A 300W 2N3055/LM723-based DC Current Load

[adinsen]

I recently decided I needed a DC current load. I looked though posts about DIY designs. I ordered a Xiaolin LM324 electronic load based on enut11's experiences as shared in this thread:

Upgraded LM324 Based 300W 72V 20A Electronic Load

Before the kit arrived, I realized I wanted to go by a different path as modern MOSFETs are generally not designed to work in DC region and all designs seem to use them past their specifications.
This DC load will sink 15A in the range of 2-30V. The LM723’s V- is negatively offset to about -2.5V with a TL431 to allow it to regulate the voltage emitter resistor voltage down to 0. Regulated voltage is averaged across the four emitters of the shut transistors, and the current limiting transistor in the LM723 is used to shutdown in case of overheating. Otherwise it’s a pretty standard 731-design.



I built it using as many parts from my stocks as possible, which explains the choice of cabinet, heat sink, passive cooling, and complexity of the power supply. I adapted a 7107 based panel meter to provide direct current read out.

Fitting of the transistors on the heat sink is critical. Silicone isolators don’t work (I tried). 0.47R emitter resistors are on the high side, but eliminate the tendency to thermal run-away provided mica-shims and good heat conductive paste is used. Only one TIP3055 died during my testing... RIP

I saw significant differences in running temperature under heavy load when measured directly at each transistor, but this was down to mounting problems (screws needing tightening).

My design draws a constant current down to about 2V. At lower voltages it becomes resistive. Being MOSFET based, the Xiaolin load works below 1V.

A few ideas for improvements are ruminating: An Arduino with a 12-bit AD/DA could make the load fully programmable. It just needs a +/-5V symmetrical power supply which could be made simpler from a stacked pair of cheap 5V switch mode modules. With active cooling and by adding more TIP3055’s, the design could be extended past 300W. Temperature compensation should be added as the resistance of the emitter resistors drift upwards with increasing temperature, so the load will reduce the current when it heats up (and the current read out will not show). Individual PCBs under the TIP's could make replacement easier.

I’ve shared my design here in the hope it may be useful for others. Feedback is much appreciated!

/Anders

[David Hess]

Before the kit arrived, I realized I wanted to go by a different path as modern MOSFETs are generally not designed to work in DC region and all designs seem to use them past their specifications.

Up to 30 volts, MOSFETs will usually meet their full power specification, or at least come close.

I built it using as many parts from my stocks as possible, which explains the choice of cabinet, heat sink, passive cooling, and complexity of the power supply. I adapted a 7107 based panel meter to provide direct current read out.

Is that why you used the TIP41?  I likely would have used another TIP3055 for the driver unless I had some random but suitable transistor to use up.

I wonder how good those ST TIP3055s really are.  The ST datasheet is bare compared to the OnSemi datasheet.

How did you tune the frequency compensation?  Or did you?

Fitting of the transistors on the heat sink is critical. Silicone isolators don’t work (I tried). 0.47R emitter resistors are on the high side, but eliminate the tendency to thermal run-away provided mica-shims and good heat conductive paste is used. Only one TIP3055 died during my testing... RIP

If I have extra transistors, I usually match them by Vbe to encourage better current sharing even when ballast resistors are present.

I saw significant differences in running temperature under heavy load when measured directly at each transistor, but this was down to mounting problems (screws needing tightening).

In a production environment a torque driver should be used, and this is not a bad idea for prototyping as well.  Belleville washers will help to get consistent mounting pressure.  Alternatively a stack of wave washers can work.  Do not bother with a split ring washer; they always compress fully without applying enough force.

[adinsen]

Up to 30 volts, MOSFETs will usually meet their full power specification, or at least come close.

Good to know! The data sheets I checked seems to have quite limited SOA area. The 75F75's delivered with the Xiaolin kit are only specified to 10ms pulses.

Is that why you used the TIP41?  I likely would have used another TIP3055 for the driver unless I had some random but suitable transistor to use up.

Yes, I got it out of my box and then I drilled the heatsink for it. I later regretted, another TIP3055 would be better/easier/look good. But the TIP41C is plenty large enough to drive the four TIP3055's even at full load when their hFE is barely 10.

I wonder how good those ST TIP3055s really are.  The ST datasheet is bare compared to the OnSemi datasheet.

Good question. I guess I will see.

How did you tune the frequency compensation?  Or did you?

I did not. I read the data sheet and chose 1N based on that.

If I have extra transistors, I usually match them by Vbe to encourage better current sharing even when ballast resistors are present.

They matched well when cool, but will they still when warming up? From what I read from various sources, I did not trust that. Also, in case of repair, I would not be able to re-match.

In a production environment a torque driver should be used, and this is not a bad idea for prototyping as well.  Belleville washers will help to get consistent mounting pressure.  Alternatively a stack of wave washers can work.  Do not bother with a split ring washer; they always compress fully without applying enough force.

