Problem with Constant Current Dummy Load design

[MrAureliusR]

I've been trying to build my own dummy load based on Dave's design (seeing as how everyone else it seems has built one), and I followed his schematic almost exactly. I'm using an FQP30N06L logic-level MOSFET instead of the MTP3055 that Dave used. Also, instead of an LM324 I'm using two LM741's. The other difference is I used a 100k pot instead of a 50k but that's not the source of my problems.

What seems to be the problem is that the second LM741 (the one connected to the MOSFET) is constantly outputting the highest voltage it can. This is because the inverting input (V-) is much lower than the non-inverting input. The non-inverting input (V+) is following the voltage from the pot perfectly, so I know the first op-amp is doing its job perfectly. I really can't figure out why V- isn't higher -- the MOSFET should be more than fully turned on (the op-amp is putting out something like 4.4V to the gate, which according to the datasheet is fully turned on -- the Vgs threshold is almost identical to the MTP3055 specs). So essentially, when I turn the pot -- nothing changes. The other problem is that even when the pot is turned all the way down the first op-amp will only go down to 0.9V (but I know this is because of the nature of an LM741 -- I suppose I should try and find a better op-amp for the job -- I've got a whole pile of different op-amps but I'm still learning about how to choose the right one for each job. In this case, just to get it working, I figured the LM741 would work)

Now, technically speaking, it does work -- the power supply has about 770mA being drawn from it constantly (there is, accordingly, about 0.77V on the source side of the MOSFET, which is, of course, connected to V- on the second LM741). So as a 'fixed' dummy load, it's working. But of course I want to be able to adjust it.

Can anyone see a glaring error I've made? I'm sure there is one... is it just that the LM741 can't handle the job? Or is there some spec on my MOSFET that I'm missing? Any insight at all would be much appreciated! I can draw a schematic if needed but honestly it's identical to Dave's minus the part changes. I also added an ON/OFF switch.

Thanks guys!!!

EDIT: Well part of the problem was because the power supply I was testing it on couldn't even supply enough current to get the inverting input up high enough for the op-amp to balance out. So I switched things around and now I have the power supply input connected to my 60W bench supply. Now with the current knob turned all the way down it draws 1.9A. Once I turn the current knob up high enough I am able to dial in current between 1.9 and about 2.2 (the limit of the voltage divider + the other losses). So now I'm wondering why it can't go below 1.9A? Should I run the op-amps from a split supply? I'm fairly sure that would help... I'm finally really wrapping my head around what is going on in the circuit. So if I were to run the op-amps from +/-5V, the op-amp would be able to 'pull' the output down more? Or am I missing something else... I know that this would solve the problem of op-amp number 1 above not being able to go lower than 0.9V.

However even with the second LM741 having 0.9V on the inverting input and 1.2V on the non-inverting input the output is still at the positive rail. This doesn't make any sense to me as my understanding of op-amps dictates that it should be trying to make its inputs equal.

[suicidaleggroll]

So you're pulling 770mA, the source of the fet and the inverting input on the op-amp are showing .77v, your pot is commanding >.77v, and the op-amp is saturated pushing 4.4v into the fet trying to increase the current.

Do I have all of that right?

If so, 4.4v should be plenty to drop the Rds(on) nice and low on that fet.  It sounds like it's working correctly and the problem is whatever you're using for your current source is dropping the voltage because it can't keep up with the load.  Have you measured the drain voltage at the fet while this is occurring to make sure it's still as high as you expect?  What happens if you lower the pot to bring the non-inverting input on the op-amp down below .77v?  Does it track correctly from there down to 0?

[sleemanj]

Do you have a scope, or are you just using a meter to measure voltages.

You might have a lot of oscillation going on and if so a digital meter might be hiding that from you.

[psycho0815]

From the lm741 datasheet you can see, that its output can only swing to within about 1V of the supply-rails, which is why you never get less that than 0.9V from your first opamp. try going directly from your pot to the non-inverting input of you second opamp. Of course that one has the same problem in theory, but 0.9V should be below the fets threshhold-voltage, so it should probably work. I'd also recommend to check for oscillations, that seems to be a common problem with that circuit.

[Pezo]

The input voltage range of the LM741 is also only to within about 2V of the supply, so without a split supply it won't go down to 0 either.

