interface program, etc) . However, as I mentioned before, I'm sure there a number of single chip current limiting PWM solutions out there that could be used for the basic functionnality needed (constant current discharge).

 

[Doctorbass]

Well.. That discussion come more and more interesting for me! 8)

Thankd you Zapat and Camlight and rkosiorek!

Maybe with alot of my measurement we together could reght here develop the ULTIMATE high current battery tester for ebike /RC crowd!

I know Camlight already developped something nice.. but it still need a CBA to work with..

Zappat I am willing to trade some stuff with you if you want to begin some desing and develop one perfect for us!

Internal resistor (that could be calculated along the discharge curve and also with many current discharge to obtain a "RI curve" depending of the SOC or discharge current would be nice!

So if you could develop a software + with rkosiorek + the help of CamLight to develop the circuit.. and my help for battery caracteristics and more.. that would be fine..

Usually these device cost thousand of dollars. so here if we could develop one for under than 300-400$ i'm sure many would like to have one!

I feel ready for that

Doc

 

[ZapPat]

Well.. That discussion come more and more interesting for me! 8)
Thankd you Zapat and Camlight and rkosiorek!
Maybe with alot of my measurement we together could reght here develop the ULTIMATE high current battery tester for ebike /RC crowd!
I know Camlight already developped something nice.. but it still need a CBA to work with..

Oh!... I just checked out Camlight's web site for the first time, and I had no idea he designs and sells dischargers already!
I don't like stepping on good people's feet much, so it would kind of feel awkward about making something so similar and impeed him, no? I see he's coming out soon with a "high amp discharge" unit (according to the web site), but only to 12V I believe.
Doc - you want high voltage testing I take it then? I was mostly thinking about a single cell tester for our cell testing needs but with high current capabilities, since this is simpler to do for a start. And how many watts are you aiming for exactly?
Camlight - how much current are you aiming for with your upcoming discharge gadget? It may be big enough for Doc, who knows?

Zappat I am willing to trade some stuff with you if you want to begin some desing and develop one perfect for us!

- -|
(drool...)

Internal resistor (that could be calculated along the discharge curve and also with many current discharge to obtain a "RI curve" depending of the SOC or discharge current would be nice!

Could be done, for sure! We could even handle all the complex math and analysis in the PC software, since the PC would be receiving the data stream already.

So if you could develop a software + with rkosiorek + the help of CamLight to develop the circuit.. and my help for battery caracteristics and more.. that would be fine..

Maybe someone could handle the PCB if they wish (although I could make the schematic it would be based on), it should be fairly simple if starting with just a low voltage single cell, high amperage unit. Come to think of it, if an external power supply is used to power the circuit, a high voltage compatible unit might not be much harder to do..?

Usually these device cost thousand of dollars. so here if we could develop one for under than 300-400$ i'm sure many would like to have one!
I feel ready for that
Doc

I'm sure you are! You're going to want to discharge all those sweet cells you have at higher currents than 35A soon, right? :wink:

Pat

 

[Doctorbass]

Well.. That discussion come more and more interesting for me! 8)
Thankd you Zapat and Camlight and rkosiorek!
Maybe with alot of my measurement we together could reght here develop the ULTIMATE high current battery tester for ebike /RC crowd!
I know Camlight already developped something nice.. but it still need a CBA to work with..

Oh!... I just checked out Camlight's web site for the first time, and I had no idea he designs and sells dischargers already!
I don't like stepping on good people's feet much, so it would kind of feel awkward about making something so similar and impeed him, no? I see he's coming out soon with a "high amp discharge" unit (according to the web site), but only to 12V I believe.
Doc - you want high voltage testing I take it then? I was mostly thinking about a single cell tester for our cell testing needs but with high current capabilities, since this is simpler to do for a start. And how many watts are you aiming for exactly?
Camlight - how much current are you aiming for with your upcoming discharge gadget? It may be big enough for Doc, who knows?

Zappat I am willing to trade some stuff with you if you want to begin some desing and develop one perfect for us!

- -|
(drool...)

Internal resistor (that could be calculated along the discharge curve and also with many current discharge to obtain a "RI curve" depending of the SOC or discharge current would be nice!

Could be done, for sure! We could even handle all the complex math and analysis in the PC software, since the PC would be receiving the data stream already.

So if you could develop a software + with rkosiorek + the help of CamLight to develop the circuit.. and my help for battery caracteristics and more.. that would be fine..

Maybe someone could handle the PCB if they wish (although I could make the schematic it would be based on), it should be fairly simple if starting with just a low voltage single cell, high amperage unit. Come to think of it, if an external power supply is used to power the circuit, a high voltage compatible unit might not be much harder to do..?

