Constant current load too sensitive

[iamdarkyoshi]

So I made a super basic constant current load like the one here:


except that I have the following differences:

the 7805 is a 7812 and a 16v power supply
the 50K pot is a 1k pot
the resistor divider is made of 1k resistors instead of 10k resistors
the MTP3055 is two paralleled IRF540 mosfets (I know, I can have issues with directly paralleled mosfets...)
the 1 ohm sense resistor is actually .4 ohms, for my purpose, I actually have to go even lower here, probably .1 ohms or lower...

The issue is that on my 15 turn pot and an 18650 connected, it seems like the pot has quite a large dead zone where no current is drawn, and then as I turn it up, it jumps up quite quickly in current. It does work though, but I would like for the pot to be less sensitive. I think my issue will be even worse when I go down to .1 ohm or lower, is this an accurate assumption?

[sleemanj]

I don't think an LM324 is a good choice for this application.  I made one once with a *324, and ended up having to use a negative supply and bias an input to correct for offset voltage.

My uneducated guess is that the input offset voltage of the 324, in both stages (buffer then driver) is throwing the spanner in the works for you.

The offset voltage is potentially anything up to 9mV, and the output can only get down to 5mV above the gnd/negative (it's waaay worse at the top of the range too several volts under the positive supply).

Aside: I recently (as in, today) finished one based on an LM358 (hmmm, now I think about it, it should be working the same as the *324 since the specs are about the same, maybe I'm misremembering how bad the 324 one was), which works a bit better, but in hindsight I should have given it a negative supply and the ability to compensate.  Also, I have resolved that next time I build one, I'm putting some smarts in it with a microcontroller so I can "cal" the damn thing in software, freaking trimmers.

[iamdarkyoshi]

I don't think an LM324 is a good choice for this application.  I made one once with a *324, and ended up having to use a negative supply and bias an input to correct for offset voltage.

My uneducated guess is that the input offset voltage of the 324, in both stages (buffer then driver) is throwing the spanner in the works for you.

The offset voltage is potentially anything up to 9mV, and the output can only get down to 5mV above the gnd/negative (it's waaay worse at the top of the range too several volts under the positive supply).

Aside: I recently (as in, today) finished one based on an LM358, which works a bit better, but in hindsight I should have given it a negative supply and the ability to compensate.  Also, I have resolved that next time I build one, I'm putting some smarts in it with a microcontroller so I can "cal" the damn thing in software, freaking trimmers.

Any op-amps you would suggest I use? do I need to use the buffer?

[sleemanj]

Any op-amps you would suggest I use? do I need to use the buffer?

Analog is voodoo to me, but ideally if you are using a single supply then you want rail-to-rail, or at least get REAL close to the negative rail (ground in single supply) due to the fact that you actually are interested in operating down there, getting close to the upper rail doesn't matter so much since you probably have heaps of head-room anyway.

The offset voltage as small as possible.

The bias current as small as possible, but I think most op-amps have very small bias current these days.

It may be easier to put in a negative rail, then you can both avoid the rail-to-rail requirement (just make your negative sufficiently more negative than you need) and more  easily correct for input offset.  You don't need much current on the negative rail.  Don't forget you don't have to have your rails "balanced" if your op-amp can handle at least 22v, you are free to use -10v and +12v for example.

Link about offset correction: http://www.ecircuitcenter.com/Circuits/op_voff/op_voff2.htm  - from memory I found that a negative rail for the op-amp is best here with your correction pot "straddling" the ground reference so one end is a negative voltage and the other a positive voltage so you can swing the wiper above/below reference as required to get your correction... correct.

[Jay_Diddy_B]

Hi,

Here are some pictures of a small load that I designed and built:










I have attached a pdf file with the schematic.



If you want to understand the design I suggest reading this thread:

https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/ It explains the design for a similar load, higher power.

Regards,

Jay_Diddy_B

[Kleinstein]

There is no real need for an extra buffer amplifier, the input of the second stage is hight impedance. Especially if the pot if only 1 K, it's no problem to drive a voltage divider without getting to nonlinear. Just the divider should be much higher impedance that the pot. With a good pot one could also just reduce the voltage at the pot instead of the divider after the pot - than the buffer is not critial if you need one.

