Improve load stability

[Bebe]

Hi! I came across this problem while designing basically a linear voltage regulator using an op-amp and a logic level N-channel mosfet as a series pass element for a project i'm currently working on. I need this circuit to software control the brightness of around 20 LED-s and the LED backlight of a 16x2 display using PWM generated by a uC. The problem with this circuit is that it's not stable enough with the load. The maximum load is around 300mA, which is not that much. Otherwise the circuit works pretty well, if i set the PWM-s duty to 50% i get spot on 2.5V (open load) on the output (input is 5V) but when i add a 300mA load, the output drops to around 2.3V which is unacceptable for me because i can clearly see that the LED-s get dimmer. Also it seems that the op-amp used matters alot, the first time i was playing around with this circuit i used a generic tl072 or 082 and the output was ok-ish but then i swaped the op-amp with a MCP6022 and the output was much stable. Am i missing something obvious? How can i improve the circuit's load stability? Or is there a better way of dimming my LED-s using PWM? Please share that with me whereas it's the first time i'm designing something like this.
Thanks a lot!
NOTE: - The components in the attached schematic are not the one i will be using.

[Benta]

How do you expect the transistor to turn on when a load is attached? The gate-source voltage will go towards zero as source voltage increases due to the load.
Use a P-channel and invert your switching logic.

[Bebe]

First time when i implemented this circuit i used a P-channel but was much worse, with open load it worked just fine but as soon as i added even a 100mA load, the output went straight to zero. I don't think there is something wrong with my design other than it's stability with load since is basically a linear voltage regulator, i need to find a way to compensate for that drop or maybe i need to use a more suitable op-amp for this job...i really don't know that now hense i made this thread to maybe find out.

[ZeTeX]

First time when i implemented this circuit i used a P-channel but was much worse, with open load it worked just fine but as soon as i added even a 100mA load, the output went straight to zero. I don't think there is something wrong with my design other than it's stability with load since is basically a linear voltage regulator, i need to find a way to compensate for that drop or maybe i need to use a more suitable op-amp for this job...i really don't know that now hense i made this thread to maybe find out.

You could just use any LDO regulator and pwm the adjust pin, for lm317 with 1.25v ref the output voltage will be always +1.25V then the set voltage, so if you feed 1V to the adjust pin it will be 1+1.25=2.25V.

[Kleinstein]

The circuit is a kind of low quality voltage buffer. Due to the needed gate to source voltage there is a maximum ouput voltage of about 2-3 V, depending on the FET used and the current. At high currents more voltage is lost and one might just run into the limit. Using the N-MOSFET as a source follower is a problem. Using an P-MOSFET would be one option.

LEDs are more suitable for current control instead of voltage control. So it might be more useful to use the standard current sink circuit with the N FET at the low side. This also avoids the problem too much voltage lost.

[Benta]

The circuit is a kind of low quality voltage buffer. Due to the needed gate to source voltage there is a maximum ouput voltage of about 2-3 V, depending on the FET used and the current.

2...3 V is optimistic, even with a rail-rail output opamp. And the TL071 is far from being rail-rail.

[MagicSmoker]

I assume all of these LEDs have individual current limiting resistors and are wired in parallel, on account of the supply only being 5V...

Headroom vs. gate threshold voltage issues aside, you also need to insert a resistor in series with the MOSFET gate because op-amps hate driving capacitive loads and that is exactly what a MOSFET looks like. Something in the range of 47 to 470 ohms usually suffices (depends on the total gate charge of the MOSFET and the op-amp's output current capability).

[Benta]

MCP602x is a good choice.
Now just replace the MOSFET with a bipolar NPN and it'll work.

[David Hess]

The limited output (and input) range of the TL071 and the Vgs of the MOSFET source follower prevent the operational amplifier from controlling the MOSFET.

Using a p-channel MOSFET and reversing the feedback will solve this but frequency compensation will need to be added and a better operational amplifier will be needed because the TL071 is not very useful with only a 5 volt supply.

Better would be Benta's suggestion.  Use an NPN transistor and a rail-to-rail operational amplifier.  A PNP transistor (or p-channel MOSFET) with the feedback reverse will allow a higher output voltage but again requires better frequency compensation because of the voltage gain added by the transistor.

[Bebe]

LEDs are more suitable for current control instead of voltage control. So it might be more useful to use the standard current sink circuit with the N FET at the low side. This also avoids the problem too much voltage lost.

Thank you for your advice! Unfortunately i can't use a current sink circuit since my LED-s are individually controlled using some TPIC6C595 shift registers whose outputs are open drains. The shift registers have an output enable pin so i had the ideea to pulse that pin to see if that will dim the lights, unfortunately i can't get my hands on one of those shift registers without ordering them from somwhere (which i can't do right now). Do you think that my ideea will work, or should i stick with a voltage regulator IC or a P-MOS pass with an error amplifier circuit as you and Benta suggested?

[Benta]

Deleted

[Kleinstein]

The easy way would be to replace the N-MOSFET with an NPN transistors. This would reduce the voltage lost at the power element from something like 2-3 V to about 0.7 V.

This still leaves the problem of having a relatively small voltage / PWM range (e.g. 60-100%) that is responsible for the useful intensity. It depends on the LEDs how bad it actually is.