I'm designing an audio circuit running on ±15V supplies, which has a need for precision +12V and -5V rails. These must source (from 12V) and sink (from -5V) up to ~20mA to power a CEM3340 oscillator and some ancillary components. Absolute precision of these +12V and -5V supplies is not critical (±1% would be OK) but I would like to minimise temperature dependence as far as possible, within a maximum budget of around $3. Load regulation needs to be good, as the CEM3340 current draw will vary with time across the oscillation cycle.

The circuit already uses several OPA1679 op-amps (which can source/sink over 20mA) and a 5V precision reference. So one solution seemed to be using two amplifiers from an OPA1679 in straightforward non-inverting and inverting configurations to give +12V and -5V from the existing 5V reference.

  1. Is using an op-amp to provide the power rails in this way sensible? If so,
  2. How should I make sure that the new power rails are properly decoupled? I understand that putting decoupling capacitors across the new op-amp outputs may cause instability, but don't know how to predict or mitigate this.

If op-amp outputs are not sensible here, what would be the best strategy for getting these power rails?

  • \$\begingroup\$ Do you already have a voltage rail higher than 12V and lower than -5? \$\endgroup\$ – JRE Mar 23 '20 at 23:21
  • \$\begingroup\$ Yes, I have ±15V supplies for the whole circuit \$\endgroup\$ – Chris Johnson Mar 23 '20 at 23:22
  • 1
    \$\begingroup\$ Sorry to have to ask. I've seen questions here before where someone wanted to "amplify" 5V to 12V with an opamp supplied from a single 5V rail. \$\endgroup\$ – JRE Mar 23 '20 at 23:26
  • \$\begingroup\$ Possibly your question could be improved if you give a better idea of what the load is on the 12V and -5V rails. It seems odd that you would need to sink current into the 12V rail, for example, rather than just source it. Is the 12V current and -5V current liable to flow in both directions both into and out of the power supply? Or am I misunderstanding what you mean. \$\endgroup\$ – mkeith Mar 23 '20 at 23:30
  • 1
    \$\begingroup\$ Sorry, now clarified. I just need to source current from the 12V rail and sink it to the 5V rail. \$\endgroup\$ – Chris Johnson Mar 23 '20 at 23:32

Generally, no it is not sensible to use an op-amp for a power rail. The main reason is stability. Op-amps are typically not stable when driving large capacitance loads. If you do decide to do it, you would typically add a resistor between the op amp output and the bulk capacitance for the rail. This will likely make the op-amp stable but will degrade transient response and make the output voltage dependent on the load. In some cases that might be OK. For example if you are driving circuitry with load that doesn't vary much and doesn't need a lot of bulk capacitance. You cannot use feedback to compensate for the drop across the resistor. That would defeat the stability benefit of having the resistor in the first place.

  • \$\begingroup\$ Many thanks. Can you suggest another option? (There don't seem to be many options for precision linear regulators at +12 and -5V) \$\endgroup\$ – Chris Johnson Mar 23 '20 at 23:43
  • 2
    \$\begingroup\$ @ChrisJohnson, you can use an adjustable linear regulator, with 0.1% resistors to set set the output voltage. Or use your op-amps with a transistor buffer to isolate them from any capacitive load. \$\endgroup\$ – The Photon Mar 23 '20 at 23:46
  • \$\begingroup\$ Well there are hundreds or maybe even thousands of linear regulators. Are you sure none of them will work for you? \$\endgroup\$ – mkeith Mar 23 '20 at 23:50
  • \$\begingroup\$ I'll look more carefully. I was rather discouraged by LDO dependence on temperature being rarely quoted in parametric searches (and even datasheets) and some orders of magnitude worse than that of the voltage reference + op-amp combination, when it was quoted. But knowing this is the way to go, I will look further. \$\endgroup\$ – Chris Johnson Mar 23 '20 at 23:59
  • \$\begingroup\$ @ChrisJohnson TI's LM1117 datasheet specifies guaranteed bounds on output voltage over the entire operational temperature range, though they don't have any 12V or -5V models (I'm just providing this as an example). Look at page 5, where it has typical output voltages for each nominal voltage at 25°C, but then it also has minimum and maximum values over the entire temperature range. That sort of information should be good enough for most applications, I imagine; you probably don't need to have an actual temperature coefficient. \$\endgroup\$ – Hearth Mar 24 '20 at 1:05

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.