Yes.... Another Dummy Load

[wigman27]

Hey again,

Yes, it is another dummy load I am having an op amp issue and im not really sure whats causing it.

I am controlling this dummy load with a micro to give my constant current, power and resistance. I am controlling it with a 12 bit DAC using a 4.096V reference to give 1mA steps up to 4 Amps, I may choose to go to 8 Amps at lower voltages, I haven't fully decided yet. With that in mind I wanted to use a 0.1 Ohm load resistor to save heat at higher currents and losses at lower voltages.

This is my first attempt


U2 was to multiply the feedback voltage to U1 by 10 so 4.096V from the DAC would give me 4.096A still. However the circuit oscillated terribly, it worked perfectly fine during the initial testing but it started to oscillate for no apparent reason. Is it something to do with the two op amps like that? should I not do that? I intend on using the LM324 in the real circuit but the LM1014 was easier to model with.

I have fixed it in my second attempt, I have removed U2 and put a voltage divider from the DAC instead and it works really well.



I am just curious what the reason for the oscillation was?

Thanks

Lee

[Rerouter]

One day someone will search the forum before posting, today is not that day...

add 100 ohm between the op amp output and the mosfet gate, and 1n between the op amp output and the inverting input, this will slow down the loop and make the op amp more happy with driving a capacitive load,

[wigman27]

Hi Re router,

I did search quite a bit before posting, obviously not enough, I'm really sorry about that. I am an admin on another forum and understand how annoying that is.

I appreciate you answering me anyway

Thanks

Lee

[toli]

Have a look here:
https://www.eevblog.com/forum/projects/yet-another-constant-current-load/

The solution I've offered starts here:
https://www.eevblog.com/forum/projects/yet-another-constant-current-load/msg146592/#msg146592
I've also explained to him what was causing it.

In the schematic you've shown, other than the pole at the output of U1 (which might in some occasions be high enough, it really depends on the OPamp and transistor parameters) there's another pole within the loop caused by U2. So you should do what was offered in the link I gave you, add a resistor to the gate of the MOSFET, add a capacitor from output to inverting input of U1, and add another resistor from inverting input of U1 to the output of U2 (or else you'll need a really large capacitor).

This will keep the DC current at the value you want (as the parts we've added won't change the circuit in DC due to the high input impedance of the MOSFET and U1 at these low frequencies), but at high frequencies where poles start piling through the loop, we've offered another path using the new capacitor, which will keep U1 stable. The resistors will help us isolate this alternate path from the rest of the loop.

[wigman27]

In the schematic you've shown, other than the pole at the output of U1 (which might in some occasions be high enough, it really depends on the OPamp and transistor parameters) there's another pole within the loop caused by U2. So you should do what was offered in the link I gave you, add a resistor to the gate of the MOSFET, add a capacitor from output to inverting input of U1, and add another resistor from inverting input of U1 to the output of U2 (or else you'll need a really large capacitor).

This will keep the DC current at the value you want (as the parts we've added won't change the circuit in DC due to the high input impedance of the MOSFET and U1 at these low frequencies), but at high frequencies where poles start piling through the loop, we've offered another path using the new capacitor, which will keep U1 stable. The resistors will help us isolate this alternate path from the rest of the loop.

Hi Toli :-)

That worked perfectly! I ended up using 100R resistors and a 100n cap and it seems very stable.

I think I will go for the second option using a voltage divider. For two main reasons, one, its one less op amp, and two, the 1st option won't let me get down to the lower currents without using a split supply, as the lt1014 will only go down to 7mV output swing and then that's amplified to 70mV at the inverting input of U1.

It obviously doesn't matter that the output of U1 can't get to 0v as the threshold voltage of the MOSFET is 1.7v

Can you see any other issue with using the second option? I will add the 100R and try a 10n to start with.

Thanks again for your help!

[toli]

The second option seems identical to what was discussed in the link I gave you. Its really a simple circuit. Other than doing what I've explained for stability, I don't think there's much to go wrong with this very simple circuit.

BTW, did you add another resistor from the inverting input of U1 to the output of U2 in the first version/the current sense resistor in the second version?
It will probably be a must for the second version. In the first version the output impedance at high frequencies can be 10's-100's of ohms (I'm too lazy to open the datasheet of the opamp to check ) so you might get away without it, but in the second version the output impedance is 0.1R so unless you add this resistor the 100n cap is going to be too little and it might also load the output of U1 too much.

[wigman27]

BTW, did you add another resistor from the inverting input of U1 to the output of U2 in the first version/the current sense resistor in the second version?

Yep I sure did :-) everything worked well.

It will probably be a must for the second version. In the first version the output impedance at high frequencies can be 10's-100's of ohms (I'm too lazy to open the datasheet of the opamp to check ) so you might get away without it, but in the second version the output impedance is 0.1R so unless you add this resistor the 100n cap is going to be too little and it might also load the output of U1 too much.

