Power supply ripple and noise rating

[brumbarchris]

Hi all,
I am evaluating some of the labs we have in the lab from a ripple and noise perspective, to select the best one for a measurement which requires low noise power supply.

From what I have seen in the datasheets of most, the common indication is the rms and/or Vpp ripple in the up to 20MHz band.

But I came across a odd one: I am looking at the specification sheet of the Keithley 2460 SMU: https://download.tek.com/document/SPEC-2460C_September_2020.pdf

In the Voltage Specification table on page 2 there is a column indicating Noise at frequencies below 10Hz. On the other hand, there is also a "voltage source noise" rating in the table on page 4, which is valid for the 10Hz...20MHz range.

I guess I will have to use this second value to compare it against the other PSUs we have, but I am somewhat puzzled about the atypical indication of the noise below 10Hz. Is this something regular for SMUs to indicate?

Regards,
Cristian

[David Hess]

Below 10 Hz flicker noise becomes dominant and because of its low frequency, it is difficult to impossible to filter out.  Some applications require low LF noise so it needs to be specified.  A standard DC to 20 MHz noise measurement does not reveal flicker noise because it will be dominated by wideband white noise at higher frequencies.

[jwet]

Flicker noise is often also called 1/f noise and as David has said is pretty much unbeatable, an inevitable physics phenomenon that you'll see everywhere.  Look at Op-Amp data sheets or app notes especially from Analog Devices.  A "1/f noise corner" is a common point where a device starts to show its 1/f noise vs. just broadband noise which tends to be flat after a few hundred Hz.  Keithley is a precision measurements company and is just being more honest frankly specifying their stuff which is universally great in my experience.

[brumbarchris]

Great, thanks for clarifying!

Cristian

[avandalen]

For a circuit with a 24bit ADC, a very low noise regulator was required, I used the LT1761ES5-5.

[David Hess]

For a circuit with a 24bit ADC, a very low noise regulator was required, I used the LT1761ES5-5.

Flicker noise is a problem even at 16 bits, and the "flicker" in the least significant digit is where it gets its  name.  Usually people do not think of it as being significant when a 40,000 or 200,000 count digital meter has more than 1 count of flicker in the last digit when connected to a clean source, but it indicates excessive noise and a design defect or damage.

For transducers which require excitation, like Wheatstone bridges, the flicker noise can be cancelled out by using the excitation as the reference for the ADC.  In practical terms this means that even a noisy regulator like a noisy 7812 can be fine as a reference if it also supplies the excitation.

Low frequency noise is dominated by the reference, and bandgap references are much worse than zener references.  Flicker noise from the error amplifier following the reference is insignificant.  It is difficult to reduce flicker noise with filtering, so a better way is to use multiple references to average it out.  With zener references levels can be low enough not to be a problem in all but the most demanding applications, and I hardly ever see multiple references to reduce noise.