Yet another DIY Electronic Load

[OM222O]

6.2k ... that's just ridiculous. Look at the time constant of that Rg and Cg.  But, as long as you are just designing load with 100Hz bandwidth to discharge batteries, then good enough.

can you please be so kind as to let us know what else you use your loads for? clearly not discharging batteries and testing power supplies! 
people give advice, yet you tell everyone they are wrong.either give constructive comments or stop whining about everything  he said he tested it and it worked fine, I tend to agree. even 10k resistors are usually fine for this application!

[MarkF]

MOSFET = IRFP064

F = 2 * 3.14159 * 6.2K * 10000pF = 2567Hz

Just how fast does it need to be?

[Yansi]

6.2k ... that's just ridiculous. Look at the time constant of that Rg and Cg.  But, as long as you are just designing load with 100Hz bandwidth to discharge batteries, then good enough.

can you please be so kind as to let us know what else you use your loads for? clearly not discharging batteries and testing power supplies! 
people give advice, yet you tell everyone they are wrong.either give constructive comments or stop whining about everything  he said he tested it and it worked fine, I tend to agree. even 10k resistors are usually fine for this application!

I am just wondering, how you are going to test a power supply transient response with a load that has a loop bandwidth of couple tens of Hz?  You know, loads are also commonly used in pulsed mode.   100Hz loop bandwidth ain't cutting the mustard even for testing basic off-line SMPS circuits.
It may be surprise for you, that static load testing for any (closed-loop controlled) power supply ain't enough.

Hence why I think ya should make the bandwidth large enough, or at least as good as it'll get.

Design decisions such as 6kohm gate resistors for 10nF mosfets are just wrong.

You have in this thread even a couple of circuits from commercial solutions, so why not inspiring there?

Also, there is this good appnote from Jim Williams: AN133, A Closed-Loop, Wideband, 100A Active Load.

[Jay_Diddy_B]

I am just wondering, how you are going to test a power supply transient response with a load that has a loop bandwidth of couple tens of Hz?  You know, loads are also commonly used in pulsed mode.   100Hz loop bandwidth ain't cutting the mustard even for testing basic off-line SMPS circuits.
It may be surprise for you, that static load testing for any (closed-loop controlled) power supply ain't enough.

Hence why I think ya should make the bandwidth large enough, or at least as good as it'll get.

Design decisions such as 6kohm gate resistors for 10nF mosfets are just wrong.

You have in this thread even a couple of circuits from commercial solutions, so why not inspiring there?

Also, there is this good appnote from Jim Williams: AN133, A Closed-Loop, Wideband, 100A Active Load.

Yansi,

There was no mention in the OP message that he was building a load for doing transient load tests on power supplies. It is possible that high loop bandwidth is not a requirement. Loop stability is always a requirement.

I presented a fairly extensive coverage on the loop stability of this type of load in this thread:

https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/

I have obtained bandwidth of around 70kHz for a 5A load and 350kHz for a 1A load.

(To get the high bandwidth, in the low power load,  I used relatively small MOSFETs to get around the problems caused by large Cg. We do not need to deal with the Miller capacitance, Cdg, because the voltage across it should be constant.)

I did not attempt to protect the op-amp from MOSFET failure in these designs. This is common practice in commercial loads. There is no evidence of op-amp protection in either the BK or the HP 6060 loads.

If I wanted to protect the op-amp, I would do something like this:





I am showing that the protection circuits I have added, the transorbs and the small signal diodes have not really compromised the control loop, especially for a static load. I would do not do this for a dynamic load.

I obtained a loop bandwidth of 10kHz without any effort, with the option to increase bandwidth at the expense of phase margin.

I used the LT1014, which is an LM324 with better offset and similar GBW product.

The MOSFET was chosen from being in the LTspice library.


Note: I have included the lead inductance in the model. As a countermeasure for the lead inductance, the snubber circuit is required. The snubber circuit can be found in the BK and HP loads.

I have attached the LTspice model.


Regards,

Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

I just want to comment on the Jim Williams AN133 circuit:



This load was designed to do a specific task. It was designed to test low voltage, around 1V, high current power supplies. It was designed to be very fast, high bandwidth. The design is limited to low duty-cycles. The maximum duty-cycle is 1.5% at 100A.

By limiting the duty-cycle and the on time, Jim was able to use MOSFETs that are much smaller than would be required for 100% duty-cycle applications.


To get the waveform fidelity shown in the application note:



Jim had to use 10,000uF of low impedance capacitors.

This picture shows the effect of adding just 20nH between the load and the power supply:





Great work for the intended application, but not a good general purpose load.

Regards,

Jay_Diddy_B

[MarkF]

6.2k ... that's just ridiculous. Look at the time constant of that Rg and Cg.  But, as long as you are just designing load with 100Hz bandwidth to discharge batteries, then good enough.

can you please be so kind as to let us know what else you use your loads for? clearly not discharging batteries and testing power supplies! 
people give advice, yet you tell everyone they are wrong.either give constructive comments or stop whining about everything  he said he tested it and it worked fine, I tend to agree. even 10k resistors are usually fine for this application!

