Electronic Load Project - TLV171 & IRFP250 with KiCad Files

[t1d]

I am posting my KiCad schematic, for my revised personal choices, for Jay’s Dynamic E-Load. Extra proofing, suggestions and comments are v-e-r-y welcome. I will be ordering the board, soon.

The only other thing that has been suggested is adding a switch, to disable some of the MOSFETs, for smaller loads. I may do that.

I would really appreciate someone open the Kicad files and confirming that the custom libraries work.

My choices, based on my stocks, or purchase preferences, are
- Dual power supply
- Dynamic, including an auxiliary input, for a Function Generator
- TLV171 Op Amps
- IRFP230 MOSFETs
- LT1019_2.5v Voltage Reference
- The MOSFETs will be wired directly to the PCB and abutted to the heat sink

The auxiliary input, for the function generator, is a new feature option. I researched it, through an independent thread. https://www.eevblog.com/forum/projects/hacking-a-dynamic-electronic-load-circuit-to-use-an-external-function-generator/
It is based on
- Injection of the FG signal, before the attenuation pot
- Signal diodes, in the controlling op amp and the driving op amps, placed in parallel to the feedback circuit.
- A 50 ohm termination resistor
I do not have the knowledge, to create this feature… I am thankful to those that helped design it.

Kia, over at TI, has modeled the circuit and it should be stable.

I am using a trick, to be able to use a cheap 100 x 100mm Chinese board… I will cut the board into two pieces, overlap them and tie the traces together, with TH pins, to get the length that I need. My heat sink is very long. So, some of the repetitive symbols, on the schematic, are to facilitate that. The remaining netwire connections will be made, by connecting the boards.



I am, presently, writing a tutorial, for this process and will release it, soon. The trick is in how to make the exact placement of the upper and lower pads so that they will line up, perfectly.

Special thanks to Jay Diddy B, Kleinstein, JS, Kia and all the others, who have so graciously helped me learn, understand and create.

Jay’s thread, with various versions; Post #144
https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/125/

You are free to make any and all use, of the files. However, in doing so, you release me from absolutely all liability. I am not an electrical engineer. I do not guarantee that there are no mistakes. You must verify absolutely everything, for yourself. I did not design the operational circuit.

[t1d]

Kia, over at TI, said:
The usual input bias cancelling resistors from the +inputs of TLV171 to GND are not needed when working with the TLV171, though, because the input bias currents are ultra low.
So, those could be removed, and the space and cost saved.

[Kleinstein]

The current sensing resistors are way to small / low power.  For a reasonable control the drop on the resistors should be more like 0.2 - 0.5 V, maybe even 1 V for a low noise version. So the power level actually used would be up to 0.5 -1 W.  To keep self heating effects reasonable small the resistor should be rated to something like 10 times that. So think about 5-10 W resistors. So this os nothing for 0805 form factor SMD, bit more like THT  wire would power resistors like these:
http://cdn-reichelt.de/documents/datenblatt/B400/KH_SERIES.pdf

Besides the power rating one should also have an eye on the TC. Some of the low ohms resistors have rather higher TC - this could be intentional for use as current sharing emitter resistors.

So the board would not be that compact - more like 3 or 4 output channels on a 100x100 mm board. No need to chop the board in pieces.
If more power is needed, use a 2nd identical board with the small input section not populated.

For just a single piece, and a 1 st test one should not make it that extremely small.

[t1d]

Thanks, Kleinstein, as usual, you have great insight. It always takes me a little time, to study out comments, as I am still a bit of a noob. But, in the meantime, I think some e-load designs use this type of current sensing shunt.

I had known that I needed a way to select the number of MOSFETs I was using, based on the test parameters. I really like your suggestion to use multiple boards.

I intend to make a test board, of a single MOSFET. It occurred to me that, if I make it rather nice, there would be no reason that it could not be used, regularly. That would do, for small tests, then, the two/three split, that you suggest, for the big unit. All of those would give me a lot of divisions and total capacity.

I'm still considering all of it. But, I think I am getting closer, truly, thanks to the good help, of folks like you.

[t1d]

Kia, over at TI, said:
The usual input bias cancelling resistors from the +inputs of TLV171 to GND are not needed when working with the TLV171, though, because the input bias currents are ultra low.
So, those could be removed, and the space and cost saved.