Usually these device cost thousand of dollars. so here if we could develop one for under than 300-400$ i'm sure many would like to have one!
I feel ready for that
Doc

I'm sure you are! You're going to want to discharge all those sweet cells you have at higher currents than 35A soon, right? :wink:

Pat

Zapat,

I agree, maybe high discharge volt is not very usefull at high voltage.. since the most importnat is to be able to test single cells..

The reason why i want something like that is because the WMR CBAII + the amplifier is just not acessible for people that want to desing a high power battery pack. + the software have still many bug that we need to compensate.. and that the service that they give is very poor!!!! I sent 5 email and thei did not answered since 5 month!.. (I have the right adress)
I tried to call and they never answered..

So something that could display the capacity and RI with good acuracy and repetability would be nice!

maybe just having up to 200A at 5V would be perfect!

I wonder what mosfet the Dewalt 36V have since it can hold up to 15A at 33V on a single unit..

The CBA too seems to already have a nice SOA curve..
the IRL2910.. able to handle 39A at 10V continuous at 100 degree C with 1.3W/degree C derating factor

Someone suggested a replacement part for it that seems better: (Bob Nuckolls) with the IRFP 2907(470Watt fets) with 149A at 10V 100 degree C with derating factor of 3.1W/ degree C . Calculated continuous current based on maximum allowable junction temperature. Package limitation current is 90A.

Using the great 50% of maximum rating for safety it seems that using 3 or 4 of them would be perfect! i saw it at 4.25$ on ebay!!


(i joined the article in attachement)
http://www.irf.com/product-info/datasheets/data/irfp2907.pdf

So for the power stage, i think it is ok..

Rick are you willing to try to desing an update with it?

Maybe we should start a new post for developing that?

Doc

 

[ZapPat]

That 2907 looks better, Doc... but check this one out for a possible CBA replacement instead (an old datasheet on my HD, haven't checked availability):


This one can take about 6A at 75V instead of 1A at 75V for that 2907 mentioned... six times more current! Both these curves are 10ms and not DC though, so they'll not actually be quite this good.

I would imagine that ST makes also an equivalent 100V part in the same FET series that should also have a relatively good SOA in linear mode operation. I guess ST's stripFET construction make for possibly better linear caracteristics in their FETs, or maybe I'm reading the datasheet wrong.

Note for Doc: Remember that this DC SOA curve is only the limit of the FET when it is acting as a variable resistor, since if you are pulsing the FET you can instead get close to the max rated voltage and the rated max current at the same time, but need something else as a resistive load at the FET's output (in the same way a motor controller behaves).

 

[philf]

Wow. This *is* an involved thread...

There are different objectives, here, with respect to faking out a load that resembles an e-bike. Finding a place to dump energy is always the sticky bit. I've read some threads, here, where people have even used space heaters as loads.

When I was confronted with this situation last year (I needed to calibrate a meter I was building against a known "good" meter, but under serious load), I went through a bunch of different devices that could do the job - even if only for a short period of time - one solution used aluminum-encased power resistors submerged in ice water.

It was while using one of those ubiquitous yellow "contractor" lights that my own light went on. This Halogen heat generator was apparently a 500 Watt lamp. Sure, measuring the resistance of the thing (cold) suggested that the "bulb" was around 4 ohms, but this obvously goes way higher when the thing is lit up - but still... I'd thought about building a "power soak" (thanks, Tom Scholz, for that term) based on these things - hell, an entire single lamp unit often comes on sale at Canadian Tire for under 10 bux, and even includes a spare bulb!

I was instantly put off after discovering that the cost of the "bulb" (tube?) in these things was nearly as much as the whole lamp unit itself... But then I happened into the "Dollar Store" one day... Holy crap!

What I did was to take two lengths of alumimum angle that I had laying around (purchased, originally, from Home Depot) and drill a bunch of holes along them. Then I screwed them to a couple of blocks of hardwood and stuck a fixed machine screw in one side, and a "floating" one on other. The "float" is the result of spring loading the untightened screws along one side of the rig - using readily available springs. I found out that you can get springs from the hardwre store that are fine, if you want to part with a lot of cash, but - conveniently - the perfect springs come "free" inside the cheap click-pens you get at the Dollar Store - six for a buck. And you can set aside the ink cartridges and use 'em later.

Anyways, the finished thing looks like this...

Up close, each "bulb" is spring loaded into position...

Simple solution, but the damned thing works well!

