Dummy load questions: Regulation characteristics, Heat Sink

[Thilo78]

After digging up the old topic concerning the Dummy Load, I finally decided to go for a separate thread on this one.

I built this dummy load* as well, using a 9k/9k divider for the set voltage at the second opamp, and a IRFZ44N n-MOSFET.
My load consists of 10 resistors 10R/2W in parallel, which add up to roughly 1.2R.

First, I ran into the same problem with supply voltage and thus switched to a 9V battery supply.
With approx. 8 V available, the dummy load works fine and is capable of more than 1.5 A (that's the limit of my homebrew bench PSU).
I just added a fan to the flimsy aluminium heat sink that I had in my parts box.
With that I'm quite stable at 70°C at 1.25 A. Without fan I'm quickly exceeding 120°C at the MOSFET case.

Note: With the fan in parallel, I have an operating voltage of ~7.8 V DC.

But:
I wondered if the dummy load would regulate the current with variations in the tested power supply.
It does not.
If I dial in a current of 1.0 A at 10 V DC, and crank up the PSU to 15 V DC, the current rises to about 1.2~1.25 A.
(current measured with an inline multimeter in 2A range and another multimeter measuring the voltage across the load resistor)

I couldn't yet figure out, why that's happening.
It's not like the MOSFET was just acting as an adjustable resistor, but the current is not really as stable as I would have expected.

I was thinking, that maybe the power supply is too low at 7.8 V DC, but I have no second bench PSU to take that part.
So I'm trying to figure out the theory first.

Any ideas?
Or even better, advice?

Second thing:
Does anyone from Europe/Germany have a suggestion for a heat sink that fits the TO-220 package of the MOSFET and is suitable for mounting a 40x40 mm 12 V DC fan?
(Links to Reichelt/Conrad/HBE-Shop are very much appreciated)

Thanks!

Thilo

*: https://www.eevblog.com/forum/beginners/dc-dummy-load-problem/

[Christopher]

How about 2x9V batteries in series with a LDO to drop down to 12/15v ?

What opamp are you using? The mosfet Gate acts like a big capacitor and opamps don't like capacitive loads. The Opamp has an input offset voltage which could be effecting your output voltage. It may be worth using an opamp with rail-to-rail inputs and outputs with dual supplies. Sensing near ground (0v) is not really a good idea.

What is your Opamp analog supporting circuitry like? Better yet post a schematic


The mosfet you are using, the IRFZ44N, has a gate threshold voltage of about 4.0v maximum. So that's okay for being driven at a low voltage...

[Thilo78]

How about 2x9V batteries in series with a LDO to drop down to 12/15v ?

I thought about that. It's bound to be revision 0.2 of my dummy load.
I'd just like to know ahead if there's a chance to resolve the regulation thing.

What opamp are you using? The mosfet Gate acts like a big capacitor and opamps don't like capacitive loads. The Opamp has an input offset voltage which could be effecting your output voltage. It may be worth using an opamp with rail-to-rail inputs and outputs with dual supplies. Sensing near ground (0v) is not really a good idea.

I'm using LM324.
Actually, it's working fine. I can dial down to 0.0 set value and have a quiescent current of about 1 mA.
That's quite what I expected from the setup and I'm happy with that.

Also the range of up to 1.5 A is perfect for my needs (=urge to play with my PSU  )

What is your Opamp analog supporting circuitry like? Better yet post a schematic

As for the schematic it's exactly the one from the original post in the dummy load thread.
Please see my link in the OP.
I didn't sketch my own circuit and I wouldn't want to post pictures that are made by other users.

The mosfet you are using, the IRFZ44N, has a gate threshold voltage of about 4.0v maximum. So that's okay for being driven at a low voltage...

Sounds good.
Typical UGS is 2 V in the data sheet. That's the reason I chose it (next to availability)

Once again: The actual performance is spot on. I just wondered why the current wouldn't be regulated when I changed the input voltage.
I noticed that the sense resistor voltage is dropping as well by means of a 0.5 ratio compared to what I'd expect from a classical voltage divider.
In other words: If I drop the voltage on a voltage resistor by 50%, I expect to have the center point drop by 50% as well.
With the dummy load I see a respective drop of 20..25 %.
That's awkward and I interpret it as a sign of misconducting regulation.