I had thought that it would be easy to eliminate the resistor and rearrange the board layout. Well, it is not so easy, with the current layout design goals.

My goal was to have long, compact MOSFET circuit blocks, to move as many components as possible away from the heat sink. So, with that in mind, I don't see a way to move things around that would be of benefit.

The circuit blocks could, likely, be compacted, by widening them. But, this brings more components closer to the heat sink. If others desire to do that, I will leave that to them, to accomplish. The KiCad files should give them a great start, toward that end. As for myself, I think I will just use a zero ohm resistor, to fill the footprint.

[Kleinstein]

The layout would change a little with a larger resistor / shunt.  I think it is still a little early to start with the layout - it helps if one knows the final circuit first  .

I don't think one would need such a large one as depicted - that one is more like a few mOhms for 10s of amps.  Depending on the requirements one can consider using a single good shunt for the combined current and than have cheap individual resistors for the separate MOSFETs responsible for load sharing. This could be interesting of a load made for very high currents. For just a 1 st test I would keep it at maybe 2 MOSFETs.

It helps if one has clear target specs.  In a hobby project these may be more lose, but one still needs to know the possible use and priorities. The next step are than a few possible circuits (e.g. different number of MOSFETs, different OPs, 1 precision shunt vs. totally separate channels)  that can be compared.

[t1d]

The layout would change a little with a larger resistor / shunt.  I think it is still a little early to start with the layout - it helps if one knows the final circuit first.

It helps if one has clear target specs.  In a hobby project these may be more lose, but one still needs to know the possible use and priorities. The next step are than a few possible circuits (e.g. different number of MOSFETs, different OPs, 1 precision shunt vs. totally separate channels)  that can be compared.

This is a true and excellent point. I thought the circuit was complete; it has been released, in various forms, for a little bit of time.

As a noob, I am learning, by relying on the circuits of others and reverse engineering how things function, by making adaptations, to my own preferences. This is an extremely poor means of learning, but it holds my interest. A little learning, a little KiCad work and a little solder smoke... I have health challenges that effect understanding and retention (I don't recommend aging, to anyone) and working, in this way, I am able to learn.

I am working on a single MOSFET model, first. That way, if there is a single problem, it can't be coming from multiples, of the same circuit. It will be easier, to track down, that way.

I don't think one would need such a large one as depicted - that one is more like a few mOhms for 10s of amps.  Depending on the requirements one can consider using a single good shunt for the combined current and than have cheap individual resistors for the separate MOSFETs responsible for load sharing. This could be interesting of a load made for very high currents. For just a 1 st test I would keep it at maybe 2 MOSFETs.

I did not think that actually using this type of shunt was of interest to me, so I did not study out why they were using them... I only saw it a couple of times...

[t1d]

Just for fun... Here is a rough draft, of the single MOSFET layout. It is intended to be used, permanently, without a case. I have a few more traces to run. I will release the Kicad files, when it is complete.

It will be cut, from a 100x100mm board. So, there is lots of room, to add on other circuits, to make a panel, of multiple designs.

[jrsikken]

You may want to look at my most recent version of my Arduino based electronic load.

https://github.com/jrsikken/ElectronicLoadR3

Here I discuss the changes to make it a better device.

I have changed the MCP6002 opamp to MCP6072 to reduce the offset and to make the output voltage swing to GND as low as possible.

I have changed the mosfet to a BTS133 mosfet which is super robust. It has over voltage, over current, over power protection. This is very useful because people will hot plug the power supply under test and then the mosfet is gone. In addition it has ESD and thermal protection. Thermal protection is also very useful because it will save you an extra temperature sensor and thermal protection circuit!

I had oscillation in my constant current circuit (opamp, mosfet, and sense resistor) and so I had tuned the passive components around it to prevent oscillation but for fastest setttling time. I managed to make it settle in 15-20us. Basically now it does pulsed loads. It's fast enough to simluate the 550us wide 2A pulses from GSM modules.

My circuit uses the MCP4725 DAC with I2C interface. Right now DAC output can change voltage about 5000 times per seconds. But I believe I can make it change at 60kHz acording to this forum post. http://www.stm32duino.com/viewtopic.php?t=1048  I have tested this guy his sketch and it outputs a 3kHz sine wave at 60kHz sample rate. I have not tested it on my electronic load yet, but it should be able to output a 3kHz sine wave. Anyway this means any waveshape can be generate by the MCU itself..

