Thank you for sharing. Your work uses basic circuits to achieve these functions. Even if the design lacks protection circuits, the project looks complete！
Very well done, especially the software is first class. It would be nice if you would reference the originator, John Scully more prominently. It's good to see that you fixed all the hardware issues that plagued his incemental design.
Very clean! It has the looks of a finished product. It actually looks better than some pre-production devices I have seen, let alone prototypes. Cost conscious choice of components too.Taking into account the limited possibilities of the LCD screen and buttons the user interface seems intuitive and quick in its use. You could add a menu with settings memories and also a constant voltage mode. For protection of both the load and the device under test you could also include overvoltage, overcurrent and overpower monitoring that can switch of the load automatically. Do you plan to add a possibility to manage the load remotely for more complex scenario's? For example with the application Test Controller? How 'fast' is it and how precise is it with very low currents? The use case I think about is simulating devices that have sleep time followed by sudden short wake-ups with peak consumption (think bluetooth, lora, wifi, etc... based sensors, smartwatch, wireless headphones and many more. Would be nice to be able to simulate such a device to test batteries or solar power/charger/battery combinations.
Thank you, that's very nice to hear! There is power motoring that makes sure it stays bellow 300W, and there's also overtemperature protection, that turns the load off if the heastink reaches certain temperature. I didn't plan on adding remote controll for this device, because I don't see a need for that. Maybe I'm wrong? It can switch currents fairly fast, though I haven't recorded any measuremets of that yet. It can go down to around 4mA and I guess it would fit the application of simulating devices that you mentioned. If minimum of 4mA isn't enough, this project could be easily modified to operate with smaller currents. This would of course come at a cost of smaller current range.
You may want to look at the INA228 it contains all the hardware needed to measure current and voltage plus it calculates power , energy and charge with high accuracy or for a pre built unit the Junctek KL140F saves having to write software to do it and they contain 20 & 24 bit adc's which alone would cost more than these devices.
The ADS1115 may be cheaper but it is only an ADC the INA228 calculates the current , power , watts , battery capacity , watt hours to a high accuracy no calculations are needed in your own code @@dominikworkshop6007
Can resistors R4, R5, R9, R10, R14, R15, R19, R20 be replaced with 0.5ohm resistors?, or can they be replaced with other types of resistors that are easier to find? Thanks sir!
Yes, they can be replaced with 0,5 ohm resistors, but that would mean more power loss on these resistors, unless the maximum current is reduced. They can certainly be replaced with other types of resistor, but I suggest ensuring that they have low thermal coefficient.
Hello Dominik, great work🎉. I do have a question for you. There is a labeled pin as “OFF” in the schematics. Since this pin is same in every mosfet and controlled by 1 output pin so how do you calibrate each mosfet seperately? Thanks!
Thank you!!! I don't have to calibrate each mosfet seperately. The op-amps ensure that the current is very evenly distributed. I used 4 resistors to sum the voltage on shunts, but it turned out unnecesary (maybe except for when there could be some hardware failure), because the op-amps are so precise, that the voltage on shunts is pretty much identical. Hope I managed to clear this out.
Very nicely done indeed! Did you take a closer look at the dynamic behavior of the load? I found that it's highly recommendable to add a RC snubber to the input of the load. (How) did you address issues related to unregulated operation, i. e. when you disconnect the power supply from the load while the load is still switched on? Without any counter measures, connecting a sensitive power supply in the unregulated state could end catastrophically for the power supply. On the other hand, connecting low impedance/high energy sources might even damage the load itself. (Assuming a minimum resistance of potentially significantly less than 0.5 Ohm, a 50V power source could momentarily force currents in excess of 100A through the load.) In the BOM it looks like you're using IRFP250. 300W is pretty high for four of these MOSFETs, especially if reliability is a concern. Unfortunately the safe operating area isn't specified for DC at all and for diy builds it seems impractical to characterize the MOSFETs. As a starting point I'd look at commercial implementations: Most loads that use these or similar MOSFETs use 8 to 10 in parallel for 200 to 300W total (they usually allow for higher voltages at the input though, often 120V to 150V). I'm in the process of building my own low power (about 50W) precision DC load with two current and two voltage ranges. It's soo much work...