[Christopher]

That opamp is gonna do whatever it can to the output pin to keep the voltage across the sense resistor the same as the reference voltage. 1V across the sense resistor is a 1A load.

If the output voltage deviates, the current should stay constant!

With a supply voltage of 7.8V, the opamp can only swing it's output pin to about 6V at best  (7.8V - 1.5V). This should be fine, as you are using a logic level fet, but it is worth measuring the Vgs of the fet while the load voltage is at 10V and 15V in case you are fully saturating the opamp's output. Also measure the voltage across the sense resistors (at the point you are tapping off) and the reference voltage.

I'd still recommend using a higher (regulated & decoupled) power supply... 10-15V would do the trick.


How about some pictures of your setup and a schematic? Your PCB routing may be to blame .

EDIT: The IRFZ44N has a maximum gate threshold of 4.0V whereas the IRLZ44N has a maximum of 2.0V. This may be of use,

[Thilo78]

OK, you called for it. 

First off, the schematic.
As I didn't draw my own yet, I took the one from Zitman in the original thread and added my components.

I decided to use 10 resistors of 10R each in parallel, as I could manage to get hold of a small batch of metal film resistors that seemed sturdy enough to handle my playing around and to tolerate some short-circuit action 

For the OpAmp I kept the 324 quad package, using two of the opamps for the voltage follower on the input and the actual regulator towards the MOSFET.

Also, I decided to go for a protoboard in order to play with plated copper wire. I like the looks and wanted to give it a try.
(btw: Yes, I did breadboard the circuit first, finding out that the IRFZ44N can take a lot of abuse  )

Pics attached.
Looks quite quirky, but I'm quite proud of my first own board, as I only needed two jumpers in total.
Okay, I later found out that one of them could have been routed better.

But actually it works and I'm quite happy.

Thilo

PS: Yes, the heat sink is crap. I salvaged that from an old PC power supply that blew up.
I'm going to get a better one as soon as I figure out what I'm doing here 

[Thilo78]

I also took several measurements.
Monitoring the load current with one UT71E, and the Gate-Source-Voltage with another UT71E (yes, my new meters arrived today, directly from China  ), I got the following results with some strange effects.

First off, load curve at 5 V power supply:
MOSFET starts to conduct at about 3.6 V
0.50 A --> 4.06 V UGS
0.75 A --> 4.19 V UGS
1.00 A --> 4.28 V UGS
1.25 A --> 4.34 V UGS
1.40 A --> 4.36 V UGS (last value is the max current of my PSU before it goes into CC mode)

Then the same with 10 V:
MOSFET starts to conduct at about 3.7 V
0.50 A --> 4.10 V UGS
0.75 A --> 4.13 V UGS
1.00 A --> 4.14 V UGS
1.25 A --> 4.15 V UGS
1.40 A --> 4.20 V UGS

And with 15 V:
MOSFET starts to conduct at about 3.6 V
0.50 A --> 4.01 V UGS
0.75 A --> 4.00 V UGS
1.00 A --> 3.99 V UGS (What's going on here?)
1.25 A --> 4.12 V UGS
1.40 A --> 4.14 V UGS

btw: Max voltage at pin 7 of the 324 is about 6.6 V (measured without power supply)

Going backwards for the regulation part, I gathered these values:
15 V --> 1.00 A set current, 4.00 V UGS
10 V --> 0.90 A, 4.13 V UGS
5 V --> 0.84 A, 4.21 V UGS

UGS seems to be rising still, but the current doesn't change.
Maybe the opamp is too slow? 

Next step will be to increase the aux voltage.
I found a wallwart supplying roughly 13 V DC, which I will modify to supply the circuit.
(must be a good one. It's from HP. So I'll modify a HP PSU for this )

[TerminalJack505]

Are you sure your circuit isn't oscillating?  A lot of people have had problems with stability with this particular circuit.