The offset voltage is a problem in this circuit. So you would like an OP with lower offset voltage. Depending on the supply this could be something like LT1013 (single supply), OP07 (needs negative supply) or AD8551, MCP6V27  (AZ OP, but only 5 V maximum). You could get a negative supply by using a virtual ground (possibly just diodes on the negative side) and a separate reference for the setpoint. The cheap solution would likely be an OP07, negative supply via a white LED as shunt regulation and a TL431 as a reference.

The other point is that the OP might get unstable and oscillate with that much capacitive load. So a compensation circuit is highly recomendet. having a resistor (e.g. 10 K) from the shunts to the OP and a smal cap (e.g. 1n) from the OPs output to negative input. In addition a resistor (e.g. 100 Ohm)  at the gate is a good idea, especially with 2 FETs in parallel.

[RR]

i would suggest to put resistor in series before that 1k pot to make divider something like 4k7. that way output voltage of the pot will be 0-~2V.  and as opamps try to have voltage on both inputs same that means 0-~1V on input of second opamp and also on current sense resistors. 0-2,5A load range if i do the math right. this should solve that big dead zone. Of course all the issues mentioned before me are still valid.

[dom0]

Even for small electronic loads an op amp with at least some decent drive capability should be used. LM358/324 is just piss-poor in this regard.

Generally speaking they are okay for slow loops, but even at audio frequencies you'll see how bizarely non-linear the 358 is, even for an op amp. This of course tends to destabilize control loops.

You could try something like a 5532 for a first try, it has much better output drive than the 358. Higher Ib, offset and drift are basically similar to the 358, speed is much higher (~ten fold bandwidth). No single supply operation. There are much better op amps suited for this, but they usually have the thing in common that they aren't "jellybean".

[wblock]

The re:load looks nice.  Simple, cheap, no additional power needed: http://www.arachnidlabs.com/blog/2013/02/05/introducing-re-load/

[Kleinstein]

The llinked circuit still needs a supply for the OP. The only special thing is using a smart fet, so it has protection from overtemperature. Still not shure the smart fet really likes linear operation.

[iamdarkyoshi]

My heatsink being used here is a CPU heatsink with two extra heatsinks bolting the power resistors down.  The two mosfets are mounted directly to the CPU heatsink. I need this thing to abuse the crap out of lithium batteries, and I was aiming for 20+ amps. This is why I need to go lower on my resistors...

I do appreciate the comments though, I will be modifying this and posting updates if I do something major. I have an ebay bid to win in 1hr 20min though...

[dom0]

If it's just about battery testing and you actually don't need continuously adjustable and regulated currents I'd recommend to get some power resistors and switch them with some really nice low Rdson MOSFETs.

Two reasons
a) Linear power dissipation in power resistors tends to be cheaper per Watt than silicon
b) They usually take higher temperatures (even the plastic-aluminum (RH-xx) cased ones take 250 °C iirc, ceramic cement ones go to ~350 °C, but don't have as low thermal resistance), which means you can use a smaller heatsink / higher power density

(One can actually make a CC load with this principle that has both continuously variable and regulated output current by having a bank of power resistors individually switched and some smaller regulated current sources ; an outer (digital) control loop switches the resistor bank and provides set points for the regulated source to achieve the wanted total current –– of course this principle doesn't work well for wide input voltages, likely not your problem...)

[iamdarkyoshi]

If it's just about battery testing and you actually don't need continuously adjustable and regulated currents I'd recommend to get some power resistors and switch them with some really nice low Rdson MOSFETs.

Two reasons
a) Linear power dissipation in power resistors tends to be cheaper per Watt than silicon
b) They usually take higher temperatures (even the plastic-aluminum (RH-xx) cased ones take 250 °C iirc, ceramic cement ones go to ~350 °C, but don't have as low thermal resistance), which means you can use a smaller heatsink / higher power density

(One can actually make a CC load with this principle that has both continuously variable and regulated output current by having a bank of power resistors individually switched and some smaller regulated current sources ; an outer (digital) control loop switches the resistor bank and provides set points for the regulated source to achieve the wanted total current –– of course this principle doesn't work well for wide input voltages, likely not your problem...)

