I have an initial accurate voltage reference, which I need to buffer. So I feed it to an opamp, which then drives a final output power transistor for a high current final voltage reference. Of course, if I want a reliable final voltage reference, keeping the noise down is imperative.

The noise from the initial voltage reference & opamp are already low. My concern now is the noise from the power transistor. Which would be typically low noise, a power MOSFET or a power BJT (Darlington)? The datasheets of the MOSFET & BJT I already bought does not mention noise levels.

Also, I didn't mention here because of website product policy that the power transistors in question are FDD770N15A & 2SD2560.

  • 1
    \$\begingroup\$ The noise of your reference will be dominated by the input noise of the op amp and any resistors you use. The noise of the power transistors is not relevant here. \$\endgroup\$
    – user69795
    Feb 26 at 19:54
  • \$\begingroup\$ Really? Is that the sole reason for no mentions in the datasheets? \$\endgroup\$ Feb 26 at 19:55
  • \$\begingroup\$ One's a Darlington with a base resistor. Does it matter much? \$\endgroup\$ Feb 26 at 19:57
  • \$\begingroup\$ Some power transistors actually have quite low noise, but they're not often used for small-signal purposes where you care about the noise level, so they don't bother to test for it. \$\endgroup\$ Feb 26 at 19:58
  • \$\begingroup\$ But, are there indications provided in datasheets that would hint at having low noise? For example, I think MOSFETs with large V_DS max rating will naturally have less channel length modulation. \$\endgroup\$ Feb 26 at 20:13

1 Answer 1


You need to distinuguish between noise within the loop bandwidth of your opamp controller, and noise outside.

Within the loop, the noise should be dominated by the opamp.

Outside the loop bandwidth, the output device will add its own noise, but here you can filter to reduce the noise. You might as well, as the output is not going to change faster than the opamp can control it.

If you really are in pursuit of the lowest noise possible, here are two tips.

Power BJTs, especially power RF transistors, are generally low noise, on account of their large junction area. A very clever engineer I worked alongside built a very low noise gain block around an LM394 super-match pair. This IC contains several thousand transistor dice connected in parallel in checkerboard fashion, so that thermal gradients are averaged out over the die. With both halves connected in parallel, it was simply a very large junction area bipolar, which with series/shunt feedback, made a 50 Ω gain block that could make -170 dBm measurements over a 10 MHz bandwidth.

I read a paper a while ago on the building of a very low noise voltage reference. The lowest noise device the author found for intermediate frequency offsets was a 50% discharged lead acid cell. It was used in circuit essentially as a many Farad capacitor in a low pass filter.

  • \$\begingroup\$ Then again, the IC is dependent on the manufacturer. Even more susceptible to parts outages now that the LM394 is obsolete. How was this engineer even sure of the internal implementation of LM394? \$\endgroup\$ Feb 27 at 20:55
  • \$\begingroup\$ Noise MOSFET & BJT aside, I've always thought BJT would be less noisy (once turned on in the linear region), but I've read in one tutorial online that MOSFET would be. All these different sources online can be conflicting. \$\endgroup\$ Feb 27 at 21:02
  • \$\begingroup\$ Just looked over the LM394’s datasheet. Looks like it wasn’t a hack & could be used that way. \$\endgroup\$ Feb 27 at 22:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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