[rdl]

I have built that exact circuit three times and it has never worked. Every time it oscillates like crazy. If I didn't have an oscilloscope I would never have been able to make it work. Good luck.

[dannyf]

You can post the schematic so people can better help you.

Generally speaking, driving a big mosfet with a weak and slow opamp is never a good idea. A worse idea is to drive the same mosfet with a fast and beefy opamp.

[MrAureliusR]

I do have a scope, and I haven't looked at it yet, I'll do that today and see what's happening. Oscillation is definitely a possibility.

Why did Dave's circuit work so quickly and easily?

I guess I will also switch to a split supply.

[c4757p]

Why did Dave's circuit work so quickly and easily?

I'd say a combination of luck and experience - getting all the parasitics in helpful rather than harmful places, in part by knowing what to expect.

It's inherently a difficult circuit to stabilize. The op amp has very high gain, and you're asking it to maneuver the output over a very narrow voltage range (only a few mV on the gate will have a great effect on the current). That's like trying to control the speed of your car with a brick strapped to your foot.

You'd make it more stable by finding ways to decrease the gain. A large sense resistor would help, though it would increase the dropout. A BJT instead of a MOSFET with a base resistor chosen to utilize most of the op amp's output voltage range would also help. The circuit as it can be compensated, but it's tricky to do (and layout-dependent, so even if you do it on a breadboard you'll have to redo it for the PCB). You can start by putting C and R together in series, from the output to the -ve input of the op amp, and playing with both values.

[mariush]

I've built this current dummy load : http://www.sleepyrobot.com/?p=136 

(it's dave's load with minor changes)

on a prototyping board, changing the mosfet to a FQP50N06 and it works fine. The opamp is probably the issue in your design.

Got to the 3A limit of my linear power supply at higher voltages (10v+), goes to about 1 amp with small voltages (<5v).

[MrAureliusR]

Yes I just fired up the oscilloscope and there's almost 2V of oscillation on the output of the MOSFET, 700mV oscillation on the output of the op-amp.

c4757p, can you elaborate on what you're suggesting? Do you mean a cap and resistor in series between the output of the MOSFET (source) and inverting input on the op-amp? (thus stabilizing the negative feedback loop?)

And what do you mean by a large sense resistor? Where would this be put, and what would the op-amp be using it for?

Thanks for all the help guys, I really appreciate it!

EDIT: looking at mariush's schematics I see what he's done, 2200pF across the output and negative feedback, and 10k in series. I'll see what caps I have that are near that value and see if that helps. Is this what you mean c4757p?

PS. My 1 ohm resistor is a 25W power resistor, and the MOSFET has a huge heat sink on it, so if I get it working I should be able to sink a few amps. At the 2A it's stuck at it only gets mildly warm, not quite hot but close.

[rdl]

 I used an LM358 but the same MTP3055VL FET. I built mine on perfboard and had to make the same modifications to mine as the one linked by mariush. Cap from output of op amp to inverting input, resistor from output of op amp to FET gate, resistor from FET drain to op amp inverting input. What worked for me was 150 ohms from op amp to FET (more was actually worse at some points), at least 120 nf from output to inverting input (more was better so I used 330 nf), between 10 and 50 ohms from drain to inverting input. With no resistor from drain to inverting input it oscillated, over 50R and it oscillated, so I used 27R.

[c4757p]

c4757p, can you elaborate on what you're suggesting? Do you mean a cap and resistor in series between the output of the MOSFET (source) and inverting input on the op-amp? (thus stabilizing the negative feedback loop?)

And what do you mean by a large sense resistor? Where would this be put, and what would the op-amp be using it for?

EDIT: looking at mariush's schematics I see what he's done, 2200pF across the output and negative feedback, and 10k in series. I'll see what caps I have that are near that value and see if that helps. Is this what you mean c4757p?

I mean a cap and resistor together (in series) right where he has C4 on that schematic - just a cap is often fine, though I usually find things a bit more reliably stable with both.

As for the large sense resistor, I mean R1..R10 on that schematic, and 1 ohm is large, so I guess that suggestion is satisfied. Didn't have the schematic on hand, and for all I know you used 0.01ohm like I did in my dummy load

[MrAureliusR]

c4757p, can you elaborate on what you're suggesting? Do you mean a cap and resistor in series between the output of the MOSFET (source) and inverting input on the op-amp? (thus stabilizing the negative feedback loop?)