Thanks for the hint! For me as a hobbyist, the problem was jsut as much getting the paste spread well enough. The temperature measurement was quite revealing of my mistakes as I saw one or two running much higher temperature than the two other under load. The ST TIP3055's have a small indent where the temperature probe could reach the metal case, even when mounted.

[coppercone2]

i like this.

What is a bigger BJT transistor that you can use for a load?

MJ11028G ?

[David Hess]

What is a bigger BJT transistor that you can use for a load?

MJ11028G ?

That is a good question.  There are higher power bipolar parts available but they tend to be either TO-3 packaged or ring emitter parts.

Ignoring TO-3 parts, the MJL21194 is the largest 4MHz 200W NPN that I see, and the MJL4281A is 230 watts but also 35MHz.  There is nothing wrong with using a fast ring emitter part, but more attention is required to prevent oscillation.

I am not sure why but the TO-3 MJ11028G is not showing up in the Mouser selection guide for me.  The MJ14002 is a similar 300 watt part but 80 volts instead of 60 volts.

Besides cost, there is some advantage to using multiple moderate power parts because doing so distributes the heat more evenly over the heat sink.

[T3sl4co1l]

Last I looked, best bang per (part only) buck was these,
https://www.eevblog.com/forum/projects/why-are-mosfets-used-as-pass-elements-in-linear-power-supplies/msg3531820/#msg3531820
also MOSFETs such as: PHP33NQ20T,127 (NRND), MCP75N10Y-BP, FDA38N30.
A list made this year would probably find still other things.

Whether per-part cost is sufficient consideration, or adjustments need to be made for cost of hardware, machining, assembly, etc., is another matter.

Tim

[Retirednerd2020]

Nicely done!  Finally a use for all those TO3055's I have laying around!  I built a 20A FET-based load years ago which still sort-of works, maybe it is time for a refresh.  Now I need a good job to put my several LM12CLK's to work.  Love those cans.

[Vovk_Z]

Good to know! The data sheets I checked seems to have quite limited SOA area. The 75F75's delivered with the Xiaolin kit are only specified to 10ms pulses.

A lot of MOSFETs will work fine in an electronic load. For example, look at IRFP240, IRFP250, IRFP260, they are still available.

[adinsen]

Nicely done!  Finally a use for all those TO3055's I have laying around!  I built a 20A FET-based load years ago which still sort-of works, maybe it is time for a refresh.  Now I need a good job to put my several LM12CLK's to work.  Love those cans.

There's a bit of nostalgia involved in my design as well as the MC1723 was given to me by my father in law who used to build big power supplies for telco's, emergency backups in hospitals and other.

In case you want to implement, the 7107-based current display, you might wonder why I put a voltage divider in front of the input (R25+R38) instead of raising REF_HI... it turned out the 7107 that came with the kit I bought had an issue with input voltages > 0.6V. It would just not measure a higher count than whatever that would correspond to given REF_HI. So the REF_HI worked, but whoever copied the original 7107 must have introduced a parasitic diode somewhere on the chip. In case you have an original 7107, you should be able to do without the TL431 as well and use the internal reference.

[Kleinstein]

Before the kit arrived, I realized I wanted to go by a different path as modern MOSFETs are generally not designed to work in DC region and all designs seem to use them past their specifications.

Up to 30 volts, MOSFETs will usually meet their full power specification, or at least come close.

[/quote]
With MOSFETs it really depends on the type how good they can stand linear operation. Modern low voltage MOSFETs may start to see a power reducion even as low as 10 V. A problem here is that there are also data sheets around that show a wrong SOA curve that is pretty use-less. With a modern low voltage ( < 200 V) part and no faster drop shown from a certain voltage on I would be suspicious. Some manufacturers (e.g. Infineon) seem to be OK with the data-sheets, but not all.

There are 3 problems with using BJTs: one is the base current that is needed - this slightly reduces the precision. The transistors can also get quite slow a lower current with no BE resistor in a darlington. Finally with a darlington on similar configuration to reduce the base current problem the minimum voltage drop can be an issue, e.g. when testing single  NiCd or NiMH cells or other low voltage sources.
The circuit as show with a tripple darlington (2 external and 1 x inside the LM723)  there is a minimum drop of some 1.2 to 1.4 V.
If one really wants BJTs, one could use a smaller MOSFET to drive the output stages and this way reduce the drop to some 0.6 V and include BE resistors (would not be an issue with the precision here).

[adinsen]

There are 3 problems with using BJTs...

Thanks for clarifiying! 

Mine can't get lower than 60mA due to offset of the 723 I'm using. Also, it does not work well below 2V, so the load is obviously NOT useful for testing low voltage batteries. I have not tested its transient response, which could be interesting, but I'm not sure how to do it in practice without a an expensive and fast high-power source. I made it to test a power supply so I don't have an issue there with the drawbacks of the triple darlington configuration.

/Anders

[coppercone2]

oh its a 3055 I thought it was a 3053 so I actually looked at TO-3 parts.