[Kleinstein]

There is a good reason that a good quality electronic load should use a reasonable size (power rating) shunt: self heating of the shunt is a serious limitation and making the resistance very small makes the demands on the OPs precision higher. So a miniature 0805 SMD shunt is not a real option for a current sink in the 1 A range - that would be more for the sub 1 mA range.

It usually also take a good precision OP. So the TLV171 is about the lower useful limit in this respect and not a good choice with a small shunt. The voltage at the shunt is still small and this offset, drift but also low frequency noise can be important. The smaller the shunt, the higher the demands on the OP.

The BTS133 might not be a good choice for linear operation as it is low voltage and relatively modern.  It has thermal protection, but I don't see and SOA protection. Large MOSFETs are usually not really sensitive to ESD from drain to source.

[t1d]

There is a good reason that a good quality electronic load should use a reasonable size (power rating) shunt: self heating of the shunt is a serious limitation and making the resistance very small makes the demands on the OPs precision higher. So a miniature 0805 SMD shunt is not a real option for a current sink in the 1 A range - that would be more for the sub 1 mA range.

Thankfully, I discovered that I had specified the wrong size footprints, for the two shunts and the snubber cap.

Here are the comments, from the original thread that were added to the schematic.
All the resistors can be 1%. The cost of 1%
versus 5% is minimal. Most parts can be 0603,
0805 or 1206 whatever you are comfortable
working with. You need 2512 resistors for the
shunts. The 2.2uF/100V should be 100V 1210 size.

All the resistors can be 1%. The cost of 1%
versus 5% is minimal. Most parts can be 0603,
0805 or 1206 whatever you are comfortable
working with. You need 2512 resistors for the
shunts. The 2.2uF/100V should be 100V 1210 size.

I have adjusted the components, on my board, to meet these size specifications. But, I continue to think on the matter.

As for the matter of component tolerances, I am having significant difficulty, finding 1% resistors and capacitors. And, when I do find one, it is ten times the price. The best that I am going to be able to do is 5% resistors and 10% caps.

It is very good practice, to spend a lot of time, going over a board design. I found several other things. You will see those, the next time I post pictures.

Kleinstein, thank you, for your continued help. I appreciate it.

[Kleinstein]

There is absolutely not need for 1% capacitors - 10% is good enough here, and even 50% would likely be OK.

For the shunts size 2512 is a little better, if the layout (copper area) can carry the heat away. Tolerance is not that import, it is more about the TC.
For the shunts it would help to really see what resistors are actually available. One should also check the parameters if they are really suitable (TC low, power rating high enough). For SMD parts the power rating usually requires a rather large copper area, this can be misleading and in sum the 2512 SMD form might need more space than a 10 W THT resistor.

Resistor tolerance would be only relevant for the scale factor of the current readout. If needed this could be adjusted independently.

[t1d]

There is absolutely not need for 1% capacitors - 10% is good enough here, and even 50% would likely be OK.

Great!

Tolerance is not that import, it is more about the TC... One should also check the parameters if they are really suitable (TC low, power rating high enough).

So, for TC, what would be an acceptable number? Looks like, in both footprints, it is mostly 100ppm... Some are less. See TT Electronics, below.

Resistor tolerance would be only relevant for the scale factor of the current readout. If needed this could be adjusted independently.

Do you mean a resistor with an add trimming pot?

How about this, for a wire wound, for the bigger boards?
https://www.mouser.com/ProductDetail/?qs=SpUj42xpVX7KvrMepLk9lQ%3d%3d

How about this, for a SMD 2512, for the single MOSFET test board? These are 2 watts.
https://www.mouser.com/ProductDetail/Bourns/CRM2512-FX-R200ELF?qs=sGAEpiMZZMtlleCFQhR%2fzQ75E81L6H6zZ7KV21VIi2U%3d

The Welwyn/TT Electronics’ model LRMAP-200-FT4 would be the best, of the 2512 footprint… 1%/3w/+/- 50ppm. But, Mouser does not have it in stock.

All said, it looks like the best cost option, even for the test board, is the wire wound model. It is expensive, but, if need arose, it could be moved to a bigger board.

[t1d]

I changed my search perimeters, at Mouser, and came up with some better sink resistor options.

