1
\$\begingroup\$

I'm trying to use a 3-terminal linear regulator LR8 (Also tried TL783 and saw the same problem...) to regulate 180V from a fixed PSU down to 116V, to power some very light loads < 1mA.

But when I power this thing on, the output voltage is all over the place!

Currently, I have the rheostat adjusted to 0, so the resistance from Adj->GND is 16K. This means the nominal output voltage should be ~97V. I put a 24K load at Vout... depending on output voltage should set the output current to single-digit mA range. (EDIT: And similar current range through the set resistor R6).

It looks exponential, but the output cap is directly connected so obviously the time constant isn't minutes.

At 120V, the load dissipates ~0.6W, the set resistor dissipates 0.9W, and the LR8 dissipates 0.75W. The regulator is definitely hot to the touch, so I was thinking temperature effects? But with thermal resistance of 81C/W, this only brings the temperature up to maybe 100-105C, within the operating range. I'm not sure how to measure the actual temp.

Also, if you have ideas for a better regulation alternative to this circuit, please leave a comment! Thank you.

VOUT drift:

enter image description here

The circuit is stupid simple:

enter image description here enter image description here

\$\endgroup\$
  • \$\begingroup\$ Keep in mind that R6 will dissipate more power than your 24k load. \$\endgroup\$ – WhatRoughBeast Apr 12 '16 at 16:08
  • \$\begingroup\$ "the output voltage is all over the place" doesn't really say how much drift you are seeing. Please give actual numbers. Otherwise it is impossible to assess whether what you are seeing is within the expected tolerances. \$\endgroup\$ – user4574 Apr 12 '16 at 16:14
  • \$\begingroup\$ @WhatRoughBeast Ah yes \$\endgroup\$ – Keegan Jay Apr 12 '16 at 16:21
  • \$\begingroup\$ @user96037 I added a title to the plot in case it wasn't clear \$\endgroup\$ – Keegan Jay Apr 12 '16 at 16:22
  • 1
    \$\begingroup\$ The device might be over-heating. Double the resistance of R4 and R6 to reduce current to LR8 and see what happens. To use a lower voltage regulator, you need to make sure the operating voltages are within limits even during power up or down. A zener diode from input to output could be a first step for that. \$\endgroup\$ – rioraxe Apr 12 '16 at 22:17
2
\$\begingroup\$

Looks like a possible temperature problem? The internal reference moves at temperature; since you have a gain of 80, you will see a volt on the output for every 12mV of drift on the reference, which usually will drift tens of mV. Check by cooling the part and watching the output.

\$\endgroup\$
  • \$\begingroup\$ The problem, as it turns out, was the set resistor R6 overheating (not the regulator). Presumably, the large power dissipation was enough to change the component's resistance such that the output voltage was set to increase. This resulted in slow positive feedback, where the output voltage just continually rises. I'm selecting your answer for the methodology you suggested. We solved it by using two higher power-rated resistors in parallel. Thermal stability was achieved even faster by bonding a small chunk of metal the the resistor surfaces. \$\endgroup\$ – Keegan Jay Apr 15 '16 at 22:30
  • \$\begingroup\$ @JayKeegan your resistors look suspiciously like the ancient carbon composition type, known for poor temperature stability. Could it be? \$\endgroup\$ – Oleksandr R. Apr 27 '16 at 1:42
  • \$\begingroup\$ @OleksandrR. Yes, they are ancient! I had no idea that could be an issue. They are from an old electronics lab stock. "Ohmite Devils or something like that" \$\endgroup\$ – Keegan Jay Apr 27 '16 at 19:42
  • \$\begingroup\$ @JayKeegan if you have a lot of them, you can sell them to people building guitar amplifiers and similar things. People pay (very) good money for these crappy resistors, especially new old stock such as these appear to be! You can buy some decent, tight tolerance and temperature-stable metal film resistors with the profit. (Carbon composition resistor values can drift by tens of percent over the years; you probably shouldn't trust their values now. The only advantage they have over most other types is their surge tolerance.) \$\endgroup\$ – Oleksandr R. Apr 27 '16 at 20:31
  • \$\begingroup\$ See here on eBay. People will pay $6+ per resistor! \$\endgroup\$ – Oleksandr R. Apr 27 '16 at 20:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.