And what do you mean by a large sense resistor? Where would this be put, and what would the op-amp be using it for?

EDIT: looking at mariush's schematics I see what he's done, 2200pF across the output and negative feedback, and 10k in series. I'll see what caps I have that are near that value and see if that helps. Is this what you mean c4757p?

I mean a cap and resistor together (in series) right where he has C4 on that schematic - just a cap is often fine, though I usually find things a bit more reliably stable with both.

As for the large sense resistor, I mean R1..R10 on that schematic, and 1 ohm is large, so I guess that suggestion is satisfied. Didn't have the schematic on hand, and for all I know you used 0.01ohm like I did in my dummy load

Well I tried adding the 10k resistor and also a few different values of caps. I didn't have a resistor between the output of the op-amp and the gate -- adding one made the oscillations worse with capacitance or without.

I don't have a 2200pF cap, the closest I have is 10000pF (10nF) and that did reduce the peak to peak value of the oscillations by about %40 but it certainly didn't stabilize anything. I'll try the resistor and cap in series, any suggestions on values?

[sync]

PS. My 1 ohm resistor is a 25W power resistor

Likely a wire wound resistor. It may have too much inductance. This can cause oscillations. Try a normal resistor or a few parallel.

[c4757p]

PS. My 1 ohm resistor is a 25W power resistor

Likely a wire wound resistor. It may have too much inductance. This can cause oscillations. Try a normal resistor or a few parallel.

If that's the case, try putting something in the neighborhood of (100 ohms in series with 10uF) in parallel with the resistor.

[rdl]

I don't have a 2200pF cap, the closest I have is 10000pF (10nF) and that did reduce the peak to peak value of the oscillations by about %40 but it certainly didn't stabilize anything.

I would go even higher 200-300 nf or more. I would also suggest to replace the 10k resistor with a pot for testing purposes. If I hadn't done that I don't know how I would have discovered the odd value required to stabilize my circuit. Might even try a pot on the other resistor and a substitution box for the capacitor until you get it figured out. That's how I did it.

[Conrad Hoffman]

OK, enough suffering. Listen to the dinosaur. Way back in January 1983, Siliconix published their MOSPOWER design catalog. Buried in the back, on page 6-60 was the application circuit "High Compliance Current Load". I've drawn pretty much that. Just pretend the divider on the left has a pot in the middle to adjust the set point. Note that if you single supply it, you'll need an opamp where the input common mode includes ground, because the signal off that current sense resistor is next to nothing. LT1013 works well. The 100k is essential for stability. I've built this circuit with all sorts of MOSFETs and all different values of sense resistor. My bench unit is powered from a 9V battery. Usually hassle free but YMMV.

[c4757p]

Conrad, isn't that pretty much exactly what he did build?

Edit: I can't read.

The 100k is essential for stability..

[MrAureliusR]

Okay, I've changed the output of the second op-amp as close as I can to Conrad's schematic (I had a 150k and 220R sitting at hand). I'm trying to re-majig the circuit for split supply. I have a little power supply splitter gadget I built up a while ago. It works great for powering op-amps.

My question is -- what needs to be 'ground' referenced in the circuit? Obviously the op-amps run off the V+ and V- from the splitter, but I'm not sure what to connect to the new virtual ground. Clearly the power supply under test should be connected to this ground? Or am I wrong there? And the other thing that I would think needs to be referenced to the virtual ground is the 1 ohm resistor's 'ground'. I'm trying to make sense of this all. BTW, I've turned the +5VDC into +/-2.5VDC. I could actually take this to +/-10V if that would be better.

(EDIT: I've figured part of that out. Both the resistor and DUT are 'ground' referenced. At least now I'm back to square one with it sinking 1.8A no matter what. The output of the voltage follower op-amp now goes from 8mV to 2.1V -- perfect. However the resistor divider (which is referenced to V-, a mistake I'm thinking) is now outputting -1.5 to -2ish. So this obviously should be ground referenced as well. I'm going to make that change now) (EDIT 2: Wait a minute, now I don't need the resistor divider, do I? The first op-amp is already outputting the voltages I want)

Also, can someone explain to me why the op-amp was oscillating so horribly in the first place? I understand why capacitance can solve oscillations in a lot of cases but with the FET in the loop I can't quite grasp it. Is it because of the switching speed of the FET?