I will have a good play and post the final schematic if that's ok :-)

Thanks again for your help!

[toli]

My pleasure 

BTW, I'd add a capacitor in parallel with the voltage divider just to lower the noise U1 will see at its input. It'll slow the response slightly, but its a DC load so who cares?! 

[wigman27]

BTW, I'd add a capacitor in parallel with the voltage divider just to lower the noise U1 will see at its input. It'll slow the response slightly, but its a DC load so who cares?! 

Will do, would 100n be too much capacitance for that?

Also I'm thinking of dropping the sense resistor to 0.01 ohms instead of .1 and change the divider to 100. Would that cause issues being that lower sense voltages do you think? Being in the noise?

At the moment it works quite well down to 2mA so that would be down to 20uV on the op amp inputs. Not sure if the lm358 will do that well. What's your thoughts?

Lee

[toli]

100n should work, there's no such thing as too much in this location as its a DC load.

Regarding the voltage drop, I'd advice against it. if you have 4.096V/10=0.4096V at the moment, there's very little gain in minimum operating voltage by dropping it lower. I'd leave it at the 0.1R value.

[rdl]

I actually built one of these "dummy loads" last weekend. It doesn't seem to work right.

I followed the exact schematic in the EEVblog episode 102 with only two additions, the 1 nf cap across the output and inverting input of the second op amp, and a 100 ohm resistor between the output and the FET.

I hooked it up to a power supply and I could adjust the current loading but every time I changed the voltage on the power supply the current would change also.

I hooked my scope up to the output of the second op amp and it was a sort of sawtooth waveform, about 1 to 2 volts and around 200kHz to 300kHZ. This should be a DC voltage pretty much right? Relatively flat without a lot of noise?

I think I'm going to replace the resistor with a pot and maybe add a pot between the FET source and op amp  inverting input. Then hook up a cap substitution box and see what it really takes to make it work correctly.

[TerminalJack505]

I actually built one of these "dummy loads" last weekend. It doesn't seem to work right.

I followed the exact schematic in the EEVblog episode 102 with only two additions, the 1 nf cap across the output and inverting input of the second op amp, and a 100 ohm resistor between the output and the FET.

I hooked it up to a power supply and I could adjust the current loading but every time I changed the voltage on the power supply the current would change also.

I hooked my scope up to the output of the second op amp and it was a sort of sawtooth waveform, about 1 to 2 volts and around 200kHz to 300kHZ. This should be a DC voltage pretty much right? Relatively flat without a lot of noise?

I think I'm going to replace the resistor with a pot and maybe add a pot between the FET source and op amp  inverting input. Then hook up a cap substitution box and see what it really takes to make it work correctly.

There are quite a few threads regarding Dave's dummy load on the forum.  I think if you follow Toli's advice in this thread you can get it working.

Here are a couple post that have schematics of circuits that are known to work:

Jay_Diddy_B's
Terminal Jack's

[rdl]

I did search, dummy load returned about 3 pages of results. Around a third were actually about Dave's circuit. I should have narrowed the search a little more I guess.

As I said, I built Dave's original circuit exactly except for the cap and resistor, both of which were recommended additions in many of the posts I read. However, there is a lot of variation in the values recommended for those parts, as well as in the actual op amps and FETS that people use. I used the same MTP3055VL as Dave and an LM358 (dual version of the LM324). Didn't work. Don't worry, I will figure it out eventually. Maybe it's my circuit layout, I built it on perfboard.

[TerminalJack505]

It sounds like you are missing a resistor between the 1R current sense resistor and the op amp's inverting input.  This was key for stability when I built the circuit.

[rdl]

No, I didn't have that resistor. I just now cut that wire and clipped in a 2K pot, but it doesn't seem to make a lot of difference. The magnitude of the waveform and the frequency change somewhat as I rotate the pot, but I'm still not getting anything close to DC out from the op amp.

I'm thinking the 1 nf cap across the op amp's output and inverting input is too small (I see you have used 10 nf). Tonight or maybe in the morning I'll solder on some wires so I can connect my cap substitution box there. Should I replace the 100R on the FET gate with a some kind of pot also?

I attached my pcb layout. It's built on pad per hole perf board. All wiring is on the bottom, bare 24 gauge solid wire in straight lines point to point (except the thicker lines are 18 gauge solid).

I will get this to work, I've spent too much time on it to give up now, besides it might actually be useful

Thanks for your advice.

[TerminalJack505]

Yeah, there will be some magic combination that makes it work so stick with it. 

If you leave the size of the resistor between the output and the gate at 100R then the resistor between the current sense resistor and the non-inverting input will likely have to be in the range of 10k.  Your circuit is similar to Jay_Diddy_B's and that's what he used.  (He has a 10k from current sense to non-inverting input, a 220R on the MOSFET gate and a 470p compensation cap.)