 

[swbluto]

Does anyone know what the electrical resistivity of salt water is? By controlling the distance and "cross-sectional area" between the two probes, it seems like you might have an infinitely adjustable resistance between 0 and the max-resistance and a huge heat-sink at the same time. When the "resistor" starts to deplete, just add water. :lol:

Ok, I just tried it out, and my power supply was reading 2.0 volt for .2 amps, so R = 2/.2 = 10 ohms. By putting the contacts closer and/or using multiple pots of salt water, it seems you might find a cheap source of low resistance. For me, though, it didn't seem like the resistance adjusted much when changing the spacing between the probes. It seemed to be pretty consistent at 10 ohms.

However, when I was trying it, bubbles started to frivolously form at the negative terminal. Until I can ascertain what gases are being created(Remember, salt is NaCl, the more frightening form of gas being chlorine gas), I wouldn't consider seriously doing it. I think it's only hydrogen and oxygen, but there's still chlorine in there...

EDIT: At http://www.energyquest.ca.gov/projects/split_h2o.html, I found out that chlorine gas is being generated by the salt water. So, that might not be a safe thing to do.

*Wonders how many brain cells were killed*

 

[ZapPat]

Very nifty dollar-sense there, Philf! Dollar-sense is kinda like spidey-sense, but warns us of buying things we can get/make for a small fraction of the price.

4 ohms cold is a bit high though if Doc is to do single cell testing, it would take many to get to his desired current levels! Too bad they don't have big 12V halogen bulbs instead of 120V ones...

I must say that I have no idea how much those big 50W resistors I'm holding cost new, but I bet they're not real cheap. Another fav of mine is nichrome wire (heater element wire), since it can be configured to get almost any resistance value needed pretty easily.

 

[ZapPat]

Does anyone know what the electrical resistivity of salt water is? By controlling the distance and "cross-sectional area" between the two probes, it seems like you might have an infinitely adjustable resistance between 0 and the max-resistance and a huge heat-sink at the same time. When the "resistor" starts to deplete, just add water.

Just don't light a fat one up while you're checking out the cool bubbles coming out of that test load!!! :lol:

I actually did this to make hydrogen when I was about 12-13 years old. Very fun indeed, but I tell ya it isn't only hydrogen that comes out of there! Lots of nice invisible, heavy chlorine too. I know because all my tools that were low-lying started rusting up pretty bad over the next many weeks... good thing I didn't have much to rust up back then!

*EDIT* :lol:
I just read the last part of your post that hadn't showed up the first time I viewed it!! And don't worry, our types have plenty of brain cells to loose anyways. It actually helps us blend in with mainstream society loosing those cells, who knows? BTW, if you want to have fun, collect some hydrogen and oxygen in a small vial and light it up!

Pat

 

[Doctorbass]

Does anyone know what the electrical resistivity of salt water is? By controlling the distance and "cross-sectional area" between the two probes, it seems like you might have an infinitely adjustable resistance between 0 and the max-resistance and a huge heat-sink at the same time. When the "resistor" starts to deplete, just add water. :lol:

Ok, I just tried it out, and my power supply was reading 2.0 volt for .2 amps, so R = 2/.2 = 10 ohms. By putting the contacts closer and/or using multiple pots of salt water, it seems you might find a cheap source of low resistance. For me, though, it didn't seem like the resistance adjusted much when changing the spacing between the probes. It seemed to be pretty consistent at 10 ohms.

However, when I was trying it, bubbles started to frivolously form at the negative terminal. Until I can ascertain what gases are being created(Remember, salt is NaCl, the more frightening form of gas being chlorine gas), I wouldn't consider seriously doing it. I think it's only hydrogen and oxygen, but there's still chlorine in there...

EDIT: At http://www.energyquest.ca.gov/projects/split_h2o.html, I found out that chlorine gas is being generated by the salt water. So, that might not be a safe thing to do.

*Wonders how many brain cells were killed*

I remember i saw a guy on youtube with salt water as a CONTROLLER for the dc motor in his electric car!!!

800A..! right in the trunk on a box like an aquarium and multiple conductive plate

Doc

 

[Doctorbass]

Wow. This *is* an involved thread...

There are different objectives, here, with respect to faking out a load that resembles an e-bike. Finding a place to dump energy is always the sticky bit. I've read some threads, here, where people have even used space heaters as loads.

When I was confronted with this situation last year (I needed to calibrate a meter I was building against a known "good" meter, but under serious load), I went through a bunch of different devices that could do the job - even if only for a short period of time - one solution used aluminum-encased power resistors submerged in ice water.

It was while using one of those ubiquitous yellow "contractor" lights that my own light went on. This Halogen heat generator was apparently a 500 Watt lamp. Sure, measuring the resistance of the thing (cold) suggested that the "bulb" was around 4 ohms, but this obvously goes way higher when the thing is lit up - but still... I'd thought about building a "power soak" (thanks, Tom Scholz, for that term) based on these things - hell, an entire single lamp unit often comes on sale at Canadian Tire for under 10 bux, and even includes a spare bulb!

