4
\$\begingroup\$

I require a voltage source which holds a large capacitor at a stable voltage. The following secenarios must be supported:

  • Something is feeding current into the capacitor. The voltage source needs to sink current to keep the voltage constant
  • Something is draining current from the capacitor. The voltage source needs to provide current to keep the voltage constant

I only need a moderate amount of current (a supported range of +10mA to -10mA should suffice). The voltage is very low, in the range of 1.5V.

Solutions which I've considered:

LDO

schematic

simulate this circuit – Schematic created using CircuitLab

Upsides:

  • Very low quiescent current
  • Stable voltage

Downsides:

  • LDO outputs can't sink current, therefore this doesn't work

LDO with constant load

schematic

simulate this circuit

Upsides:

  • Very stable voltage
  • Works

Downsides:

  • 10mA of quiescent current. This is prohibitive since the circuit is battery-operated
  • Unpleasant surprises when the load current goes even slightly below -10mA

Zener diode voltage divider

schematic

simulate this circuit

Upsides:

  • Low complexity, cheap
  • Works

Downsides:

  • Very unstable (+/- 100mV over the current range with the shown 700 Ohm resistor)
  • High quiescent current (10mA with the shown 700 Ohm resistor)

This solution is inferior to the "LDO with constant load" variant: You can adjust the resistor, but then you either get even more quiescent current, or you get even more instability.

OpAmp in voltage follower configuration

schematic

simulate this circuit

Upsides:

  • Easily adjustable (could be controlled by uC etc.)
  • Low quiescent current
  • Very stable voltage

Downsides:

  • OpAmp outputs are not stable when connected to capacitive loads; therefore this doesn't work.

Any alternate ideas?

\$\endgroup\$
  • 2
    \$\begingroup\$ OpAmp outputs are not stable when connected to capacitive loads; therefore this doesn't work. I disagree as that depends on the opamp. Make the opamp a trans impedance amplifier (OTA, voltage to current amplifier) so that the output impedance (of the OTA) // C1 are the dominant pole and you can make this stable. Also: LDO outputs can't sink current I disagree as it depends on the LDO, I have worked on an LDO which could sink (a modest) current. You make a lot of (wrong) assumptions about the behavior of these circuits and dismiss them for the wrong reasons. \$\endgroup\$ – Bimpelrekkie Apr 23 '18 at 20:53
  • 1
    \$\begingroup\$ Stability of LDO with regard to details of load capacitance is/was always a concern. So there are no differences. If one needs to regulate voltage both ways, the last circuit is the only valid one. Stability needs to be worked out of course. \$\endgroup\$ – Ale..chenski Apr 23 '18 at 21:04
  • \$\begingroup\$ @Bimpelrekkie, the vast majority (by volume sold) of linear regulators use pass transistors and therefore can't sink current, so OP is not wrong there. Your 'an LDO' would be a rarer case and it only sank a modest current anyway. \$\endgroup\$ – TonyM Apr 23 '18 at 21:19
  • \$\begingroup\$ If you add a resistor between your opamp and the cap then the voltage follower should work without becoming unstable. \$\endgroup\$ – brhans Apr 23 '18 at 21:35
  • 1
    \$\begingroup\$ What problem are you actually trying to solve? If you keep an absolutely constant voltage on the capacitor then it never sinks or sources current itself, and it does not affect whatever other circuits are connected to it. So why not just remove the capacitor, and supply a fixed voltage to the rest of the circuit with whatever kind of regulator circuit is appropriate? \$\endgroup\$ – The Photon Apr 24 '18 at 2:33
5
\$\begingroup\$

PC motherboards use DDR busses which are transmission lines terminated by resistors to a certain voltage, which is managed by a "VTT regulator" which is able to sink or source current. The interesting bits here are the fact that motherboards are high volume, therefore many chips are available to do this job at cheap prices. Try googling the terms above, or using the search feature on digikey, looking for "termination regulator".

  • Opamp designed for capacitive loads.

Opamps usually dislike capacitive loads (due to phase margin issues) but there are ways around this. Some opamps are rebellious and will drive "unlimited capacitive loads" which are terms you should search. There are a few candidates.

  • Brute Force

If the max current the regulator will have to sink is known, a simple linear voltage regulator and a resistor sinking the max current will do the job. It is quite inefficient.

  • Roll your own

Most LDOs are voltage controlled current sources. The error voltage (output voltage minus reference) goes into a transconductance amp which creates a proportional current, and the output cap integrates this current and acts as a dominant pole for loop compensation. You can implement this yourself if you want, it isn't difficult.

If you want to create a voltage regulator which will work well with a known capacitance at the output, all you have to do is consider the output cap's ESR and capacitance, and create a circuit which synthetizes an impedance which is equivalent to some inductance and enough resistance in series to make sure the resulting circuit is nicely damped. This can be implemented with an opamp and some R/C in the feedback network, for example.

| improve this answer | |
\$\endgroup\$
1
\$\begingroup\$

Not sure where you get the idea that op amps are per se unstable with capacitive loads.

Try it.

Every power supply is an opamp with a capacitive load. Mostly they have a big fat capacitor on the output.

It would be true to say that some designs (LDO = PNP or PFET pass element) tend to be unstable with small capacitive loads. i.e. they have a minimum C to make them stable, usually a few micro farads.

| improve this answer | |
\$\endgroup\$

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.