I am just wondering, how you are going to test a power supply transient response with a load that has a loop bandwidth of couple tens of Hz?  You know, loads are also commonly used in pulsed mode.   100Hz loop bandwidth ain't cutting the mustard even for testing basic off-line SMPS circuits.
It may be surprise for you, that static load testing for any (closed-loop controlled) power supply ain't enough.

Hence why I think ya should make the bandwidth large enough, or at least as good as it'll get.

Design decisions such as 6kohm gate resistors for 10nF mosfets are just wrong.

To start, I have NO need to test power supply transients.
Second, you keep spouting 10s of Hz bandwidth.  Show me the math!

[2N3055]

Maynuo 9812 has 25kHz max pulse repetition rate, and can achieve slew rates of 2.5A/usec.
That is sufficient for testing "normal" PSU, LED drivers, and such.
For testing special fast drivers you need something faster, like that in Jim Williams appnote Yansi linked.

You can use load for static test, like battery discharge, or testing power supply thermaly.
You can use load to pulse test output of PSU to see regulation, recovery time, to see if it's stable (no overshoot or undershoot, no oscillation) etc. For that, load has to be faster than DUT's regulation loop, preferably order of magnitude faster, so it will look to DUT as something called "step response" (a term from regulation theory). That "step response" will reveal how the DUT responds to such a rapid load change. It will show DUT step response function that tells us a lot about it's regulation, stability, speed etc.

That is why active loads have limited use if they are very slow.
Which is not  problem if you don't need dynamic load testing.

But a word of advice: Jay_Diddy_B made his load, he has shown design, calculations, explanation of loop stability, what it is and how it is calculated. There is literally nothing to be said more on the topic after his very detailed postulation. So use his design for analog part and just replace signal generation part. He is extremely nice gentleman and if you get stuck I'm sure he'll help you if he can. He has already posted few posts here..

[Yansi]

Jay_Diddy_B: No offense sir! But noone prevents anyone to  use a different mosfet with apropriate heatsinking, different current range, yet keep he same circuit tricks from there, like for example the gate buffer stage, feedback network + compensation type, etc.

[OM222O]

Jay_Diddy_B: No offense sir! But noone prevents anyone to  use a different mosfet with apropriate heatsinking, different current range, yet keep he same circuit tricks from there, like for example the gate buffer stage, feedback network + compensation type, etc.

you are so ignorant and delusional! he has explained EXACTLY WHY he did what he did and RAN ACTUAL SIMULATIONS compared to your bluh bluh bluh, I am such a great engineer and I KNOW EVERYTHING, SO YOU ARE ALL WRONG AND WHAT I SAY IS RIGHT. responding to you is worse than banging my head against a brick wall, because at least the brick wall breaks at the end, but you keep on going  there are no rhyme or reason to your comments, rather you just keep saying everyone is wrong. please design a load with full schematic and post it, then add comments as to why you designed it the way you did. otherwise stop shitposting and wasting everyone's time. no wonder you have 2.3k posts!

there's actually a very good saying about arguing with people like you which unfortunately I can't use here ...
I'm really not sure what you're trying to compensate for, but nobody really cares how great you are. please keep your attitude for yourself and only post constructive comments.

there needs to be more people like john who actually take time out of their day to give good, detailed explanations which have been extremely helpful MULTIPLE TIMES, at least to me personally. I think others agree too.

[wilfred]

Jay_Diddy_B: No offense sir! But noone prevents anyone to  use a different mosfet with apropriate heatsinking, different current range, yet keep he same circuit tricks from there, like for example the gate buffer stage, feedback network + compensation type, etc.

Jay_Diddy_B has contributed some of the best quality posts on this forum that I have seen. I'm glad you don't want to offend him. I don't want you to either.

[MarkF]

For a reference, here is an enhanced version of the HP 6060 Load put forth by Jay_Diddy_B

[pinoccio]

This project is not abandoned
I was a bit overloaded with the mass of info and felt a little demotivated too, so took a break from this one...

Well, there is always something to improve in a circuit, better ways to do it, change a part for better performance, better this and that. But if I don't set and hold on to some boundaries, it's going to be an endless spiral of improvements and the project will never become something real.

I appreciate the efforts and time of everyone who posted something relevant in this thread. Especially those who contributed improved versions of my circuit or provided alternative ones, and actually explained how and why...
I liked the modified version from OM222O as a concept, even though I may not build it as the first prototype, took some ideas thou. Same for circuits from MarkF, Jay_Diddy_B, Yansi ...

Btw. never planned to test transient response etc. Might be nice, if the final circuit was somewhat capable if I learn more about that stuff in future. But for a beginners project, this is currently not considered.
Well, actually any further info and insight about dynamic loads, transients, loop bandwidth, stability, op amp protection, and so on - is obviously welcome ... For future study and also other beginners who might read this

Now here is my next version:

available also at the public project page: https://easyeda.com/pinoccio/Electronic-Load-DIY-open

- my IXYS fets arrived
- swapped op amps for possibly better ones
  //both ebay so ...hopefully no knockoffs or faulty rejects
- changed sense resistors to 1/3 R and related modification to divider
- 12V supply
- played with the current sensing part...

