Dynamic Electronic Load Project

[t1d]

The square wave generator part does not need a precision OP - here an LM358 / 741 could be good enough. It's mainly the output drivers that need to be precision due to the low voltage at the shunts.

There tons of function generators circuits built on the XR2206 and the AD8038, too.

[Kean]

Nope, that fix is not going to work.
R21 is a shunt resistor and has to go between the FET source pin (M2 pin 3) and GND.
Your monitor output then needs to average out the voltages across all N shunt resistors by having an Nx50 ohm resistor from the high side (non grounded side) of each shunt resistor.
You've got the configuration correct with M1, R2, and R6 - but it is totally messed up around M2.

Another minor error - the 100R resistor for the monitor output (near R21 labelled R?) connects to ground instead of the junction of R21 and M2 source pin.

Nice one, Kean! The two unnumbered resistors have been assigned as R5 and R6. Here's your fix:

[t1d]

Thank you, for seeing this through, Kean. Hopefully, this better. If not, please keep after me. I appreciate your help.

[hugo]

Hi Jay,

How about replacing the IRF530/IRFZ20 mosfets with some low side power switch like BTS117/BTS133 to make it bulletproof, what do you think, will those work?

[Jay_Diddy_B]

Hi Jay,

How about replacing the IRF530/IRFZ20 mosfets with some low side power switch like BTS117/BTS133 to make it bulletproof, what do you think, will those work?

The issue with the BTS117/BTS133 is that they are SMD packages with no tab to bolt to a heatsink.


If I was to buy MOSFETs for a load I would look at PSMN4R8-100PSE from Nexperia (NXP) These have extended SOA.

Link: https://www.nexperia.com/products/mosfets/power-mosfets/PSMN4R8-100PSE.html

I also like the power dissipation protection implemented in this load:

https://www.eevblog.com/forum/projects/complete-project-constant-current-dummy-load/




This circuit would work better if the transistors were replaced by a transistor array like the LM3046. If you use the transistor array the substrate must be connected to the most negative part of the circuit.

I would need to do some analysis to see if the discrete transistors are good enough.

Regards,

Jay_Diddy_B

[t1d]

T1d,
You need to sort out the op-amp pin numbers.
In an earlier message you had decided to use the LT1013 op-amp with the non-standard pin connections:
But you are showing the, standard pinout, quad op-amp on the schematic:
Regards,
Jay_Diddy_B

Jay, I noticed that the original schematic release used Quad TL074’s. This gives me understanding, of the error I made, in crossing up the Dual LT1013 identifier, with the Quad LT1014 footprint.

Now that I understand what occurred, the solution becomes clear. I will label the existing Quad schematic, with “LT1014,” because the pinout is the same as the TL074. And, I will provide an additional schematic, using the LT1013 and its non-standard pinout.

Here is an example of both. These have my dual ac power supply setup and the LT1019 V Ref, that I have in stock. My final release will include both, of these options. Plus, I will include the battery power supply and the LT6656 V Ref options.

I have inserted 1x3 connectors, between the MOSFETs and their Op Amps. I know that, because of the power involved, I should not use connectors. I will just use the holes to hard wire my control board, to my MOSFET board. I am still looking for the wire pad symbol, to label these holes, correctly.

If you would provide me with the formula, to calculate the resistor value, needed for changes in the number of MOSFETs, I would like to include it, as a note, on the schematic. Then, people can scale the number of MOSFETs, as they like.

Thanks.

[hugo]

The issue with the BTS117/BTS133 is that they are SMD packages with no tab to bolt to a heatsink.

They can be bolted to a heatsink, I guess:

 https://www.digikey.ca/product-detail/en/infineon-technologies/BTS117BKSA1/BTS117BKSA1-ND/5410153

[t1d]

Jay,

Is the information contained in the attachment what I need to learn, to be able to scale the value of the MOSFET heat sinks, for a given number of MOSFETs? I just don't want to be learning the wrong thing.

The KiCad files are just about ready to go. I just have a few more tweaks and I want to add the heat sink formula, as a note.

Thanks

[t1d]

This is the current revision, of the Single Model. It had the same symbol name/symbol icon issue. I cleaned that up, by assigning the LT1014 name, to match the existing quad symbol.

I have finished all the KiCad schematics that I intended to contribute. The only other thing that I would like to do is to add a note about how to calculate the sink resistor values, for changes in the number of MOSFETs.

What about spreading the current through multiples of these resistors. Do we need to add a comment about that? I see that some of the designs use one and others use two.