I was instantly put off after discovering that the cost of the "bulb" (tube?) in these things was nearly as much as the whole lamp unit itself... But then I happened into the "Dollar Store" one day... Holy crap!

What I did was to take two lengths of alumimum angle that I had laying around (purchased, originally, from Home Depot) and drill a bunch of holes along them. Then I screwed them to a couple of blocks of hardwood and stuck a fixed machine screw in one side, and a "floating" one on other. The "float" is the result of spring loading the untightened screws along one side of the rig - using readily available springs. I found out that you can get springs from the hardwre store that are fine, if you want to part with a lot of cash, but - conveniently - the perfect springs come "free" inside the cheap click-pens you get at the Dollar Store - six for a buck. And you can set aside the ink cartridges and use 'em later.

Anyways, the finished thing looks like this...


Simple solution, but the damned thing works well!

great and cheap idea Phil.. but this is a resistive load.. and the current is not constant as the voltage decrease...

unless you use it as a part of the load in serie with mosfet that have a feedback with the sensed current..?

 

[Doctorbass]

Very nifty dollar-sense there, Philf! Dollar-sense is kinda like spidey-sense, but warns us of buying things we can get/make for a small fraction of the price.

4 ohms cold is a bit high though if Doc is to do single cell testing, it would take many to get to his desired current levels! Too bad they don't have big 12V halogen bulbs instead of 120V ones...

I must say that I have no idea how much those big 50W resistors I'm holding cost new, but I bet they're not real cheap. Another fav of mine is nichrome wire (heater element wire), since it can be configured to get almost any resistance value needed pretty easily.

guys.. we must think LOW OHM load instead of HIGH WATT... 500W at 120V is 4 ohms

50W at 12V is 4A and 3ohm

the load required to cary like 100A at 3.0V is 30 miliohm load!!

 

[swbluto]

Low ohm... and nichrome wire... hmmm....

I can imagine you could have two posts, one negative and the other positive, and by shortening the distance between the two and by wrapping nichrome around it, you would be able to efficiently and accurately lower the resistance coming from the nichrome wire assembly - Decreasing the distance decreases the "series" resistance and increasing the wraps around it would increase the amount of "resistors in parallel". By adding a fan over the nichrome wire(Or, hey, why not submerge them in good old water?), you can create a pretty good way of centrally dispersing the heat. But simple wrapping would increase its inductance which may be undesireable. Maybe there's some way of wrapping that minimizes the increase in inductance?

But I'm assuming nichrome wire is pretty flexible which it might not be.

Alternatively, you could use more common wire. You just have to adjust the wire to account for the resistance at the temperature it stabilizes at.

As far as ensuring precise "constant current", I'd definitely use a low resistance load(lower than the amps you want at the lowest output voltage) and use a switching buck converter with a feedback loop coming from a shunt. This has the effect of lowering the output voltage, which decreases the current going out of the battery and the feedback loop ensures a tight control on the current(Personally, I'd program a microcontroller to do just as ZapPat offered, but you could use op-amps if you're more analogically suave). By using an ultra low resistance load, you can get a wide current range as well.

Honestly, it sounds like a project like this should cost much less than $400 in raw material costs(USD).

 

[webfootguy]

I wonder what mosfet the Dewalt 36V have since it can hold up to 15A at 33V on a single unit..

Doc,

The MOSFET in the Dewalt pack BMS is FDP038AN06A0 has an RDS of 3.5mΩ and is good for 80 Amps.

Ref: http://www.rcgroups.com/forums/showthread.php?t=587041&page=3&highlight=a123+dewalt+circuit+fuse

 

[CamLight]

Wow, you guys really discussed a lot in just one day!

The new load is not on the CamLight web site because I only put together a 25pc. batch as a limited release, but I do have a 500W (400W continuous), 2V-60V, 150A electronic load available...the CC-400:
http://www.rcgroups.com/forums/showthread.php?t=938266

It's priced at $389, which might seem expensive to some, but there's nothing like it on the market at that power level. And especially at that size.

Two customers have already put five of these in "parallel" for up to 600A, 2000W discharges. Several other customers are using one to three of them for various discharging needs. I'm currently assembling a 4,000W, 750A system for another client for IR, burst capability and continuous testing of LiFePO4 packs.

The CC-400 load can work with the CBA but can also be directly controlled by a standard ESC servo signal (50Hz frame rate, 1mS-2mS pulse width) or "manually" by passing 1A thru the current sensor to start it. It can also function in "Burst Mode", where it alternates between a "burst" current level and a "rest" current level....ideal for checking for IR throughout the discharge curve. The CC-400 is designed to increase the discharge capabilities of cycling chargers and other electronic loads so it does not have a LVC in it. More info at the link above and in the CC-400 User's Guide (link available in the RC Groups thread listed above).