Thank you! Yes, I checked how it responds to sudden input voltage changes and current changes. I made sure that there are no oscillations. The load looks very stable. Regarding connecting/disconnecting a power source when the load is on - I just assumed that you have to turn the load on when the power supply is already connected, because I don't see how this could be solved otherwise. I wonder how it is in proffesional loads. One thing worth mentioning is that the current sink circuitry is very fast, so the load would short the loaded power supply for just a brief moment (though it's obvious that this could still damage the power supply or the load itself, as you said), which isn't so obvious, becouse I saw designs where the power MOSFETs were controlled directly by some kind of DAC, so there was a lot of time between connecting the power supply and microcontroller recognising that the current is WAYYY too high. The 300W rating of this load isn't like a continuous rating, it's more like how much power you can dissipate for a few minutes (assuming that you start with a cold heatsink). I agree that if you use it at it's max power it could fail after time. I was planning on using more MOSFETs in parralel, but that would require balancing resistors, which I was avoiding, because I wanted to minimalize the minimum input voltage for full 8A. I'm curious how they do it in proffesional loads, because I'm controlling each MOSFET individually with an Op-amp, which ensures that each transistor dissipates the same amount of power. But I'm pretty sure that's not how it's done in commercial loads. In DIY designs I saw, the MOSFETS were simply parralelled with an equalizing resistor in the source path. This makes a compromise between the even current distribution among the transistors, and the minimum input voltage for full current rating. I'm curious how your project will turn out! And I strongly agree, that building a nice load is quite a hassle.
@@dominikworkshop6007 In principle, you're absolutely right: One should never disconnect or connect a power supply to the load while the load is switched on. However, in normal use, one might accidentally forget that the load is still on. You're right again saying that the short circuit lasts only a short period of time. I would argue that "short" is relative and the poor opamp has to drive the gate (with its capacitance) from about opamp V+ (e. g. 12V) down to about 3-5V. Also, MOSFETs have a tendency to turn themselves on when there is a large dV_DS/dt due to the internal capacitance, i. e additional charge is introduced to the gate. This makes the problem potentially worse. So the opamp's slew rate, the opamp's output impedance, the "speed" of the control loop and the gate resistor will likely determine how bad the overshoot will be. In my opinion it's all but impossible to avoid the problem completely. E. g. the load can't distinguish between a power supply being disconnected (effect: voltage decreases) and the voltage being decreased on purpose. What is an option though: Reducing the effect of this issue. I have a HP 6060B (schematics available: google HP 6060b Service Manual) which uses a pure analog control circuit for CC, CV, CR (CP is missing completely). They implemented a fairly complex analog circuit to detect the unregulated state (signal UNREG) and likely limit the conductance of the MOSFETs in this situation. Still, connecting a power supply with the load enabled leads to a certain overshoot relative to the programmed value. For my design I use a very similar approach: I effectively limit the gate voltage to a certain value that depends on the current setpoint. (Since the MOSFET's gate voltage varies between different samples and is highly temperature dependent one has to be careful not to limit the voltage too much.) With this the maximum gate voltage would be rather between 6 to 8 V instead of say 12V. This reduces the overshoot massively since the opamp doesn't saturate (hence eliminating the significant recovery time) and doesn't have to work as hard to bring the circuit back into regulation. In addition it might be a good idea to sample the voltage continuously and switch the load off if the voltage falls below a certain disable threshold (e. g. 0.3V). Using a hysteresis and a time delay one could re-enable the load after a certain period of time if the voltage rises above an enable threshold (e. g. 0.5V). This feature could also be made optional. I think this is something that you could implement fairly easily without any hardware modification. Pretty much all commercial designs I know of use one opamp per MOSFET plus a current measurement resistor. Opamps are cheap and this is the only way to achieve optimal load sharing. However, some use inverting, some non-inverting topologies. Some loads use an additional current sensing opamp per current measurement resistor and sum up the currents (or rather voltages proportional to the currents) afterwards (e. g. HP 6060B), while others measure the total current across a high power current shunt in addition to the local current regulation (e. g. BK Precision 8600 series).
