Author Topic: Increasing accuracy of linear voltage regulator  (Read 1210 times)

0 Members and 1 Guest are viewing this topic.

Offline mbless

  • Regular Contributor
  • *
  • Posts: 221
  • Country: 00
Increasing accuracy of linear voltage regulator
« on: April 25, 2017, 02:20:28 pm »
I'm looking to make an accurate, adjustable current source (<1% up to 750mA) using a known voltage and accurate resistors. For currents less than 15mA I can use a reference voltage such as the MCP1501T which is nominally 0.1%, so no problem there. The issue is above 15mA I can't find an accurate voltage regulator; the best is about 1% accuracy, such as the AZ1117C, but that is really 2% over the full temperature range due to the internal reference voltage range.

So I was wondering if anybody has improved the accuracy of LDOs by using the an accurate voltage reference and op amp at the adjust pin of an LDO. I've attached a schematic roughly showing what I had in mind. (The 0.1% accuracy I'm showing is ignoring temperature and load effects.) I haven't worked out the specifics since some LDOs have source current out of the adjust pin.

So has anybody attempted this?


Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 7906
  • Country: de
Re: Increasing accuracy of linear voltage regulator
« Reply #1 on: April 25, 2017, 03:13:21 pm »
The circuit is a little like an an buffer with the LM317 to boost the OPs output current. So its an constant voltage source. Usually one would use a different OP and would likely need to add some extra compensation to make it stable with a capacitance at the output.  The LM317 is used a little like a transistor - so one could as well use a Transistor if one does not need the over-temperature / over-current protection.

The max4239 is wrong in three ways: it is not unity gain stable and as a AZ OP it has quite some current noise spikes and is not really low noise. Without extra compensation it is also much to fast compared to the LM317. An OP that might work (to a certain capacitance) without extra compensation would be something like an OP193 - so a really slow OP.

For a constant current one would use something like the constant current sink circuit with a power MOSFET, precision resistor on the source side and OP in feedback. Power is coming from a raw voltage source of suitable value.

Offline mbless

  • Regular Contributor
  • *
  • Posts: 221
  • Country: 00
Re: Increasing accuracy of linear voltage regulator
« Reply #2 on: April 25, 2017, 03:36:19 pm »
  :palm: D'oh, I should have thought of using the dummy load setup! There's plenty of designs floating around here.

So then I can use the voltage reference as the non-inverting opamp input and different resistor arrays at the mosfet source to set my current.

Offline MrAl

  • Frequent Contributor
  • **
  • Posts: 481
Re: Increasing accuracy of linear voltage regulator
« Reply #3 on: April 27, 2017, 06:04:09 pm »

Actually yes i did just that, but with a voltage regulator not a current regulator because i needed a steady voltage for something and it had to be very stable with temperature and input line changes.

I started with an LM317 but then upgraded to an 8 amp version of that chip.  The upgrade just gave it more current ability though and did not change the performance otherwise.

The thing i found out about the LM317 and similar is that the voltage reference for those chips is in the same package as the pass transistor that does the main regulation of the output, so when the internal pass transistor gets hot, the voltage reference gets hot.  This changes the temperature profile by a lot more than the data sheet would have to believe because the internal reference temperature keeps rising as the device heats up.

To get around that i figured that the simple solution would be to have the reference off chip so it would not heat up as the internal transistor heated up.  It's a simple concept.  The circuit is similar to yours, where you use an op amp with a good voltage reference and use that to drive the LM317.  When the LM317 heats up, the reference stays at room temperature so there is very little change from that.  Of course if you have a wide temperature range for your ambient then you might want to put your external reference in a small oven with regulated temperature, or you could buy a very high quality reference which you can get these days for around 5 dollars (USD).

It works pretty well, and i later added a millivolt adjustment to it so i could set the output voltage by millivolts to get a precise setting which i needed at the time.

I also added a diode on the output of the op amp so that it effectively OR's the output of the op amp with the normal resistor that goes to ground in regulators like the LM317.  I can set the resistor to keep the voltage at maybe 11v and set the op amp to regulate to 10v, so the output can not get above 11v if something goes wrong.  So it's a secondary safety mechanism for over voltage with normal output regulated to precisely 10v unless something goes wrong with the feedback and then it should only get up to 11v and that could trigger an error LED for example.

I used an LM358 for the op amp to start with but intended to upgrade to a faster op amp in the future, like with slew rate at least 10 times that one.  The LM358 isnt really suitable for power supplies that have to keep close output regulation over time because of their poor slew rate, but it works for less critical stuff like battery charging.
« Last Edit: April 27, 2017, 06:10:16 pm by MrAl »

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo