Data sheets often give uncertainty in dB. If uncertainty is specified as a percent, I believe that means it is a relative uncertainty. Assuming the coverage factor is one, the relative standard uncertainty is the percentage given. When uncertainty is specified in dB and is converted back to a percentage, the interval is no longer symmetric. For example, +/-3 dB would be +100%/-50%. How should I interpret this?


In The Measurement of Power Spectra by Tukey and Blackman, the interval is given as the spread in dB. "Spread is the difference between the upper boundary expressed in db, and the lower boundary expressed in db." And "All intervals are symmetric in the probability sense, half of the non-included probability falling above and half below the interval."

If we think of the interval in terms of total spread, does it make sense to add the positive and negative uncertainties? For example, +/-3 dB, would be a spread of 6 dB or +/-150%.

  • 1
    \$\begingroup\$ if the uncertainty covers from 0.5 to 2 with a mean of 1, then the mean value is no longer centred in the linear sense by a significant amount. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Aug 8 '18 at 3:36
  • \$\begingroup\$ Fractional values do not work well with dB, which is normally expressed as a whole integer. A sound engineer is not concerned with fractional dB levels, but a ADC designer would be. \$\endgroup\$ – user105652 Aug 8 '18 at 4:21

1dB voltage is 12% change, as is 1dB current

1dB power is 25% change.

Linearizing these, 0.1db voltage is about 1% change.

0.1dB power is 2% change.

  • \$\begingroup\$ When the numbers are small, the percentages about the mean are roughly equal. Are you suggesting it only makes sense to specify uncertainty in dB for small percentages? \$\endgroup\$ – Todd Aug 8 '18 at 13:16
  • \$\begingroup\$ it depends why the uncertainty is non-linear \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Aug 8 '18 at 17:19
  • \$\begingroup\$ @TonyEErocketscientist In general, I'm struggling to understand what that +/- db really means. But I'm specifically interested power variance. See edit. \$\endgroup\$ – Todd Aug 9 '18 at 14:52
  • \$\begingroup\$ It is a log scale , so on big log numbers with big tolerances it is like saying plus or minus an order of magnitude when a percentage does not make sense. Where the uncertainty is some fraction of the exponent for uncertainty. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Aug 9 '18 at 15:04

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.