@@sebastian_harnisch In my load, the maximum gate voltage is only 5V (due to max supply voltagee of the op-apm) which limits the maximum current, but of course as you said, due to the miller capacitance as the Vds rapidly raises, the gate voltage can raise as well. Now that I think about it, turning the load off when the voltage is close to zero is a great idea, if this feature is made optional. I will definitely implement this in the next software update. I'm pleasently surprised that commercial loads also use one op-amp per transistor. Now I think that I could try using some more generic op-amps for regulating the current through each MOSFET separatelly, and using a precision op-amp for measuring the total current through a precision shunt resistor. My only concern is stability, I'm afraid that using so many op-amps might produce too much gain and the system could get unstable. Mayby by creating suitable negative feedback with RC networks would do the trick, but I'd be afraid that the phase and/or gain margin would be small, and I could assume it's stable, but with some environmentall changes, the design could get unstable. I'm not quite sure how to get bode plots in such device to check it's stability. Do you have any tips for that? Thank you soo much for your feedback! It is very valuable for me! (and sorry it took me soo long to reply)
@@dominikworkshop6007 Ah, interesting. 5V might be fine, I guess you have checked it, but depending on the conditions (temperature etc.) it might not be quite enough to achieve a low Rds(on). (If not for normal use a slightly higher voltage might help with the „short“ feature many commercial loads offer. On the other hand that might be something you‘re not interested in.) Just lowering the voltage imho is only part of the solution. The saturation issue is at least as important I think, so the way to go is to both reduce the gate voltage and make sure that the opamp never saturates. Some opamps might be less susceptible to this issue than others, but it might be better to just implement a circuit that avoids this situation completely. Yep, I use this approach in my design. I think this makes the whole thing much, much more complicated and stability is a huge concern. (Just having many opamps doesn’t necessarily increase the loop gain though.) A good starting point would be to simulate the circuit in something like ltspice (or maybe qspice, seems to be a good/better alternative created by the original author of ltspice) where you can do bode plots reasonably easily. I spent more time with that than I care to admit ;) This though helps enormously to figure out, what might work and what not to try at all. I optimized the inner loop first and only then proceeded with the complete control circuit. I then used a simple bread board prototype to take some measurements with my Siglent SDS 2104 plus scope. It has the bode plot feature. However, I mostly modulated the setpoint of the control circuit and measured the current through the current shunt to determine the transfer function/bandwidth. I did not inject the signal directly into the feedback loop (via an isolation transformer) and did not measure before and after the "injection point" which you would do for actual stability measurements (gain/phase margin). What I did do was looking at the step response etc. There are some videos on TH-cam specifically about measuring stability of power supplies etc with frequency response/low frequency network analyzers (in case you have the equipment available to you) which might help. I currently don't have the setup (suitable isolation transformer), although I was thinking about building one for stability analysis at some point. Getting the loop gain right is important. The loop gain can be adjusted for example by varying the gain of the current sense amplifier. DC loads often feature two current ranges, but they only attenuate the setpoint by 10x and don’t switch to a different current shunt, hence the loop gain stays the same - easy. It gets much more complicated if you change the current shunt, because then the loop gain changes and you have to switch *something* to achieve a lower gain… During the simulation I found quickly that I have to use compensation for the „outer“ control loop (integrator/compensator) as well. This helps massively with the step response. Also, as mentioned before, having a snubber circuit at the input is really important (if not necessary in many cases). This is especially true if you want to build a reasonably fast load. Regarding loop stability vs environmental changes (temperature): I would argue that resistors and capacitors (especially C0G/NP0) are relatively stable over temperature and the MOSFETs are probably the main concern (or more specifically their transconductance). The inner control loop should be designed to handle the variations between two or more mosfets of the same type anyway, so the temperature is just another thing to look out for… Haven’t done extensive testing on my design yet. Simulations can help here too, but not all models may reflect the true behavior well or at all. In the git repo you mentioned that you'd like to add reverse polarity protection. What do you have in mind? The load is kind of protected already - the body diodes will limit the reverse voltage to a low value and are conveniently placed on a heatsink. Some commercial products seem to rely on this form of protection (e. g. HP6060B). Also typical for power supplies. I see the disadvantage that this potentially "shorts" the power supply.