[wigman27]

Hey again 

Sorry for the delay.. Work got in the way of electronics!! How RUDE!

Ok, So I have come up with a circuit that was very stable across all ranges, all currents from 2mA to 8A all voltages from 1.5V to 24V using the LT1014 which I thought was reasonably close "replacement" to the LM324 which I planning on using in my final circuit. It worked exactly as you would expect.

I had reached a stage I was happy with so I thought I would find an LM324 model to try, just to be sure, when I replaced the LT1014 with the LM324 the circuit went crazy! It didn't oscillate but the MOSFET was very slow to respond at low currents and huge overshoots at high current settings then it would finally settle. I wasn't able to get the LM324 anywhere near the same low current outputs that I was getting the LT1014 down to. I'm guessing it could be a few things, does it sound like any of these?

• Dodgey LM324 subckt
• The LM324 is not as precise as the LT1014 and not able to sense the lower voltages at the 0.1ohm resistor R1
• It has something to do with the phase shift I have done for the LT1014 isn't satisfactory (or just a different requirement) to the values I should be using for the LM324?

I would really appreciate some help as to what may be going on. I have attached a zip of the folder as there are two library's and also I haven't worked out how to save the .asc file to text nicely as yet.

Thanks for your help!

Lee

[Rerouter]

lets rule out the basics,

common mode input range includes ground, check
output can get close to ground, check

sink and source current are not the same, (may not be pulling down fast enough at current time constant)
figure 13 is interesting, http://www.ti.com/lit/ds/symlink/lm124-n.pdf

equally is common mode rejection is about 40db worse than the LT1014,

[Kevin.D]

Yes the lm324 model your using is very broken I would say . I tried your circuit with using two  lm324 spice models I have  LM324/NS and LM324/ST and both work ok in your circuit .
Having said that you are still not using very good values in your circuit to achieve a good fast but stable control loop when using lm324 and it's clones.. I would put a bigger resistor on the gate ~ around 470 ohms . And decrease c2 to around ~ 22n and decrease R5  to  ~1K . This should give you good stability for most load conditions without slowing the feedback loop to much .

[wigman27]

That's good!! Thanks very much guys.

Yes the lm324 model your using is very broken I would say . I tried your circuit with using two  lm324 spice models I have  LM324/NS and LM324/ST and both work ok in your circuit .
Having said that you are still not using very good values in your circuit to achieve a good fast but stable control loop when using lm324 and it's clones.. I would put a bigger resistor on the gate ~ around 470 ohms . And decrease c2 to around ~ 22n and decrease R5  to  ~1K . This should give you good stability for most load conditions without slowing the feedback loop to much .

I will try these new resistor and cap values out. Can I ask, how did you come up with these values? Is it experience or is there a good way of working them out? I did do a bit of reading initially on phase shift angles but still havent found a way to work it out thy doesnt require a doctorate in all things electrical... I'm just asking so I will know for next time :-)

Would you mind sending me a link to the models you have?

Thanks again :-)

Lee

[TerminalJack505]

So far as I can tell you have two problems. 

One, your bandwidth is about 600Hz.  I changed things to what's shown in the attachment a got a lot more bandwidth. 

The other problem is voltage offset.  By using a 0.1R resistor you're getting a gain of 10x as well as seeing 10x worse offset.  Which could account for up to 70mA of error on the output with the LM324's worse-case offset of 7mV.

[rdl]

I built the exact circuit as shown in EEVblog #102 and it oscillates like crazy. I built it again adding the resistor between the op amp and the FET and the cap across the op amp output and inverting input (C2 and R4 in your circuit) and it still oscillated (see my previous posts in this thread). If I didn't have an oscilloscope I probably never would have been able to fix this circuit.

I replaced C2 with an IET Labs capacitor substituter and replaced R4 with a 2K pot. Then I added a 5K pot between the FET source and the op amp's inverting input (R5 on your schematic). 5K turned out to be way too large, I had to change it to 500 ohms.

Based on my tests, using LM358 and MTP3055VL as in the original circuit:

R4 is required to prevent bad oscillation when the load voltage is outside 3 to 8 volts range.
I was able to stabilize the circuit with R4 only, but it required a very large value (over 47K).
Around 100 - 150 ohms worked fine here when used with C2 and R5.

C2 needed to be 110n minimum, going higher didn't seem to hurt, so I settled on 330n just to be safe.

R5 was where it got strange. The required value was between about 5 ohms and 50 ohms only.

Anything outside this range an the circuit would oscillate badly, without the oscilloscope, I probably would never have figured this out.

[Jay_Diddy_B]

Hi group,

I recently posted a design for a dynamic load. This thread can be found here:


https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/msg288313/#msg288313

The design shows how to use more than one MOSFET to share the dissipation.