Do you have a scope you can use to check for oscillation?  If not then I'd find one of the circuits in this forum that has frequency compensation, such as Jay_Diddy's. 

I've attached the circuit that I used when I built the circuit which is similar to Jay_Diddy's but probably isn't as good since he's a much better circuit designer than me.

[Thilo78]

Bang, you're right.
I wasn't aware of that issue until now and used my scope to check for load voltage (across the 1R load resistor) and UGS. Both are oscillating nicely.

Apparently that's my problem there...

Would have been too easy 

At least the circuit itself is usable as long as I don't have to care about oscillation too much.
Next thing will be to throw in the filtering components.

[Christopher]

Small gate resistor (10-100R) to slow the whole thing down and local negative RC feedback should stop it oscillating .

[Thilo78]

I've attached the circuit that I used when I built the circuit which is similar to Jay_Diddy's but probably isn't as good since he's a much better circuit designer than me.

Nice!
Just did a little more fiddling with the dummy load.

Converted my HP wallwart in a regulated 12 V power supply, and worked with that.
(also on proto board)

Next, I added the 2 1k Resistors and the 3 caps to my board.
After checking the function I hooked up my freshly refurbished Fluke 93, and the oscillations are gone.

As for the function: When I change the voltage of the PSU under test, the current remains stable within around 20 mA.

Very nice!

Thanks for the help.
I think I'll go on with this design.

Thilo

PS: Sounds strange for a guy of 35, but this was the first electronic thing I built from scratch. So I'm kind of proud 

[TerminalJack505]

   Right on!

From what I remember when I built the circuit, the compensation cap (C1) can be made smaller.  I don't remember exactly but I likely used a cap that was 10x bigger than necessary just to be on the safe side. 

Using a smaller cap will give you better bandwidth (the circuit will settle faster) but watch out for oscillation.  Instability will likely first show up at one of the two extremes (low current or high current) so if you try using a smaller compensation cap be sure to check for stability at the extremes. 

If you have a sine wave generator then you can hook it up to the non-inverting input and send a low frequency signal through it and watch the response on the oscilloscope.  If the output starts getting fuzzy then you've made the compensation cap too small.

You can also check the response by sending a low frequency fast rise time pulse through the non-inverting input and watch the output for any overshoot or ringing which will likely indicate instability.

Probably the ultimate way to check the circuit and get the optimal bandwidth is to use a network analyzer and get a Bode plot.  This will show a circuit's bandwidth and phase margin.  I've got a Digilent Analog Discovery that comes in handy for that sort of thing.

[Thilo78]

OK, now for step two:
How would I proceed to parallel two of the MOSFETs?

First idea would be to keep the sensing resistor, and spend a third opamp for the second MOSFET, including the same circuitry with the 1k resistors and the cap in the output and sense line (output and negative terminal of the opamp).
Positive input to the two opamps would just be branched off of the existing output of the voltage follower directly.

Would this work to share the current between the MOSFETs equally?

[Jay_Diddy_B]

Hi,

You have the right idea. This is the way that I did it:



HP and many other companies do it the same way. An op-amp is used for each MOSFET to force the current to share.

The circuit was shared in this thread: https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/

Good luck with your project.

Jay_Diddy_B

[Thilo78]

Hi,

You have the right idea. This is the way that I did it:
...
Good luck with your project.

Jay_Diddy_B

Thanks a lot!
With one MOSFET it's working really great.
I'm still lacking a proper heat sink, but have it working with 30V/5A for a short time.
Those MOSFETs are pretty impressive little beasts 

Next step is using a dumpster dived metal case for it. In the course I plan to add a solid heat sink and double up the MOSFETs.
The latter is a step towards extending life time. I don't think it's really necessary, as the MOSFETs can take a lot, but I think it's fair to share the load. 
Also, I will add an ammeter and proper banana jacks to it.

[rob77]

OK, now for step two:
How would I proceed to parallel two of the MOSFETs?