@@sebastian_harnisch I don't mind MOSFETS not being able to short the input completely, I didn't aim for that. When it comes to simulation in LTspice, I've had problems with importing simulation models of elemnts used in my projects. I guess I have to play with it a bit more. But the most confusing part for me doing the bode plots, I managed to get some sensible results, but I'm just never quite sure if I got everything right (eg. if I injected the signal in the right place?). Thank you for all that information, I will try implementing all the avice I got! I find such reverse polarity protection, that shorts the power supply, to be a bad protection, I wanted to implement something, that would prevent reverse current flow alltogether. Of course I wouldn't want to put a diode in series with the load, but I'm struggling to find a solution for that.
Thank you! Having a PCB made is very easy. There is a gerber file on my repository (link in the description). You just have to upload it to a PCB manufacturer, like JLCPCB, and order it.
@@dominikworkshop6007The PCB is ordered now, we just have to wait until it arrives. What program is used to upload the code? I can't find the list of components, maybe I don't have the right programs. Sorry for the questions, I'm a beginner, thank you very much
@@marecek19891989 I used platformio to upload the code. There are many tutorials on TH-cam. There is a pdf with full schematic on my repository. You could also open this project in Eagle to be able to inspect what goes where more easily. Let me know how this project is going for you!
@@dominikworkshop6007 Hi, my project has progressed. The PCB arrived from China looks very good. I am ordering components, but it is very difficult here in Slovakia. I can't buy measuring resistors :( where did you get them? Thank you
Did you use an input trigger? I mean define in defines.hh is wrong. It corresponds to ENCODER_BUTTON and the signal input trigger is connected to pin A7 on the PCB. It's just an analog input. Has it ever been tried? Miroslav
You are right, thank you for pointing it out. I haven't used the input trigger yet, and there is an error in defines.hh, where I defined this pin to pin 3, instead of A7. The two pins I know can work as external interrupts were used for the ENCODER_A and ENCODER_BUTTON signals. I'm not sure if I will be able to configure the A7 input as external interrupt, but in that situation I planned to just check the state of this pin frequently, and act accordingly. I know it's not perfect, but the uC used in this load has some limitations. If I were to make a V2 of this project, I would probably base it on some STM32 uC.
@@dominikworkshop6007 Maybe the PB3 pin can be used on the current system, which is still unused. It's a keyboard spare pin and MOSI too. And there is PCINT3 attached.
![อาจารย์ยอด : พระเตะผี [ผี]](http://i.ytimg.com/vi/zusDOIFasaE/mqdefault.jpg)
Reminds me of the Scullcom DIY DC Electronic Load.
I enjoyed his series because he goes through all of the design steps.
Yeah, my load is heavily inspired by his construction. I mainly changed the analog part of the circuit.
Anyone know what happened to Skullcom ?
@@tonybell1597 I read that he has personal projects in progress, AFAIK is everything ok
Terrific good looking DIY project, really neat. Congrats and thank you for sharing.
Thank you!
Man you did a wonderful job on this! I definitely want to build one when I get some time
Thank you so much!
Very nicely done and well explained. The PCB is well designed and looks as good as something that would cost hundreds. Thanks!
Thank you!!!
Beautiful workmanship
Thank you!!!
Awesome. A lot of work here! Good job!
Thank you so much!!!
Very nice!
I will watch for advancement of the product.
Until now, this is the BEST DIY LOAD which I found.
Thank you, I'm glad to hear that!
Thank you for sharing. Your work uses basic circuits to achieve these functions. Even if the design lacks protection circuits, the project looks complete！
Thank you!
You've earned yourself a subscriber! Great Job!
Thank you!!!
Awesome project, I love it ❤❤❤
Thank you!!! :D
Amazing work. Thanks a lot for share it
Thank you!
Great work!
Thank you!!!
Very very nice!
Thank you Charles!!! :D
excellent job. You should have more views and subs. Keep up the great work
Thank you!!!
Good job.. ❤.
Thank you! :D
No way this is diy, very nice!