Regarding the IRFP2907 and IRL2910...neither MOSFET has a DC load line in the SOA graph so they aren't even rated for DC use, i.e., as a load. You can't extrapolate from the 10mS load line either as the change from 10mS to DC can be huge depending on the internal structure/fab method for the MOSFET. This is why the IRL2910 in the CBA blows up so much. For more information on using MOSFETs as a load, and the things you need to consider, see my analysis of the CBA II:
http://www.camlight.com/techinfo/CamLight_Systems_CBA_II_Thermal_Analysis.pdf

Same problem with the STB160N75F3, STP160N75F3, STW160N75F3 MOSFETs that ZapPat mentioned...no DC plot line in the SOA graph means they're very susceptible to hot-spotting and burning up when used as a load.

There are many switching MOSFETs that can be used very successfully as loads though. I've had the prototypes for the CC-400 operating for several thousand hours at 140C-175C junction temperatures, tens of thousands of room-temperature-to-175C-to-room-temperature cycles It takes a lot of time to go through the various data sheets and check the specs, but as long as you're using one as a load within its SOA DC plot line in the SOA graph, you won't have to worry about hotspotting. Keeping the MOSFETs cool enough is a whole different matter though!! It took me a year to get the CC-400 down from a shoebox to its present size (a bit bigger than a CBA).

The best TO-220-cased MOSFET I've found for use as a load is the Infineon IPP048N06L. As a load, it's good for 10A up to 18V. The junction-sink thermal resistance is approx 1.0, which isn't as good as, let's say, the IRFB4110, but the IPP048N06L won't blow up. There are other FETs in TO-247 cases that have lower thermal resistance values but you need to balance that with their cost. Make sure to verify that they can pass enough current at the lowest voltage we'll need them for, let's say 2V.

I've modded 2 of my 3 CBA's with the IPP048N06L and have sold a couple dozen modding kits for others to use for upgrading their CBA's and all of the installations have been successful and the MOSFET is running at up to 125W continuous without problems. I don't recommend exceeding 105W-110W though as that is the power level at which the MOSFET reaches its rated max junction temp, 175C.

ZapPat, no worries about you guys creating a load.
In fact, if it's a open-source design, I could create the circuit board and make it available for sale. Perhaps CamLight could even stock parts kits or complete units? I have a lot of experience designing high current circuit boards.

 

[CamLight]

Camlight is right, since we already see this SOA problem with the FETs in the original CBA. If you really want something that'll eat up lots of current without a problem (like hotspots developping in your linear biased FET), just use the FET as a switcher as it is designed for.

Of course, by using the FETs as a switch, it obviously won't be dissipating much heat anymore. You would have to find yourself some real chunky heatsink-mountable resistors to put on that huge aluminum heatsink, plus an appropriate inductor (coil) and capacitors to filter the FET output going to the resistors. There would have to be some kind of current feedback from a shunt to control the FET's duty cycle. This may sound complicated, but even the very simple brushed DC motor controllers work this way once they hit their current limited operating zone.

For this load solution one would basicaly need:
1 FET + 1 diode + small heatsink (or multiples for really high currents)
1 Shunt (for current feedback)
PWM controller (could be a specialized IC or use a microcontroller)
Inductor + capacitors (to filter FET switcher output)
Large heatsink such as Rick or Doc have posted here
Heatsink mountable load resistors such as these:


I'm more of a digital-type guy, so this is my non-linear way of seeing this solution. I'm sure there are simple to use IC's out there that could be the brains of this circuit, but one could also write a fairly simple and small MCU program to do this in a customizable way. Another advantage of not using the FETs themselves as the load is that we can change the power resistor hookup quite easily since it is external to the PWM regulator circuit.

Interesting idea and worth discussing along with a linear design IMHO.

Some random thoughts of mine without any regard to their intelligence level...
- The resistors can very quickly add up in cost but are more forgiving at high temperatures than a MOSFET.
- Surplus resistors are often available but can disappear from the market at any time. I'd love to see a unit designed that others can still build a year from now...if not five years.
- Using resistors in a PWM (switching) design will mean that the available current levels won't be as flexible as a linear design, but perhaps the available power levels and greater number of usable MOSFETs will offset that? Different banks of resistors might be required for different voltage/current discharge ranges.
- A switching design means a lot more electrical noise and a decent amount of prototyping and testing to make sure is well handled/suppressed and that a good PCB is designed. A linear design has a very low bandwidth (is slowwww) and IMHO would be easier for others to simply breadboard up. No PCB required (but still recommended). But, I don't have enough experience with power supply and pwm designs to say this with any authority.

