DigiPot 1/f noise

Is there more noise data for 1KΩ DigiPots than the 1KHz spot noise spec in the data sheets? For example, the AD5253 (1K) shows 3 nV/√Hz at 1KHz which makes sense for a pure resistance in the neighborhood of 1KΩ. I am interested in the region of 0.1Hz to 100Hz. Since the devices are not simply resistors but semiconductors, is there 1/f noise? If so, has it been characterized? Any data?

BTW, with such low noise, these devices might be attractive as programmable attenuators for low level signal processing. The noise in multiplying DACs is rather high. IMHO.

Parents Reply Children
  • Hi, the original questioner wanted to know about frequencies from 0.1 Hz on. Is there any data available on noise voltage density in that range (0.1 to 10 Hz)?

    Regards,

    Gerd

  • 0
    •  Analog Employees 
    on Mar 17, 2020 5:23 AM in reply to GerdF

    We do not have any data available on the voltage noise density from 0.1-10Hz.

    As stated above, there will be noise contributed by the CMOS switches at the wiper position, but that will fall off at 1Hz.

    Do note that the thermal noise of sqrt(4kTRB) is the dominant factor in the resistor noise.

  • If you do not have any data, how can you be sure that it will fall off at 1Hz?

    Also resistors always have intrinsic noise. It's usually current noise, but it's not like current noise flowing "out of" the resistor as it does with the input terminals of OpAmps. It rather is resistance noise where the resistance value changes with 1/f characteristic. As soon as you put some voltage across the resistor, there will be according current noise.

    While thin film resistors (I assume the on-chip resistors in your DigiPots are thin film ones) seem to be superior to thick film resistors, they still have some 1/f noise.

    With your capabilities (you have very low noise voltage regulators and also very low noise OpAmps) it should not be a big problem to measure the low frequency noise behavior of your DigiPots.

    An LT3045 at +15V, an LT3094 at -15V, both with some 1000µF across Rset, an AD5293 in between the two outputs, set to midscale (would not even have to be programmed) and an ADA4625 as amplifier (or some of them in parallel) should to the job.

    Regards,

    Gerd

  • 0
    •  Analog Employees 
    on May 8, 2020 5:26 AM in reply to GerdF

    Let me know what your actual requirements and your application. 

    As mentioned above, TFR does not have any significant 1/f noise compared to the resistor thermal noise.