Tried to design the current sensing summing amp, so that it can still work with single supply... well, not really sure
If this is not a reasonable way to go, other options are:
- proper dual supply
- separate direct sensing of each channel (got the free ADC port)
- dedicated current sense module high/low side -like INA* or similar

[Inconel-oh-el]

I'm designing a fairly similar load at the moment. I'm not sure at all why Yansi pushed for low side current sensing. The shunt resistors will heat up, and that will affect accuracy (especially with .33R/2.5A). While testing a power supply for a period of time, you may notice the reading changing- you then have the hassle of determining if it is the supply or the load causing that.

Adding a summing amp/potentially dual rail supply is an inelegant, brute force solution.  High side current sensing would add a single op-amp and low value current sense resistor before the load, completely decoupling the current measurement from the current passing through the load. It is much simpler, more accurate, and also allows you to add any number of parallel paths later on without requiring a new PCB.

[2N3055]

I'm designing a fairly similar load at the moment. I'm not sure at all why Yansi pushed for low side current sensing. The shunt resistors will heat up, and that will affect accuracy (especially with .33R/2.5A). While testing a power supply for a period of time, you may notice the reading changing- you then have the hassle of determining if it is the supply or the load causing that.

Adding a summing amp/potentially dual rail supply is an inelegant, brute force solution.  High side current sensing would add a single op-amp and low value current sense resistor before the load, completely decoupling the current measurement from the current passing through the load. It is much simpler, more accurate, and also allows you to add any number of parallel paths later on without requiring a new PCB.

Low side, current sink design topology is industry standard that pretty much all professional loads use. Low side sensing allows current sensing without large common mode voltages, making current sensing more accurate and with larger bandwidth.  Self heating of shunts is universal problem that is addressed by using lower resistance shunts with better tempco.
There are many pro load schematics on the net. Study them, and you will see what they use..

[Inconel-oh-el]

Low side, current sink design topology is industry standard that pretty much all professional loads use. Low side sensing allows current sensing without large common mode voltages, making current sensing more accurate and with larger bandwidth.  Self heating of shunts is universal problem that is addressed by using lower resistance shunts with better tempco.
There are many pro load schematics on the net. Study them, and you will see what they use..

I've looked at several, I'm not disputing the low side current sinking. My design follows the same, as well as using low value shunt resistors, and thus an amplifier for each. Pinoccio's design is uses .33R so that no amplifier is needed for regulation, keeping things much simpler (and I don't see anything wrong with this for the intended purpose). However 2W (and especially 8W if he intends for 5A) will heat the resistors and cause readings to drift... I don't think anyone recommends building instruments that give you faulty, untrustworthy readings, no matter how simple or low accuracy they need to be.

Adding a high common mode range amp like the INA286 as recommended earlier in this thread (CMR of 80V) is still much simpler than one amplifier per resistor, and then trying to add those signals/using additional ADC's on the MC. Multiplexing may also work, but the point still stands. Obviously local pricing may vary, but the INA286 is not the only suitable option.

[pinoccio]

..next iteration
- slightly lower shunt
- split current sensing with pre-amp (but increases complexity)
- simple op amp protection


edit: updated the schematic pictures:
- fixed error/s
- added snubber
- lower caps for better bandwidth possibly
- other minor edits

Can the resistor heating be such an issue here?
Now the projected load is about 5A total, so 2.5A each branch, with dual shunt resistor we get 0.78W each... so I plan to use 5W aluminum shell -those should not heat up too much having such headroom.



Another possible improvement would be a better DAC /with more resolution, maybe DAC8571 (16bit). But I'll leave that for later as this is planned as a module plugged into a header, so it can be replaced any time.

[pinoccio]

...updated pictures in the previous post (just small corrections and updates)

[thomasx]

Hi

I do realize I am posting in a 5 year old thread. But there's good and interesting info in here from people with apparently the right skill set

I have some linear MOSFETs on order to start building my own electronic load. However, they are a week away or so, and eager as always I want to at least try this out in a simple setup right away.

Looking at the very simple setup in this post in the beginning of the thread

https://www.eevblog.com/forum/projects/yet-another-diy-electronic-load/msg2448360/#msg2448360

and checking my "stock" of components I find a BUZ10 that seems to have fairly good specs, looking at Output and Transfer Characteristics.

https://www.mouser.se/datasheet/2/389/stmicroelectronics_cd00000673-1204298.pdf

For initial tests I am thinking of sinking 1-2 amps at 5-12 volts, and then to go on from there with 12 volts and see how far I can get, with about 85Watts as maximum load, meaning that at about 7 amps the DUT voltage will start to drop to maintain the output power. But according to the SOA, the BUZ10 will give up before then. Step two I guess will be to try with two parallelled BUZ10, with an op-amp each of course.

Any thoughts on a simple test like this, with the BUZ10?