Perhaps working up and testing a single "channel" PWM design to get a good handle on the voltage and current ranges available/possible for a given cost? We can then compare it to the costs and advantages/disadvantages for a linear design.

 

[CamLight]

Did a little looking around and found that one of the FETs I had originally considered for my loads, but wasn't available, is now in stock...the Fairchild Semi FDA69N25 (http://www.fairchildsemi.com/ds/FD%2FFDA69N25.pdf).

It's a "planar-stripe" MOSFET which means it's inherently less susceptible to hotspotting when operated in its linear region, i.e., as a load, than the standard "trench" switching MOSFETs like the IRFB4110. And the specs bear this out.

- It's under $3 in 100pc. lots.
- Can handle 20A at 20V as a load, about 10A at 45V.
- Can still pass over 50A at a drain-source voltage of 2V (need at least 8V gate drive).
- Has a junction-to-case sink thermal resistance of approx 0.5 degrees-C/W.
- Comes in an easy to mount TO-3P case. It's not the big metal TO-3 case, but a plastic case similar to the TO-247.

This MOSFET still looks really good for use in a load. I'm very, very seriously considering it for one of my products and I'm glad that this thread got me looking back at some old FETs I had researched, but couldn't find available anywhere at that time.

While I am a bit torn about helping out with the design of a load that could be competition for my CC-400 (and other loads in development), I'm a huge fan of having the right product available for all levels of users and I think there's a place for something we can come up with here. If there's any way I can help, I'm in.

Besides, it will give me an outlet to use more of all the stuff I had to learn over the past two+ years spent developing my loads.

 

[philf]

There are different objectives, here, with respect to faking out a load that resembles an e-bike.

great and cheap idea Phil.. but this is a resistive load.. and the current is not constant as the voltage decrease...

unless you use it as a part of the load in serie with mosfet that have a feedback with the sensed current..?

Wow - busy thread!

Anyways, yup - I *know* this is a resistive load, and we're looking to build an active one (my opening sentence was a bit too subtle). I saw that the heatsink encased power resistors in ZapPat's earlier post (these are not cheap), and was responding to that.

Those lightbulbs (cold) are 4 Ohms EACH. By building a rig with the spring-loaded "holders', I can parallel as many as I need to get the load I'm looking for. The resistance of a light bulb goes up when its filament heats. Populated with all 10 bulbs (I left room for double that many), this thing pulls just short of 24 Amps at 38 volts. For calibrating an ADC circuit (against a known "good" meter), I like to run with just enough bulbs populated to give me between 10 and 12 Amps - which is only half of them...

 

[CamLight]

The limited availability of the FDA69N25 was worrying (not just to me) so I did a bit of spec sheet reading and the following other MOSFETs would also be candidates for use in a load. Each has varying DC SOA graph specs but any would be OK at 10A each. Some are good for up to 30A each but the cost needs to be balanced with current requirements. But, it gives us a good list of swappable MOSFETs for use in the load.

All are Fairchild Semiconductor:
FDA59N25
FDA59N30
FDA62N28
FDA70N20
FDA79N15

I do not recommend the FDA75N28 as its thermal resistance is higher than the ones listed above, isn't any less expensive, and is limited above 10V when used as a load.

 

[rkosiorek]

i was also looking at the FETs and the other numbers are also not stocked or require large minimum orders. but they do give a larger seletion.

how about the FDA33N25? it is cheap and readily available. with an RDs of 94mR it's a bit high but at just over a buck each we can afford to put a whole bunch in parallel. also they have a much lower drive requirement. worse comes to worse could we not give each it's own voltage follower as a driver? quad op amps are cheap.

i would recommend using old pentium4- socket 478 heatsinks. the original ones from Intel are cheap and easy to find. size is very standard. big enough for a few watts dissipation. and easy to modify for a regular fan. each should be good for 400W dissipation.

at super low voltages even the shunt resistor will be a current limiter. so maybe we would have to settle for a bit of a clip at both the low and the high end of the scale. i was thinking of where to get cheap high power low value resistors and i thought why not just use a 100A or 200A meter shunt? you can usually get these for under $20.00 for the Auto/Marine variety. 10mR values are not uncommon.

just quick figuring the using 6 of the FDA33N25 in parallel with a 10mR shunt. the RDSon resistance plus the shunt would limit to 40A at 1V. up to the design limit of 50A @ 1.3V. the full 50A will be available up to 15V because of the 750W limit. this will decline linearly to 7.5A @ 100V not sure but i think that the LM324 should be able to drive all 6 of these at 10 to 12V.

since i am looking for a manual control, most of my circuit would remain the same. but changing IC1b,c,d analog power limiter for a pic based design would be a better idea. anything has got to be better than trying to use a carefully biased diode operating in it's linear region to generate a linear ramp to set the cutoff.

if i can get some feedback on the viability of using the FDA33N25 in parallel and how many of them i can drive with a single op-amp from an lm324 - i'll redraw the output stage.

rick

 

[CamLight]

Hmm...I checked http://www.findchips.com and all five of the other FETs were in stock in single piece quantities at one or more places.