First idea would be to keep the sensing resistor, and spend a third opamp for the second MOSFET, including the same circuitry with the 1k resistors and the cap in the output and sense line (output and negative terminal of the opamp).
Positive input to the two opamps would just be branched off of the existing output of the voltage follower directly.

Would this work to share the current between the MOSFETs equally?

1.  simply paralel the mosfets - the traces should have the same length (or wires) and ideally you should pick ones with same/similar Ugs vs Rdson characteristics. i've built mine with 4x IRFZ34N and i didn't bother to measure them despite the fact that i have a lot of them on stock (but i had all the mosfets from the very same production batch)

2. heatsink with fan ? nothing better than the old CPU heatsinks with fan the ones for old socket370 cpus are good enough. i've used a nice black fanless heatsink from a Pentium 2 processor, with a standard socket370 heatsink with fan bolted to it. it's a lot of aluminium

3. oscillation - mine was a beautiful powerful almost-sine-wave generator (i was even tempted to wire a transformer in series just for fun  ) during the initial tests (using a unregulated wall wart adapter to power the opamp) - btw.. the oscilations have apparently something to do with the positive feedback through the + rail. 
in the final version (still using that wal wart adapter)  i've added ar RC filter to the opamp output (kill me for that if you wish  ) and the set voltage is coming from an ATmega (arduino nano) - the PWM is 490Hz with 3rd order RC filter. basically the set voltage is quite well decoupled from the + rail of the opamp (atmega is powered from a linear regulator connected to that + rail) - and there is not a single sign of any oscillation - the ripple current through the load is non-existent

by final version i mean the circuit , the device is still not finished - the power module is finished (heatsinks, power mosfets, thermistors, comparators and mosfet for on/off control of the fan) and the rest is still on a breadboard ( LM324 + passives, 1602 LCD display + arduino nano) once done, i'll share it including the software. i posted the picture of the power module in one of the other dummy load thread.

[Thilo78]

1.  simply paralel the mosfets ...

Thought about that as well.
I'll maybe try. But asking before you produce (unneccessary) magic smoke appeared smarter to me 

2. heatsink with fan ? nothing better than the old CPU heatsinks with fan the ones for old socket370 cpus are good enough. i've used a nice black fanless heatsink from a Pentium 2 processor, with a standard socket370 heatsink with fan bolted to it. it's a lot of aluminium

Also considered that.
But I have this nice case, and it would fit a 140x40 mm heatsink just perfectly. Maybe I can find something at Conrad.
Shopping time tomorrow after work 

3. oscillation - mine was a beautiful powerful almost-sine-wave generator (i was even tempted to wire a transformer in series just for fun  ) during the initial tests (using a unregulated wall wart adapter to power the opamp)

Oh yeah. I had this one, too.
I wonder how much you would have to add to smooth that out.
With the wall wart the circuit goes crazy. With my (new) linear power supply it works like a charm.

( LM324 + passives, 1602 LCD display + arduino nano) once done, i'll share it including the software. i posted the picture of the power module in one of the other dummy load thread.

Yes, please share the project.
I still wonder if I want to invest the time to digitalize this dummy load. 

I think I have all the parts necessary, but currently I don't see the point (except the usual "'cause I can!")

[mariush]

Isn't there a risk with parallel mosfets due to the high input capacitance of some mosfets?  Two or more mosfets would only increase the input capacitance... I don't know off the top of my head how to figure out if an opamp is capable of driving a particular capacitance, maybe Jay_Diddy_B or someone more experienced could add their suggestions.

Maybe a better idea to use a quad opamp and use one opamp for each mosfet, each getting feedback from same resistor?

[Thilo78]

Maybe a better idea to use a quad opamp and use one opamp for each mosfet, each getting feedback from same resistor?

I'm not sure about the capacitance. So I'm going the way with separate opamps.
I'm using LM324 quad anyway, so it has another benefit:
I like the voltage follower on the input, so I'd use one opamp on one side for the input and the other two on the opposite side for the MOSFETs.
This gives me a nice setup on the board as well, as the output part can be built symmetrically.