In reading back through the entire thread, I realized that a good many of the questions, that I was asking repeatedly, and that had gone without response, were answered, in prior posts. My sincere apologies, for that. However, I did not see the answer, regarding the sink resistor calculations. If I missed that, in the posts, please direct me to the post number.

As this version, of the Single Model, has undergone significant change, I would appreciate help in proofing it. Thank you, in advance, for your help. In the meantime, I will be finishing the release notes.

Hmm... I can't get the pdf image, to open, in the text frame. I am using the image function and inserting the file location link, but no joy?

[Kleinstein]

The shown circuit is odd in that the upper 2 MOSFETs get a 10 times higher current than the lower ones. The divider at the OPs output is only for the lower ones.  There is also no need to use the same type of OP for the whole circuit.

With the divider before the output stages it is possible to use a much lower grade OP (e.g. LM358) for the initial part. It does not make sense to include the TL084 option in the single supply version - it just won't work. Something like an OPA4171 would be a suitable single supply replacement.

The IRFZ20 likely not a viable option for linear operation 

For the parallel connection of the shunt resistors, it can be an option, but it does limit the accuracy. Also keep in mind that for accurate current shunts one has to operate them at well below nominal power: at nominal power the self heating is usually too high and causes excessive change in the resistor value. The 0.1 Ohms shunts should operate at something like up to 1 A and thus 100 mW. Still it would make sense to choose resistors that are good for a 0.5 or 1 W power rating. 250 mW types are too small - even with just 100 mW used.

[t1d]

For the parallel connection of the shunt resistors, it can be an option, but it does limit the accuracy. Also keep in mind that for accurate current shunts one has to operate them at well below nominal power: at nominal power the self heating is usually too high and causes excessive change in the resistor value. The 0.1 Ohms shunts should operate at something like up to 1 A and thus 100 mW. Still it would make sense to choose resistors that are good for a 0.5 or 1 W power rating. 250 mW types are too small - even with just 100 mW used.

Good point. I will add this, as a note, to the schematics.

As for the other matters, they are above my pay grade. But as for the upper MOSFETs seeing more current, aren't all the MOSFET nodes connected in parallel?

Do you know how/formula, for calculating the MOSFET's sink resistor values, based on the number of MOSFETs used? That information is all that I need, to finish the schematics, for publication. I want to publish them, this weekend, and move on to other matters.

Thanks!

[Kleinstein]

In the last schematics, there is a mistake (obvious drawing fault)  to give a higher control voltage to the upper stages.

How much voltage to use for the resistor for current regulation is not easy to answer.  Using a small resistor and thus only low voltage (e.g. 100 mV) has the advantage of a lower minimum working voltage and also keeps the power dissipation low. A larger resistor can give lower noise and drift, as the voltage is larger and thus easier to handle. This is especially important if only a small fraction of the maximum current is used.
So the choice is a kind of compromise, likely with something like 200 mV ...500 mV at full current. So the shunts need to be high power: depending on the value one could consider 5 W or even higher power rating. So SMD may not be the best choice. There is a reason the commercial electronic loads often use quite bulky wire type resistors.

One could slightly ease on the problem by a kind of range switching: for low current use only 1 power stage, possibly even a special stage made for lower currents.

The TL084 is cheap, but also rather poor, when it comes to DC and low frequency noise performance. The OPA4171 (alternative slightly lower grade TLV4171) is not that much more expensive.  As the power stages are quite large anyway because of the heat sink, I would personally prefer dual or even single OPs anyway. Another affordable  alternative would be RC4558 (dual)  - still better noise and DC performance than a TL082. A better OP could permit saving on the shunt (smaller and lower power). So for a higher performance solution I would even consider something like OP27 or OPA211 - noise can matter.

For the power MOSSFETs, I would no take the risk with low voltage switching types. There is a real risk they could fail (usually short) due to internal thermal runaway.  Something like an IRFP250 (TO247) is not that much more expensive and could replace about 2 of the TO220 ones.

[t1d]

In the last schematics, there is a mistake (obvious drawing fault)  to give a higher control voltage to the upper stages.

I thought the MOSFETs were all in parallel, at this point. Could you draw a picture, for me, like Jay did? Remember, I am still something of a noob, when it comes to design. That's why I have stayed out of the design conversations and made my contribution to only be the KiCad drawings.

How much voltage to use for the resistor for current regulation is not easy to answer.

I worded my question poorly. What I want to know is how to calculate the resistor value, for the MOSFET sinks, based on the number of MOSFETs and a 50mV (design spec) output.