The FDA33N25 could be used but there are a couple of considerations...
- It's junction-to-case thermal resistance is approx 50% higher than the others. Larger heat sinks (or more efficient ones) will be needed or more FETs to lower the total thermal resistance. This will require more $$ for the sinks or more drivers for the FETs, possibly eating up all the $$ savings by using the less expensive FETs.
- The Rds-on value will quickly jump from approx 90mOhm to approx 180mOhm as the junction heats up. This will severely limit the current we can pass thru the FET at lower cell. But, at 2.5V we should still be able to get at least 10A thru each FET, including current sense resistance and PCB trace/wiring resistance.
- The SOA graph shows it handling 10A at up to about 25V...very good.

The MOSFETs can't be directly paralleled without risking all sorts of problems (the hotter MOSFET will start hogging the current, get even hotter and eventually burn out) so we'll need to drive each one separately in its own control loop. But, as you mentioned, op-amps are cheap.

Mounting the FETs to socket-478 sinks might be a problem? Holes will need to be drilled and tapped for each FET. Not especially hard to do but not easy either for a lot of folks.

Those sinks have a typical sink-to-ambient resistance (thete-sa) of 0.5 degrees-C/W. If we beef up the fan, we might get it down to 0.35 degrees-C/W...maybe. Using, let's say, six FDA33N25 FETs gives us an equivalent junction-to-sink thermal resistance (theta-js) of 0.75 / 6 = 0.125. Adding that to the 0.35 of the sink and we have a junction-to-ambient thermal resistance (theta-ja) of 0.475 degrees-C/W.

If we assume a max of 30C ambient and a max desired junction temperature of 120C (80% of the max 150C rating, the recommended max for "high reliability" applications), then we can allow a 120C - 30C = 90C max junction temperature rise.

Taking the max temp rise, 90C and dividing it by the theta-ja of 0.475 and we get 190W maximum for the heat sink before the six FETs will rise to the max temperature. If we allow the junction temp to rise to 150C, the absolute rated max (not a good idea IMHO), the max power level will be 253W. Not too bad but I think my numbers for the theta-sa for the heat sink with the new fan are optimistic and the achievable power levels will be closer to 175W-200W. Still very usable for a module that can be paralleled with others.

I agree with you on using the PIC. Gives you a lot more freedom to modify things via code changes instead of replacing components too.

Using 20mOhm current sense resistors for each FET operating at, let's say, 10A each, gives you 2W in dissipation in each resistor. Using 5W resistors will keep the temperature drift of the sense resistors down to manageable levels. If we measure the voltage of each current sense resistor to easily calculate the total then the 0.2V current sense resistor max voltage level will give you 41 steps using a 10-bit PIC ADC with a 5V Vref. At 10A, this allows us to read to within 0.25A for each FET....not great. Using a 12-bit ADC PIC and a 2.048 Vref, we can get 400 ADC steps for that same 0.2V. At 10A, this gives us 25mA resolution for each FET....pretty good! No need to use a large external shunt then. The 5W 20mOhm current sense resistors allow for easy "automatic" current control at each FET since each has its own local control loop.

If we used that large 100A-200A shunt (Empro or other) you mentioned, that lets us measure the total current flowing and allow for easy adjusting of the voltage to the op-amps. Using 1% current sense resistors lets us assume that each FET will probably be operating within 5% of each other and we can take that into account when calculating the modules max power rating.

An idea...using a pot on one of the PIC ADC inputs (and a DAC to the op-amps) would allow for easy manual control and also "higher-level" current control via serial comms.to the PIC.

For max safety, I recommend not spec'ing out an AC-DC power supply but instead have a 15V wall-wart to power the unit and use a 12V regulator to the op-amps and a 5V regulator to the PIC. Somebody, somewhere will zap themselves with a badly built AC-DC power supply that "you told them would work" and sue the heck out of you.