For the test board, I think I will go with this WW, Chassis Mount, 17mmx17mm, 10w, 5%, 50ppm, $3.65USD.
https://www.mouser.com/ProductDetail/279-HSA5R10J

For the bigger board, I am thinking of going with this guys bigger brother, which is cheaper.
WW, Chassis Mount, 29mmx28mm, 25w, 5%, 50ppm, $2.73USD.
https://www.mouser.com/ProductDetail/279-HSA25R10J

For the two above, Mouser mistakenly has a sale flyer, for the Data Sheets. Find the data sheets, here.
https://www.te.com/commerce/DocumentDelivery/DDEController



I even found an interesting SMD, 2818, 10w, 1%, 75ppm, $1.28USD.
https://www.mouser.com/ProductDetail/71-WSHP2818R1000FEA

[t1d]

This is very close to finished. There are only two considerations, before printing the board on paper and confirming the component sizes…

1) Should the FG Aux Input BNC be turned toward the front, or the side, as pictured? I say the side, to keep the front uncluttered, as the FG will be only occasionally used. What do you think and why?

2) I would like to add a 12vdc output, to drive a cooling fan. But, I am not sure how to create the PSU, for it, because it needs a full bridge rectifier and there is already a half bridge rectifier on the board. Crossing up the ground with the negative source is the concern. See picture.

I will be researching this. If you know the solution, please clue me in… Maybe, use the rectified +9vdc and -9vdc, being 18v and regulate that? Input = +9vdc, ground = -9vdc, output = 12vdc, with the fan input 12vdc and fan ground tied to circuit ground?



I am using these trace sizes…
- 25mils, for the input signal.
- 50mils, for the first op amp’s output, to the second op amp’s input.
- 75mils, for the positive and negative power legs.
- 125mils, for the DUT load.

Here is the schematic and board. Please help me proof them. The more eyes, the better.

Thank you, for your help.

[t1d]

Oops, I see that I have forgotten the DUT Load Fuse, yet again. I will add it, right now.

[t1d]

Updates to the Test Board Design...
1) Shunt resistor footprint tweaked
2) Power input components relocated farther away from heat sink
3) Power and signal ground planes separated
4) Ground plane backed away from heat sink
5) Auxiliary fan power connector added

A couple of considerations
1) The MOSFET ratings are for AC power. I understand, that when used in DC circuits, they should be derated sixty to seventy percent. The advertised ratings are 200v and 30a. At a case temperature of 150*C, the amperage rating drops to 12.5a. If I get 3a and 30v out of it, I will be extremely pleased. What do you think?
2) The fan supply is 12v, half wave. This makes for a 50% duty cycle. So, I can either use a 12v fan and suffer the duty cycle speed decrease, or, maybe, use a ~10v zener clamp, and a 5v fan. My noob brain wonders which would create more airflow... Thoughts?

I am putting together the release of the KiCad files, for the test board. Have I forgotten anything?

[Kleinstein]

The MOSFET ratings are usually not for AC, but in most cases maximum values for one parameter at a time. So maximum current with good cooling (low temperature) and fully switched on. The maximum voltage is in Off mode, or very small current.

For the electronic load it would be the DC (forward biased) SOA that matters, at a higher temperature (e.g. around 50 C case).

For a TO247 MOSFET, 90 W is already quite a bit and would requite a very good heat sink (e.g. fan with high air flow).  So I would be a little more conservative and use only around 2 A, maybe 2.5 A if the voltage is really only 30 V max. Besides the thermal limit, there could be stability limits of the SOA at more than about 30 V. So 1 A to 60 V is way more critical than 2 A at 30 V. One has to remember that not all sources specify an DC SOA - so nothing is for sure. One should definite do a test (possibly destructive for the FET) before real use. The smaller the allowed load the less likely is a failure of the FET.  The good thing in that unless very much on the edge (which is unlikely) the test should not do harm to a good transistor, but only blow a bad one.

It depedends on the fan how well it works with a half way rectified supply. Many fans with BL motor have an buffer cap inside an would thus run not much slower, but mainly put strain on the transformer, diode and capacitor. A simple resistor or low voltage zener in series might be better to reduce the speed.

[t1d]

Kleinstein, it is good to hear from you. I was hoping you would stop by. I can always count on you, for good information. I think I will add footprints, for a few optional passives, in the fan leg, and call it done.