It also shows how to make a pulsed load that can be used for doing transient load test on power supplies.

Regards,

Jay_Diddy_B

[Kevin.D]

The two ltspice models i use you can find here .
http://ltwiki.org/?title=Components_Library_and_Circuits
just click on the 7.-opamps . and you will find lm324/ns model there .

on this same page both the lm324/ST and lm324/ns are in the "A Large LTspice Folder from Bordodynov"  in the "extra" folders of that download . These are what I used .

 On how did I came up with these values , It was from both a bit of experience and from looking at the bode plots of these circuits . If you want to do the bode plot of your circuit try the "loopgain2" method you will find in the examples directory of lt spice .Bode plots  are  fairly easy to do (in spice anyway ) but youl need to do  a bit of reading and practice to interpret them and use them (but basically it's phase shift at the 0db gain crossover point that's the important bit ) .
What your trying to do here is get a  stable but as fast control loop as you can ,so this means adding as little compensation as you can get away and still be stable .(it's very easy to stabilize any feedback control loop just by keep adding more compensation and reducing it's bandwith to lower values but then  you have reduced it's abilility to react fast to any changes which is the whole point of the feedback loop in the  first place,.).
An easy way for most hobyiest and beginers to check  control loop stabilty  with out having to do Bode plots(and it works for 99% of the time)  is to just do a step load response and  you can get an good indication of the phase margin ( stabilty) that way . the nice thing about this is that you dont need expensive equipment and to can do this in both Ltspice simulation and on the real  circuit , you can then compare the results . (to do this in ltspice on your circuit just step the load voltage souce (V2) rapidly a few volts  in your simulation and then plot the response across the current sense resistor (R1) ).you can then compare to chart like I have posted below to get the phase margin, the amount of overshoot/ringing indicates phase margin ,you  aim for about ~ 60 deg margin for a optimized loop .(but you should also test under the worst working case senariou your likely to see to check is still stable (not oscilating but will have much reduced phase margin).  :-that is with a resistive load  in the Drain of Fet for Electronic load type circuits like yours ).

Or you could of course avoid all the above and just copy what everyone else does  .It's up to you 
 Anyway It's sunday ,so my day of rest so enough typing for me today .
Good luck .

[wigman27]

So far as I can tell you have two problems. 

The other problem is voltage offset.  By using a 0.1R resistor you're getting a gain of 10x as well as seeing 10x worse offset.  Which could account for up to 70mA of error on the output with the LM324's worse-case offset of 7mV.

Hi TerminalJack505,

Thanks for your reply

My original circuit did have an offset problem as it was using a x10 gain op amp picking up the voltage drop across the 0.1ohm resistor. However with this circuit I don't have any gain in the op amp and offset isn't really a problem a the threshold voltage of the MOSFET is around 1.7v so 7mV offset hopefully won't worry me too much.

I built the exact circuit as shown in EEVblog #102 and it oscillates like crazy. I built it again adding the resistor between the op amp and the FET and the cap across the op amp output and inverting input (C2 and R4 in your circuit) and it still oscillated (see my previous posts in this thread). If I didn't have an oscilloscope I probably never would have been able to fix this circuit.

I replaced C2 with an IET Labs capacitor substituter and replaced R4 with a 2K pot. Then I added a 5K pot between the FET source and the op amp's inverting input (R5 on your schematic). 5K turned out to be way too large, I had to change it to 500 ohms.

Based on my tests, using LM358 and MTP3055VL as in the original circuit:

R4 is required to prevent bad oscillation when the load voltage is outside 3 to 8 volts range.
I was able to stabilize the circuit with R4 only, but it required a very large value (over 47K).
Around 100 - 150 ohms worked fine here when used with C2 and R5.

C2 needed to be 110n minimum, going higher didn't seem to hurt, so I settled on 330n just to be safe.

R5 was where it got strange. The required value was between about 5 ohms and 50 ohms only.

Anything outside this range an the circuit would oscillate badly, without the oscilloscope, I probably would never have figured this out.

Thanks mate! I will definitely check this out, I will have to do some research on the right values to use, I will start with the values Kevin.D and you have suggested as a start.

Hi group,

I recently posted a design for a dynamic load. This thread can be found here:


https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/msg288313/#msg288313

The design shows how to use more than one MOSFET to share the dissipation.

It also shows how to make a pulsed load that can be used for doing transient load test on power supplies.

Regards,

Jay_Diddy_B

I saw that! it looks very nice!! thank you.

Or you could of course avoid all the above and just copy what everyone else does  .It's up to you 
 Anyway It's sunday ,so my day of rest so enough typing for me today .
Good luck .

Thats way too easy! sometime I just wish I could do that... but no, I have to know EXACTLY how it works or its not good enough for me..

Thanks for taking time on Sunday to answer these questions.

I will let you all know how I go.

Lee