I built this not only to test batteries but to be able to test PC power supplies and stuff, I needed a dummy load for my workbench anyway. Although I will have to be adding a large resistor bank and bigass mosfets to do huge loads from said PC supplies, as the existing config is probably not going to handle 500W XD This is currently something I made with scrap parts in a couple hours. I still have four more ceramic .1 ohm power resistors and about 100 pounds of scrap CPU heatsinks... Once I get a good proof of concept, I will head on over to mouser.

[rdl]

The schematic in the first post is exactly what Dave built in EEVblog #102. You might want to make sure yours isn't oscillating. I've built it several times and it always oscillates madly.

[dom0]

Bad schematics have a habit of not working out

[iamdarkyoshi]

Well I reduced my resistor to .1 ohms, and put the wiper of the pot directly into the opamp. Now when the input voltage goes above 12V, the current draw goes insane. But from 12 down to like nothing, it seems to regulate quite well. Is there leakage somewhere that is causing the opamp to turn the mosfets completely on?

[dom0]

What  op amp are you using again? LM324/358 have phase reversal (i.e. pull all the stops) when the input(s) go outside their common mode range (something like V+ - 1.5 V or so).

Best if you draw a schematic of what exactly you now have built.

[iamdarkyoshi]

What  op amp are you using again? LM324/358 have phase reversal (i.e. pull all the stops) when the input(s) go outside their common mode range (something like V+ - 1.5 V or so).

LM324.

Edit: I will go grab a copy of DaveCAD, BRB

[Kleinstein]

Without extra cap for frequency compensation, the circuit is prone to oscillation. It could be just the higher voltage to start it.  So add the cap and resistor. You might even need an extra RC combination at the output in case you have inductive loads.

Leakage should not be a problem, the OP has plenty drive capability to counteract any leakage even at high temperature.

Input comon mode voltage should not be a big deal, as the voltage at the shunt will hardly get to high unless the fet isblown. But the LM358 might like to oscillate in some circuits due to it's dead zone of the output stage - an extra load to GND might help here.

[iamdarkyoshi]

I DEFINITELY have an outdated version of DaveCAD, I need to run updates.  But here is a horribly drawn schematic:



Anything high current is using solid core mains wiring, because why not. Anything else is on a perfboard. The resistors and mosfets are mounted directly to the heatsink.

[Jay_Diddy_B]

Hi,

You are missing two things:

1) You should add a resistor at the top connection of the Pot to limit the maximum voltage.

2) You are missing the feed back connections




I would also recommend adding an RC network to the input:



These values are not calculated, but should prevent any oscillations.

Regards,

Jay_Diddy_B

[iamdarkyoshi]

Hi,

You are missing two things:

1) You should add a resistor at the top connection of the Pot to limit the maximum voltage.

2) You are missing the feed back connections




I would also recommend adding an RC network to the input:



These values are not calculated, but should prevent any oscillations.

Regards,

Jay_Diddy_B

Ha, I knew I would forget to draw something 

I need to try this thing called "sleep" I have heard it makes you less derpy. I do have ADHD, blame that. Thanks for the suggestions though, I will try them as soon as I can and post an update.

[Jay_Diddy_B]

Quote:

I need to try this thing called "sleep" I have heard it makes you less derpy.


I have outsourced my sleeping, I pay a guy minimum wage to sleep for me 

Regards,

Jay_Diddy_B

[dmwahl]

1) You should add a resistor at the top connection of the Pot to limit the maximum voltage.
2) You are missing the feed back connections

Exactly what I was going to suggest. I just recently built a similar load and a 1nF ceramic between the negative input and output of the op-amp plus a 100R resistor between the top of the shunt and the negative input. Slightly underdamped, but settles in around 20uS.

If you're going to use 2 parallel FETs, you really should drive each one separately. They will only effectively current share when out of the linear region (which is where you're operating). In linear mode, the variations in Vgs-threshold will vary enough that you may not current share between the FETs effectively. Since the LM324 is a quad-amp, you should be able to do that without much more work.

[sarepairman2]

make sure the pot is good