Thank you! I'm glad to hear that! :D
Very well done, especially the software is first class. It would be nice if you would reference the originator, John Scully more prominently. It's good to see that you fixed all the hardware issues that plagued his incemental design.
Thank you, I'm glad to hear that!
I will start making it too soon.
Very nice
Thank you!
Perfect
Thank you!
Awesome
Thank you!
Amazing project
However, I separate display for measure value will be helpful. A diff display to select the modes will make it much more user friendly
Thank you! I get your point, though that would make this project a little bit more expensive.
Very clean! It has the looks of a finished product. It actually looks better than some pre-production devices I have seen, let alone prototypes. Cost conscious choice of components too.Taking into account the limited possibilities of the LCD screen and buttons the user interface seems intuitive and quick in its use.
You could add a menu with settings memories and also a constant voltage mode. For protection of both the load and the device under test you could also include overvoltage, overcurrent and overpower monitoring that can switch of the load automatically.
Do you plan to add a possibility to manage the load remotely for more complex scenario's? For example with the application Test Controller?
How 'fast' is it and how precise is it with very low currents? The use case I think about is simulating devices that have sleep time followed by sudden short wake-ups with peak consumption (think bluetooth, lora, wifi, etc... based sensors, smartwatch, wireless headphones and many more. Would be nice to be able to simulate such a device to test batteries or solar power/charger/battery combinations.
Thank you, that's very nice to hear!
There is power motoring that makes sure it stays bellow 300W, and there's also overtemperature protection, that turns the load off if the heastink reaches certain temperature.
I didn't plan on adding remote controll for this device, because I don't see a need for that. Maybe I'm wrong?
It can switch currents fairly fast, though I haven't recorded any measuremets of that yet. It can go down to around 4mA and I guess it would fit the application of simulating devices that you mentioned. If minimum of 4mA isn't enough, this project could be easily modified to operate with smaller currents. This would of course come at a cost of smaller current range.
Please make a video about 16bit adc voltage current measurement..
You may want to look at the INA228 it contains all the hardware needed to measure current and voltage plus it calculates power , energy and charge with high accuracy or for a pre built unit the Junctek KL140F saves having to write software to do it and they contain 20 & 24 bit adc's which alone would cost more than these devices.
Thank you for the suggestion! This IC seems great, though I must say that the ADS1115 does the job and is a bit cheaper.
The ADS1115 may be cheaper but it is only an ADC the INA228 calculates the current , power , watts , battery capacity , watt hours to a high accuracy no calculations are needed in your own code @@dominikworkshop6007
Good Project, Can you explain how ADC is designed and working and explain working of electronics similar to your Power supply design explanation.
Thank you! I might do that in the future. It is very likely that there will appear some documentation on the github repository.
Can resistors R4, R5, R9, R10, R14, R15, R19, R20 be replaced with 0.5ohm resistors?, or can they be replaced with other types of resistors that are easier to find? Thanks sir!
Yes, they can be replaced with 0,5 ohm resistors, but that would mean more power loss on these resistors, unless the maximum current is reduced. They can certainly be replaced with other types of resistor, but I suggest ensuring that they have low thermal coefficient.
Wow, very nice project, the final construction is gorgeous. What is the cost of the enclosure ? Thanks
Thank you!
I don't know the cost of the enclosure, since I had it manufactured for free.
What power mosfets are you using?
IRFP250
Hello Dominik, great work🎉. I do have a question for you. There is a labeled pin as “OFF” in the schematics. Since this pin is same in every mosfet and controlled by 1 output pin so how do you calibrate each mosfet seperately? Thanks!
Thank you!!!
I don't have to calibrate each mosfet seperately. The op-amps ensure that the current is very evenly distributed. I used 4 resistors to sum the voltage on shunts, but it turned out unnecesary (maybe except for when there could be some hardware failure), because the op-amps are so precise, that the voltage on shunts is pretty much identical. Hope I managed to clear this out.
Very nicely done indeed!