The Spice model used two 0.2 ohm resistors, for two MOSFETS. I carried the two resistor design (and resistor value), over to the schematics, in which I had included 5 MOSFETs, to reach a 90 watt capacity (to handle testing of PSU designs up to 30v/3a.) Jay said that 0.2 ohms was incorrect, but he did not tell me the formula, to calculate the resitor values correctly. I want to add a note, to the schematics, as to how to calculate the resistor value, so that folks can change the number of MOSFETs, as they would like.

The TL084 is cheap, but also rather poor, when it comes to DC and low frequency noise performance.

As said, I will leave the design elements up to you guys. But, I hope to learn enough to join in, in the future.

[Kleinstein]

The choice if resistor value depends on the maximum set voltage and the maximum current per MOSFET. So the choice is the described design question: a low resistor allows cheaper low power resistors, but causes more noise and drift. So it depends on the priorities - there is no universal correct value.

The IRF530 and maybe IRFZ20 may be good for 1 A up to 30 V, maybe a little more. IF higher voltages can be present, a lower limit might be a good idea.

Assuming a 2,5 V reference (there are different voltage versions of the LT1019), the summing amplifier gain of 2  and with the divider 18K/2K the set voltage goes up to 0.5 V.  So for a 1 A limit the shunts should be 0.5 Ohms  or 2 of 1 Ohms in parallel. With 2 x 0.2 Ohms the current could go up to 5 A per FET and thus too high with more than about 10 V.

Besides the more obvious drawing error, there is a more general problem with the circuit: It lacks a circuit to limit regulator windup. So when sink circuit is powered before the source is connected, the OPs in the output stages will force the gate all the way up, because no current can flow.  If than an external voltage source is connected, it takes some time for the MOSFETs to turn down - so quite a significant current puls could flow. depending on the power of the source this might even cause some damage.
It would be a good practice to turn off/down the current sink if the external supplied voltage drops too far (e.g. below 0.7 V ).


Here the more obvious error in the last drawing  (Yesterday at 04:22:57):

[Wolfgang]

... not powering up before a certain drain voltage is reached could result in a "stuck" situation when the power supply attached does not provide enough current to overcome this.
Another solution would be to limit gate voltage so that the MOSFETs are not driven into full saturation.

[t1d]

Thank you, Kleinstein, for the great explanation and the wonderful drawing. I have to admit, I am going to have to chew on the variables, a bit... lol... In the meantime, I will straighten up the schematics.

Super information, Wolfgang!

Thank you, both.

[t1d]

Here's the fix... Thanks!

[Kleinstein]

... not powering up before a certain drain voltage is reached could result in a "stuck" situation when the power supply attached does not provide enough current to overcome this.
Another solution would be to limit gate voltage so that the MOSFETs are not driven into full saturation.

It would be a rather unusual supply that has trouble to overcome the minimum start voltage with the load turned off. The danger is more in getting an oscillation where the voltage breaks down below the limit when the load is turned on.

Limiting the gate voltage would help a little: it makes the reaction faster and could act as an additional current limit if set accurately.

A good solution would be to limit the effective resistance to something like twice the shunt value. This would be by limiting the set voltage to something like twice the output voltage minus a small offset. So at very low voltages the load would switch over from a constant current mode to a kind of constant resistance mode, not allowing to go below about 2 times the shunts.

[t1d]

Hi, contributors,

In my mind, the development of the KiCad files had reached the point of only needing a comment, as to how to calculate the meter output resistors. I was way, way, over thinking that. The information had, previously, been provided. My apologies, for repeatedly asking for the formula. The light finally came on!

So, as promised, here is the (first) release, of the KiCad files. I need some help, please…
- Open the KiCad files, download my custom symbol library, and add the custom library, to your library preferences. Make sure that the schematics include the custom symbols, when opened. See the following post, for files.
- I have made additional (minor) changes, to all three schematics. Please proof them… hard. Pdf’s provided, below.
Thank you, for this help.

The meter output is specified as 50mV/A. I would like to have a V/A ratio of 10:1. That way, whatever my multimeter volts reading, that is the number of amps being sunk. Is there any problem with using a 9K/1K resistor divider network?

A note to non-KiCad users, making use of the pdf files:
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]

In appreciation, of Jay_Diddy_B’s ingenious design, hard work and gracious sharing, of his Dynamic Electronic Load Project, I am contributing the KiCad files, for his schematics. KiCad is a powerful, FREE program, to make schematics and board designs. Find it here: [Domain Removed] It will be very easy, to create PCB boards, with the files.