 

[rkosiorek]

very valid point about the Wall Wart. i'm actually using one for most things these days. i usually get them at the salvation army or St. Vincent DePaul thrift shops for a couple of bucks each. with the nubers of these things being used for cameras, routers, games, etc. they are no longer hard to find at all.

same goes for 478 style cpu coolers. the local computer store has a whole collection of them. you know all of those guys who buy the fancy ones to replace the standard stock ones. i just drop by and ask for them every couple of weeks. sometimes they got them sometimes they don't but i could collect a half a dozen per month and i live in a city of 65,000. you guys in the bigger city would have a bigger selection.

i'm sort of surprised that they are only good for 190W or so. i would have though double that. but you are using science and that has to mean something. plus it does not matter what we specify for coolers. stock coolers/heatsinks come undrilled and holes will have to be drilled and tapped anyway. that is unless they are supplied as part of a kit. the only ther cheap and readily available heatsinks out there would be the halfbrick and fullbrick types sold for industial use.

so let me see if i understand this correctly. Each FET needs it's own shunt resistor and fet driver in a closed loop. the resistors can be 20mR 5W. i gues 1% resistors would be a good choice.

would there be any advantage to using precision op amps like the LMC6082 to keep the response between different fets and loops closer together?

rick

 

[CamLight]

same goes for 478 style cpu coolers. the local computer store has a whole collection of them. you know all of those guys who buy the fancy ones to replace the standard stock ones. i just drop by and ask for them every couple of weeks. sometimes they got them sometimes they don't but i could collect a half a dozen per month and i live in a city of 65,000. you guys in the bigger city would have a bigger selection.

LOL, I never thought of doing that! But, I'm pretty sure I wouldn't be the only one here in New York City who is thinking of going down to the local stores and checking. Having so many people here may actually be a disadvantage.

i'm sort of surprised that they are only good for 190W or so. i would have though double that. but you are using science and that has to mean something. plus it does not matter what we specify for coolers. stock coolers/heatsinks come undrilled and holes will have to be drilled and tapped anyway. that is unless they are supplied as part of a kit. the only ther cheap and readily available heatsinks out there would be the halfbrick and fullbrick types sold for industial use.

I'll run the numbers again to be sure. If you can get me the thermal resistance spec for the heat sink you're considering, that would help a lot. Or, even a model number would be good and I can search for the spec.

But, I'm pretty sure of the numbers as I have a 0.19 degree-C/W thermal res. custom heat sink on my CC-400 load along with MOSFETs with a much lower thermal resistance and the CC-400 is limited to 400W. And that's when using MOSFETs with a max temp rating of 175C. With a max of 150C, like the FDA MOSFETs we're considering, the max power level for the CC-400 would be lower.

I was thinking though about the load using PCB mounted heat sinks like the Aavid Thermalloy sink here: http://search.digikey.com/scripts/DkSea ... e=HS385-ND

Widely available, inexpensive, and would allow at least 30W each. This is without fan cooling. Arraying these on a PCB and using a down-blowing fan above them would increase their power handling. No drilling/tapping of holes in the heat sink required, no wires would need to be run from the PCB to the MOSFETs mounted remotely on the sinks, and assembly would go much faster. I have some data I gathered when testing these (using a fan) for use in my loads a couple of years ago. I'll dig it up.

What do you guys think? Some users would have monster sinks like Doc, or access to CPU coolers like you, but others wouldn't and these sinks would make for an easy kit. The PCB could have alternate mounting holes to support other types of heat sinks or, with some clever work on our part, one set of MOSFET holes could support multiple mounting styles.

so let me see if i understand this correctly. Each FET needs it's own shunt resistor and fet driver in a closed loop. the resistors can be 20mR 5W. i gues 1% resistors would be a good choice.

That's correct. Attached is a great circuit to work off of. A 5W, 20mOhm resistor will work well for 10A, possibly up to 12A (especially with fan cooling).
1% resistors are what I use and they work pretty well. They get me to within +/-1.5% on average before calibration with my loads.
Be sure the TCR spec number is very low as I use current sense resistors with a TCR of 20ppm and the load drifts up to 0.6% when only operating the resistors at 60% of their rated max power...and that's WITH fan cooling. There are 5W resistors out there with a TCR of 2,000 PPM! These would be a disaster for use in this load. The lower TCR resistors are more expensive but they are a must-have component to prevent drift.

would there be any advantage to using precision op amps like the LMC6082 to keep the response between different fets and loops closer together?

I was thinking about that earlier....I'm not sure. I use precision op-amps in my loads and never tested any "standard" op-amps. The offsets and other specs will be worse for the "standard" op-amps, but if those specs don't vary much from chip to chip, they can be worked around. It's possible big chip-to-chip differences in op-amps like the LM324 that I'm worried about.

We could just make sure that the pinout matches the standard dual or quad op-amp pinout so swapping between op-amps would be easy?