Did you take a closer look at the dynamic behavior of the load? I found that it's highly recommendable to add a RC snubber to the input of the load. (How) did you address issues related to unregulated operation, i. e. when you disconnect the power supply from the load while the load is still switched on? Without any counter measures, connecting a sensitive power supply in the unregulated state could end catastrophically for the power supply. On the other hand, connecting low impedance/high energy sources might even damage the load itself. (Assuming a minimum resistance of potentially significantly less than 0.5 Ohm, a 50V power source could momentarily force currents in excess of 100A through the load.)
In the BOM it looks like you're using IRFP250. 300W is pretty high for four of these MOSFETs, especially if reliability is a concern. Unfortunately the safe operating area isn't specified for DC at all and for diy builds it seems impractical to characterize the MOSFETs. As a starting point I'd look at commercial implementations: Most loads that use these or similar MOSFETs use 8 to 10 in parallel for 200 to 300W total (they usually allow for higher voltages at the input though, often 120V to 150V).
I'm in the process of building my own low power (about 50W) precision DC load with two current and two voltage ranges. It's soo much work...
Thank you!
Yes, I checked how it responds to sudden input voltage changes and current changes. I made sure that there are no oscillations. The load looks very stable. Regarding connecting/disconnecting a power source when the load is on - I just assumed that you have to turn the load on when the power supply is already connected, because I don't see how this could be solved otherwise. I wonder how it is in proffesional loads.
One thing worth mentioning is that the current sink circuitry is very fast, so the load would short the loaded power supply for just a brief moment (though it's obvious that this could still damage the power supply or the load itself, as you said), which isn't so obvious, becouse I saw designs where the power MOSFETs were controlled directly by some kind of DAC, so there was a lot of time between connecting the power supply and microcontroller recognising that the current is WAYYY too high.
The 300W rating of this load isn't like a continuous rating, it's more like how much power you can dissipate for a few minutes (assuming that you start with a cold heatsink). I agree that if you use it at it's max power it could fail after time. I was planning on using more MOSFETs in parralel, but that would require balancing resistors, which I was avoiding, because I wanted to minimalize the minimum input voltage for full 8A. I'm curious how they do it in proffesional loads, because I'm controlling each MOSFET individually with an Op-amp, which ensures that each transistor dissipates the same amount of power. But I'm pretty sure that's not how it's done in commercial loads. In DIY designs I saw, the MOSFETS were simply parralelled with an equalizing resistor in the source path. This makes a compromise between the even current distribution among the transistors, and the minimum input voltage for full current rating.
I'm curious how your project will turn out!
And I strongly agree, that building a nice load is quite a hassle.
@@dominikworkshop6007 In principle, you're absolutely right: One should never disconnect or connect a power supply to the load while the load is switched on. However, in normal use, one might accidentally forget that the load is still on. You're right again saying that the short circuit lasts only a short period of time. I would argue that "short" is relative and the poor opamp has to drive the gate (with its capacitance) from about opamp V+ (e. g. 12V) down to about 3-5V. Also, MOSFETs have a tendency to turn themselves on when there is a large dV_DS/dt due to the internal capacitance, i. e additional charge is introduced to the gate. This makes the problem potentially worse. So the opamp's slew rate, the opamp's output impedance, the "speed" of the control loop and the gate resistor will likely determine how bad the overshoot will be.
In my opinion it's all but impossible to avoid the problem completely. E. g. the load can't distinguish between a power supply being disconnected (effect: voltage decreases) and the voltage being decreased on purpose. What is an option though: Reducing the effect of this issue.
I have a HP 6060B (schematics available: google HP 6060b Service Manual) which uses a pure analog control circuit for CC, CV, CR (CP is missing completely). They implemented a fairly complex analog circuit to detect the unregulated state (signal UNREG) and likely limit the conductance of the MOSFETs in this situation. Still, connecting a power supply with the load enabled leads to a certain overshoot relative to the programmed value. For my design I use a very similar approach: I effectively limit the gate voltage to a certain value that depends on the current setpoint. (Since the MOSFET's gate voltage varies between different samples and is highly temperature dependent one has to be careful not to limit the voltage too much.) With this the maximum gate voltage would be rather between 6 to 8 V instead of say 12V. This reduces the overshoot massively since the opamp doesn't saturate (hence eliminating the significant recovery time) and doesn't have to work as hard to bring the circuit back into regulation.