(EDIT 3: This circuit is making me go mildly insane. For some reason the split supply is mucking everything up. I'm going to bed, maybe a clear head will help tomorrow. Thank you guys so much for all the help, I really, really appreciate it!)

[mrflibble]

My question is -- what needs to be 'ground' referenced in the circuit? Obviously the op-amps run off the V+ and V- from the splitter, but I'm not sure what to connect to the new virtual ground. Clearly the power supply under test should be connected to this ground? Or am I wrong there? And the other thing that I would think needs to be referenced to the virtual ground is the 1 ohm resistor's 'ground'. I'm trying to make sense of this all. BTW, I've turned the +5VDC into +/-2.5VDC. I could actually take this to +/-10V if that would be better.

Not sure what you are asking... Taking the circuit of Conrad as example, you would connect the Sink_In to the + terminal of your power supply under test, and the ground (lower side of that 1 Ohm sense resistor) to the - terminal of your psu under test. Let the other nodes float where they may.

[IanJ]

Hi guys,

I have posted this before......but hopefully it'll be useful here......from my own design notes for my own load. I still get instability but only at much higher loads/voltages. My design basically follows Dave's but with the following components. PS. Don't faint at the cost of the mosfet!

MOSFET:
These are N-type Linear Power Mosfets, the IXTK46N50L (500v, 46A). Conventional mosfets can be used but you would be limiting the range of voltages/loads the Dummy Load would be able to cover due to running the mosfet in it's linear mode. The IXTK mosfets overcome these limitations by extending the transistors’ FBSOA. I.E. ETI (Electro-Thermal-Instability) as a result of positive feedback within the mosfet when used in linear mode. The IXTK is ideally used in programmable loads, battery chargers & current regulators etc.

OP-AMP:
This is single supply capable and rail to rail in order the output can swing down to 0vdc. The LM8272 chosen here is also ideal as it is able to drive large capacitive loads such as power mosfets. This is crucial for the operation of this circuit across the full voltage range of the load.
PS. Standard op-amps may work, but again more than likely you'd be limitting the voltage range of the load.

Ian.

[dannyf]

Note that if you single supply it, you'll need an opamp where the input common mode includes ground, because the signal off that current sense resistor is next to nothing.

R2R or split supply here isn't that important because you need 2-3v to turn on the mosfet. ie. the output doesn't need to go to 0v to generate very small output current.

An opamp with a pnp input stage would do here.

The 100k is essential for stability.

The 200ohm resistor is absolutely essential for stability and is likely highly tailored for the particular mosfet used. ie. you will need to experiment with your own.

[mrflibble]

Also, can someone explain to me why the op-amp was oscillating so horribly in the first place? I understand why capacitance can solve oscillations in a lot of cases but with the FET in the loop I can't quite grasp it. Is it because of the switching speed of the FET?

The mosfet gate has a nice bit of capacitance. Essentially when you are turning the mosfet on/off you are charging/discharging this capacitor. Capacitor fully charged ==> channel fully open ==> low Rds. Mosfets with lower Rds_on tend to have a bigger channel and as such you need to pump more charge around during switching.

And this is also why you want that series resistor (the 200 Ohm one in Conrad's schematic), to prevent ringing. Basically you get an RC filter formed by the 200 Ohm and the gate capacitance. How much resistance you need here depends a bit on how you drive it, and the specific mosfet. As dannyf said, you'll have to experiment a bit. 200 doesn't sound too bad as a starting point for something driven by an opamp output. For stiffer driving sources you can go to lower value like 10 Ohm.

On the subject of multiple supplies, I'd think you could do with just one supply. Just use an opamp that can go close enough to ground. That said, when things do not work for strange reasons, it definitely cannot hurt to use an extra negative supply for the opamp during debugging. At least that way you can check if your problems are due to the supply rails or something else. But for the finished product I'd say single supply should do the trick.

[dannyf]

I understand why capacitance can solve oscillations in a lot of cases

Adding capacitance for stability generally is bad, unless you know what you are doing.

The gate stopper (that 200ohm resistor) essentially weakens the opamp's current drive and slows it down - in that regard, a slow and weak opamp is a blessing.

However, a gate stopper CAN indeed cause oscillation itself so use it with precaution. My suggestion for something like this is to use a bjt as much as you can.

If you have to use a mosfet, use it with precaution and / or buffer it with a bjt - that allows the use of a small value gate-to-source resistor.