Just FYI alot of people route wire to TO-3 parts. I guess bandwidth suffers but it probobly is not that bad.

For mounting them, you can use any heat sink, if its modified with helicoils, then you don't need nuts on top and you can thread into the heatsink without ripping the fins off, and if you use say teflon heatshrink to make it so its OK for the lead to touch a fin, but that is getting into some serious macguyver/desparado

But if you do board mount TO-3 for any reason, use a socket, even if you got a real heatsink. Soldering a PCB mount TO-3 on heatsink to PCB is pain. it might look like some mil-spec satcomm receiver, but probing it requires a PHD in acrobatics.

But you already have wires there. It just needs to be around 3 times more creative to use TO3

[Retirednerd2020]

OH!  I assumed TO-3 too.  I thought the others were driving the base or something.  Shows I didn't look that close.

[MathWizard]

There are 3 problems with using BJTs...

Thanks for clarifiying! 

Mine can't get lower than 60mA due to offset of the 723 I'm using. Also, it does not work well below 2V, so the load is obviously NOT useful for testing low voltage batteries. I have not tested its transient response, which could be interesting, but I'm not sure how to do it in practice without a an expensive and fast high-power source. I made it to test a power supply so I don't have an issue there with the drawbacks of the triple darlington configuration.

/Anders

Looks like a nice project, I need to make an electronic load again some day too. So for your under 2V and low current testing, if you think you need it sometimes, wjy not just make another little circuit/PCB, and with another switch or 2, to disconnect it, form the output, or even the power too, when not needed ?

If you went to all this trouble, 1 more little PCB shouldn't matter!

[T3sl4co1l]

With MOSFETs it really depends on the type how good they can stand linear operation. Modern low voltage MOSFETs may start to see a power reducion even as low as 10 V. A problem here is that there are also data sheets around that show a wrong SOA curve that is pretty use-less. With a modern low voltage ( < 200 V) part and no faster drop shown from a certain voltage on I would be suspicious. Some manufacturers (e.g. Infineon) seem to be OK with the data-sheets, but not all.

Most MOSFETs have 2nd breakdown (and yeah, you can call it 2nd breakdown, it's the same basic dynamics as in BJTs) above maybe 30 to 80V or so, making low voltage operation trivial, and you merely have to select datasheets to find parts with wide SOA when higher voltage is needed.

I'm not sure what you mean by "wrong curve"; how else would you determine it?  Have you discovered parts with outright wrong curves, that failed under well-defined conditions (esp. temperature/heatsinking)?  (Reported to mfg? Name and shame?)

I've seen some people wave hands about Spirito effect, but it's meaningless without actual parameters of the die itself -- power density and distribution, tempco, die dimensions and conductivity, tab thickness, etc.  Indeed one of the referenced commonly cited, specifically throws up their proverbial hands for this reason; unfortunately it's a short paragraph admission, so is easily missed.

Anyway, despite a power density higher than ever, there are still SJ types with full SOA; I don't know how or why that happens, but, I have tested a few and confirmed them.

There are 3 problems with using BJTs: one is the base current that is needed - this slightly reduces the precision. The transistors can also get quite slow a lower current with no BE resistor in a darlington. Finally with a darlington on similar configuration to reduce the base current problem the minimum voltage drop can be an issue, e.g. when testing single  NiCd or NiMH cells or other low voltage sources.
The circuit as show with a tripple darlington (2 external and 1 x inside the LM723)  there is a minimum drop of some 1.2 to 1.4 V.
If one really wants BJTs, one could use a smaller MOSFET to drive the output stages and this way reduce the drop to some 0.6 V and include BE resistors (would not be an issue with the precision here).

Base current is easily mitigated by measuring total current at the load terminals; which can also be by summing source/emitter resistors, which you need anyway for balance (with or without op-amps to enforce it; source resistors are reasonable if you don't mind a couple volts dropout).  Can get a bit tedious, wiring everything up, need a couple more op-amps to ensure you've got all the sums and differences in the right place, but I mean, you do it and it's done, it's fine.

Tim

[boB]

I would pick parts with a large die.   I think the OnSemi part(s) that Richard mentioned may be ones.

I would probably have used a good audio amplifier part like the MJ15024  but that is a TO-3

[adinsen]

oh its a 3055 I thought it was a 3053 so I actually looked at TO-3 parts.

Sorry, I realize the subject line was misleading, it should have said TIP/2N3055. There are no TO-3 devices in my box.

[adinsen]

Looks like a nice project, I need to make an electronic load again some day too. So for your under 2V and low current testing, if you think you need it sometimes, wjy not just make another little circuit/PCB, and with another switch or 2, to disconnect it, form the output, or even the power too, when not needed ?

If you went to all this trouble, 1 more little PCB shouldn't matter!

It was fun making it and I'm happy people find my idea useful I have thought about something similar, but now that I have both my load and the Xiaolin LM324 MOSFET kit, I plan to do the switching manually