What Is included
Circuit development branched into two divisions; the original Dual (Power) Supply model and the Single (Power) Supply model. The Dual Model schematic was further developed to use a common pinout quad op amp (the LT1014, or the TL074) and the dual LT1013 op amp (having a unique pinout.) Therefore, there is a total of three schematics.

The schematics are ready for you to use, as is. You will only need to:
- select your choice of power models,
- select your choice of op amps,
- select your choice of power supply and remove the one you will not be using,
- select your choice of voltage reference and remove the one you will not be using,
- select your components and
- create your own PCB board design.

Feel free to customize. Use components that you like, or have in stock. The selection of the component types, for the board, is done, with the CvPCB function, found on the schematic page. Depending on your selection, from the Power Supply and Voltage Reference options, you may have to relocate the power flags. Whatever you plan to do, you will be miles ahead, using these files.

A note on customization… Changes in components will effect performance; you need to know what you are doing, or get help. If you change the number of MOSFET nodes, you will need to calculate the new value, for their meter output resistors. If you want to use a LT1013 dual op amp, on the Single Model, you will have to change the connections, to the new op amp’s pinout, as the does not have a common pinout.

The wattage, seen by the shunt resistors, can be spread over parallel resistors, but it does limit the accuracy. For accurate current shunts, one has to operate them at well below nominal power; at nominal power the self heating is usually too high and causes excessive change in the resistor value. The 0.1 Ohms shunts should operate at something like up to 1 A and thus 100 mW. Still it would make sense to choose resistors that are good for a 0.5 or 1 W power rating. 250 mW types are too small – even with just 100 mW used. So, the value of parallel shunt resistors equals 0.1 x number of parallel shunt resistor used.

The value of the meter output resistor should be 50 ohms x number of output stages. For example, with four output stages use 200 ohm resistors.

A Note To SMD Users
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.

A Note To Non-KiCad Users
KiCad allows the selection of component footprints, regardless of their schematic symbol. 01x03 connectors, labeled as "Wire Pads," are used here, to represent the selection of Wire Pads, on the PCB board. This method enables the use of separate Control and MOSFET PCB boards. However, these connections should be hard wired. Because of high power draws, connectors should not be used, to connect boards, to MOSFETs. If a single, Control-MOSFET board is used, remove these pads, from the circuit.

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.

Edited by moderator 02/12/2021 - Removed reference to an old URL which is now potentially malicious.

[Kleinstein]

With the shunts at 0.1 Ohms, the meter output is 100 mV/A /N. Where N is the number of power stages. So the scaling of the output depends on the number of power stages. If the output is large enough, it would be acceptable to have a divider to adjust the range to an easier scale.

One would have to adjust the R18/R33 divider to something like 1:50 to limit the current to an acceptable level for the power stage.
With such a small shunt resistance it would be a good idea to have really low noise / low drift OPs.  For the single supply version this could be something like OP213 or similar. The LT1014 might still be ok. However the TL074 is not working single supply  and too noisy to be a practical choice with such a low shunt value.

The alternative would be keeping the 500 mV full scale range and this use something like 0.5 Ohms shunt, that than would need to be high power (like 5-20 W) types and thus likely THT form factor like a TO220, larger case cemented ones or bare wire shunts or quite some size.

[t1d]

Excellent answer, as always, Kleinstein. I will continue to think on it.

[t1d]

I have asked this before, but is there any reason that I could not use AD8032ANZ op amps and MTP3055VL MOSFETs, in Jays design? I have these, in stock. Thanks.

[bicycleguy]

..snip snip..
Jay, I have this 500V 46A "linear" mosfet IXTN46N50L (photo) , datasheet (HERE). 

It has a whopping gate capacitance (Ciss) at 7000 pF, just wonder if the circuit needs a major revision just to make it work as pulsing dummy load ? Say at "reasonable" rise & fall time for working at "common" power supply types. Also when running at the static continuous current, does need modification too ?

A while back in this thread Jay posted a lot of LTSpice work with the IXTN46N50L.  Has anyone done anything with this?  Seems like the package isolation and low RthJC and RthCS would make it easier to deal with than a bunch of parallel mosfets.

[t1d]

As I recall, the author already had these, in his stocks. At $42/ea, at Mouser, it is out of my price range... I imagine that would be so, for most hobbyists. But, yes, you can sink more load, with fewer MOSFETs.