[rob77]

Yes, please share the project.
I still wonder if I want to invest the time to digitalize this dummy load. 

I think I have all the parts necessary, but currently I don't see the point (except the usual "'cause I can!")

sure i will once finished. what i can share right now is a few pictures (pics taken with my phone - sorry for the poor quality and reflections)
on pictures you see the power module construction - heatsinks with fan , big ass 50W 1R resistor, thermistors glued (on on resistor , one near mosftes) and a small protoboard with lm393 dual comparator and bs170 switching the fan

i'm sensing the temperature on 2 places because at low voltages and high current the resistor is dissipating the majority of the heat while on higher voltages and lower currents the mosfets are taking the load.

the last picture is showing a run when it's eating 75W (it goes approx 40C above ambient during long runs with such a load).

[rob77]

Isn't there a risk with parallel mosfets due to the high input capacitance of some mosfets?  Two or more mosfets would only increase the input capacitance... I don't know off the top of my head how to figure out if an opamp is capable of driving a particular capacitance, maybe Jay_Diddy_B or someone more experienced could add their suggestions.

Maybe a better idea to use a quad opamp and use one opamp for each mosfet, each getting feedback from same resistor?

actually the gate capaciance is not an issue at all in this case. for example i'm using 4x IRFZ34N - which is 4x 700pF max gate capacitance - therefore almost 3nF. the mosfets are working in the "linear" mode , they're not rapidly switched on and of => no big gate currents. furthermore i have a RC low-pass (47k + 200nF) on the opamp's output - so the 3nF gate capacitance is "half a bee's dick" in that case

4 opamps with 4 mosfets sharing 1 current sensing resistor ? well... actually... how would those distribute the load evenly ? if one opamp is faster than the others - it will drive it's mosfet to sink all the current to achive the goal - and the others will do nothing - because the voltage drop on the sensing resistor is already ok.
please correct me if i'm wrong...

[Jay_Diddy_B]

I don't know off the top of my head how to figure out if an opamp is capable of driving a particular capacitance, maybe Jay_Diddy_B or someone more experienced could add their suggestions.

I have posted a reasonably detailed analysis on how to design a stable load circuit and what the effect is of the gate capacitance in this thread: https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/msg462562/#msg462562

(I put it my original thread, because it may be a stretch for the beginners section).

Essentially the input capacitance of the MOSFET and the gate resistor form a pole. This is (should be) at a higher in frequency than the dominant pole made with the feedback around the op-amp.

Regards,

Jay_Diddy_B

[sangseu]

Hi everyone.
Please look at my circuit, it has a problem as show in picture:

in picture, Column A is raw ADC data measuring current via 5 resistor 0.5R. Column C is V_DS. Column E is raw PWM DAC value.
When PWM DAC value increase from 0 to 47, current increase as calculated. But, when PWM DAC value increase one step, to 48, current has a big jump up from 33 to 180. At the end of chart, current increase suddenly to 0.
Can anyone clean my problem and have solution ?
Here is schematic:

Sorry for my bad English.

[Jay_Diddy_B]

Hi,
Welcome to the Forum!!

We need to decide if this is a software problem or a hardware problem.

I would fix the voltage at the top of RV1, the PWM signal, with a d.c. voltage and use the potentiometer RV1 to control the output.

May be you can try this and let us know what happens?

Regards,

Jay_Diddy_B

[sangseu]

Thank @Jay_Diddy_B. I'm going to do your suggested then post the result as soon as possible.

[sangseu]

Hi. I run my circuit with DC voltage and change RV to regulate current. Picture show the result:

You can see current jump a big step again when a increase I_set.

I'm using some MOSFET 12N60C to sink current. DC 3.3V, RV 10k, atmega328ADC to tracking U, I, I_set. Using Fluke 15B to tracking current across MOSFET and see current. I set I_set to 0, power on V_DS, and increase I-set. Current increase and has big step from about 0.1A up to 0.9A so I turn off V_DS.