In addition it might be a good idea to sample the voltage continuously and switch the load off if the voltage falls below a certain disable threshold (e. g. 0.3V). Using a hysteresis and a time delay one could re-enable the load after a certain period of time if the voltage rises above an enable threshold (e. g. 0.5V). This feature could also be made optional. I think this is something that you could implement fairly easily without any hardware modification.
Pretty much all commercial designs I know of use one opamp per MOSFET plus a current measurement resistor. Opamps are cheap and this is the only way to achieve optimal load sharing. However, some use inverting, some non-inverting topologies. Some loads use an additional current sensing opamp per current measurement resistor and sum up the currents (or rather voltages proportional to the currents) afterwards (e. g. HP 6060B), while others measure the total current across a high power current shunt in addition to the local current regulation (e. g. BK Precision 8600 series).
@@sebastian_harnisch In my load, the maximum gate voltage is only 5V (due to max supply voltagee of the op-apm) which limits the maximum current, but of course as you said, due to the miller capacitance as the Vds rapidly raises, the gate voltage can raise as well.
Now that I think about it, turning the load off when the voltage is close to zero is a great idea, if this feature is made optional. I will definitely implement this in the next software update.
I'm pleasently surprised that commercial loads also use one op-amp per transistor. Now I think that I could try using some more generic op-amps for regulating the current through each MOSFET separatelly, and using a precision op-amp for measuring the total current through a precision shunt resistor. My only concern is stability, I'm afraid that using so many op-amps might produce too much gain and the system could get unstable. Mayby by creating suitable negative feedback with RC networks would do the trick, but I'd be afraid that the phase and/or gain margin would be small, and I could assume it's stable, but with some environmentall changes, the design could get unstable. I'm not quite sure how to get bode plots in such device to check it's stability. Do you have any tips for that?
Thank you soo much for your feedback! It is very valuable for me! (and sorry it took me soo long to reply)
@@dominikworkshop6007 Ah, interesting. 5V might be fine, I guess you have checked it, but depending on the conditions (temperature etc.) it might not be quite enough to achieve a low Rds(on). (If not for normal use a slightly higher voltage might help with the „short“ feature many commercial loads offer. On the other hand that might be something you‘re not interested in.) Just lowering the voltage imho is only part of the solution. The saturation issue is at least as important I think, so the way to go is to both reduce the gate voltage and make sure that the opamp never saturates. Some opamps might be less susceptible to this issue than others, but it might be better to just implement a circuit that avoids this situation completely.
Yep, I use this approach in my design. I think this makes the whole thing much, much more complicated and stability is a huge concern. (Just having many opamps doesn’t necessarily increase the loop gain though.)
A good starting point would be to simulate the circuit in something like ltspice (or maybe qspice, seems to be a good/better alternative created by the original author of ltspice) where you can do bode plots reasonably easily. I spent more time with that than I care to admit ;) This though helps enormously to figure out, what might work and what not to try at all. I optimized the inner loop first and only then proceeded with the complete control circuit.
I then used a simple bread board prototype to take some measurements with my Siglent SDS 2104 plus scope. It has the bode plot feature. However, I mostly modulated the setpoint of the control circuit and measured the current through the current shunt to determine the transfer function/bandwidth. I did not inject the signal directly into the feedback loop (via an isolation transformer) and did not measure before and after the "injection point" which you would do for actual stability measurements (gain/phase margin). What I did do was looking at the step response etc. There are some videos on TH-cam specifically about measuring stability of power supplies etc with frequency response/low frequency network analyzers (in case you have the equipment available to you) which might help. I currently don't have the setup (suitable isolation transformer), although I was thinking about building one for stability analysis at some point.
Getting the loop gain right is important. The loop gain can be adjusted for example by varying the gain of the current sense amplifier. DC loads often feature two current ranges, but they only attenuate the setpoint by 10x and don’t switch to a different current shunt, hence the loop gain stays the same - easy. It gets much more complicated if you change the current shunt, because then the loop gain changes and you have to switch *something* to achieve a lower gain… During the simulation I found quickly that I have to use compensation for the „outer“ control loop (integrator/compensator) as well. This helps massively with the step response. Also, as mentioned before, having a snubber circuit at the input is really important (if not necessary in many cases). This is especially true if you want to build a reasonably fast load.