[t1d]

I have ordered the needed parts to assemble the single MOSFET test model. When they arrive,  I will print the board, to scale, on paper and confirm the footprint.

I am confident, enough, that there will be no problems, to release the Single Model KiCad files, now. The zip will include the complete KiCad project files, including custom symbol and footprint libraries. Get the schematic and board, as jpg, here.

The zip file is too big, to combine it with other attachments. You will find it, in a separate, following post.

You will see that I fully developed the Full Wave Fan PSU.

[t1d]

You are free to make any and all use, of the files. However, in doing so, you release me from absolutely all liability. I am not an electrical engineer. I do not guarantee that there are no mistakes. You must verify absolutely everything, for yourself. I did not design the operational circuit.

EDIT: I see that I left off the resistor value, for the current sink resistor. I have updated the files, for that.
HSA10R10J = 0.1R @ 10w. I also added the custom library paths...

EDIT #2: The pinout for Pot1 is set for CCW operations. This is a mistake. You will need to rotate the pot symbol, on the schematic, to swap pin 1 and 3, rebuild the netlist, read the new netlist, into the board, and correct the traces, on the board. As this is a test board, I will not be modifying the files. You can do that, if it is important to you.

EDIT #3: KiCad does not show the values/names of the MOSFET and the BNC connector, on the PCB board editor image, or on the Gerbers (IIRC.) But, they show up, in manufacturing. The fix is to edit these two components and make their values/names invisible, or place them, in a clear spot.

[t1d]

I am also working on a Two MOSFET layout. I am wondering, if I can use a single, 0.1R/25w, sink resistor, for this design. See drawing. Do I have it drawn correctly? If not, how should it be?

[Kleinstein]

One can use a single sense resistor. However one would still need additional individual source resistors. There would be one loop (OP) to control the overall current and a second OP to make sure the current sharing is working. The OP for the current sharing does not needs to be as accurate and fast as the one for the sum. The resistors for current sharing can also be lower grade and lower power.
So a single sense resistor version can make sense if high accuracy is wanted and thus an expensive shunt is used, precision OPs (e.g. OP27) are not that expensive anymore that it really matters.
A downside is that the main current regulator would need to drive 2 gates with quite some capacitance, which can make the whole thing slower. There is also a higher minimum drop, due to the extra source resistors.

The plan as shown has another error: The upper power stage input is connected wrong, it should have a separate resistor similar to R9.

[t1d]

Thanks, Kleinstein!

One can use a single sense resistor. However one would still need additional individual source resistors. There would be one loop (OP) to control the overall current and a second OP to make sure the current sharing is working. The OP for the current sharing does not needs to be as accurate and fast as the one for the sum. The resistors for current sharing can also be lower grade and lower power.
So a single sense resistor version can make sense if high accuracy is wanted and thus an expensive shunt is used, precision OPs (e.g. OP27) are not that expensive anymore that it really matters.
A downside is that the main current regulator would need to drive 2 gates with quite some capacitance, which can make the whole thing slower. There is also a higher minimum drop, due to the extra source resistors.

While I can follow what you are saying, this would be beyond my (present) skill set. Maybe you would like to sketch it out?

The plan as shown has another error: The upper power stage input is connected wrong, it should have a separate resistor similar to R9.

This is the correction...

[t1d]

The parts arrived. I printed the board layout, on paper, and verified all the component sizes. Various layout tweaks, followed.

This is what I intend to be the final design, before ordering the Single MOSFET Test Boards. I am using this post, to show the layout, and bump the thread, to ask for a final review, by everyone.

Kleinstein, does any of this apply to the single MOSFET version?

One can use a single sense resistor. However one would still need additional individual source resistors. There would be one loop (OP) to control the overall current and a second OP to make sure the current sharing is working. The OP for the current sharing does not needs to be as accurate and fast as the one for the sum. The resistors for current sharing can also be lower grade and lower power.
So a single sense resistor version can make sense if high accuracy is wanted and thus an expensive shunt is used, precision OPs (e.g. OP27) are not that expensive anymore that it really matters.
A downside is that the main current regulator would need to drive 2 gates with quite some capacitance, which can make the whole thing slower. There is also a higher minimum drop, due to the extra source resistors.

The plan as shown has another error: The upper power stage input is connected wrong, it should have a separate resistor similar to R9.

Thanks to everyone, for your continued help.