Regarding loop stability vs environmental changes (temperature): I would argue that resistors and capacitors (especially C0G/NP0) are relatively stable over temperature and the MOSFETs are probably the main concern (or more specifically their transconductance). The inner control loop should be designed to handle the variations between two or more mosfets of the same type anyway, so the temperature is just another thing to look out for… Haven’t done extensive testing on my design yet. Simulations can help here too, but not all models may reflect the true behavior well or at all.
In the git repo you mentioned that you'd like to add reverse polarity protection. What do you have in mind? The load is kind of protected already - the body diodes will limit the reverse voltage to a low value and are conveniently placed on a heatsink. Some commercial products seem to rely on this form of protection (e. g. HP6060B). Also typical for power supplies. I see the disadvantage that this potentially "shorts" the power supply.
@@sebastian_harnisch I don't mind MOSFETS not being able to short the input completely, I didn't aim for that.
When it comes to simulation in LTspice, I've had problems with importing simulation models of elemnts used in my projects. I guess I have to play with it a bit more. But the most confusing part for me doing the bode plots, I managed to get some sensible results, but I'm just never quite sure if I got everything right (eg. if I injected the signal in the right place?). Thank you for all that information, I will try implementing all the avice I got!
I find such reverse polarity protection, that shorts the power supply, to be a bad protection, I wanted to implement something, that would prevent reverse current flow alltogether. Of course I wouldn't want to put a diode in series with the load, but I'm struggling to find a solution for that.
What is the maximum current at 12v?? 8A too?? or 25A??
It would still be 8A.
Can i have the gerber for current updated main pcb? I dont use eagle, so i dont know how to export gerber.
There is a gerber file on my github repository. There were no major PCB changes, and this gerber file is the most recent version of the main PCB.
Jakaś seria z cyklu "Jak projektować układy elektroniczne"? Zwłaszcza w 3D :)
Może w przyszłości kiedy będę mieć wiecej czasu o tym pomyślę ;)
@@dominikworkshop6007 spoko dzięki za odpowiedź 👍
There is no BOM file in github, sir
Very nice work. I would like to make one myself, but I don't know how to have a PCB made. Shouldn't you have one piece of PCB?
Thank you!
Having a PCB made is very easy. There is a gerber file on my repository (link in the description). You just have to upload it to a PCB manufacturer, like JLCPCB, and order it.
@@dominikworkshop6007 Thank you. I´ll try it.
@@dominikworkshop6007The PCB is ordered now, we just have to wait until it arrives. What program is used to upload the code? I can't find the list of components, maybe I don't have the right programs. Sorry for the questions, I'm a beginner, thank you very much
@@marecek19891989 I used platformio to upload the code. There are many tutorials on TH-cam.
There is a pdf with full schematic on my repository. You could also open this project in Eagle to be able to inspect what goes where more easily.
Let me know how this project is going for you!
@@dominikworkshop6007 Hi, my project has progressed. The PCB arrived from China looks very good. I am ordering components, but it is very difficult here in Slovakia. I can't buy measuring resistors :( where did you get them? Thank you
Did you use an input trigger? I mean define in defines.hh is wrong. It corresponds to ENCODER_BUTTON and the signal input trigger is connected to pin A7 on the PCB. It's just an analog input. Has it ever been tried?
Miroslav
You are right, thank you for pointing it out. I haven't used the input trigger yet, and there is an error in defines.hh, where I defined this pin to pin 3, instead of A7.
The two pins I know can work as external interrupts were used for the ENCODER_A and ENCODER_BUTTON signals. I'm not sure if I will be able to configure the A7 input as external interrupt, but in that situation I planned to just check the state of this pin frequently, and act accordingly. I know it's not perfect, but the uC used in this load has some limitations. If I were to make a V2 of this project, I would probably base it on some STM32 uC.
@@dominikworkshop6007 Maybe the PB3 pin can be used on the current system, which is still unused. It's a keyboard spare pin and MOSI too. And there is PCINT3 attached.