Normally, vendors of LDOs and DC/DC converters specify the output voltage error in the range of 1.5% to 2.5%. However, the analysis often suggests that the error can range from 3% to 6%. Lately, I saw a demonstration on youtube, showing the significant difference between the actual min/max and the spec'd error of TI TPS784. The error is mainly due to the variations of Vref, Ib, and resistors' values. Why is the datasheet's value so low compared to the calculation results CircuitEXP provides? Do they emphasize different aspects?

  • \$\begingroup\$ The TPS784 datasheet has a note on the output accuracy spec saying "Feedback resistors tolerance is not included in overall accuracy" Which on the one hand, they can't control what resistors use. On the other hand, calling it "overall accuracy" while neglecting a major error contributor is kind of silly. \$\endgroup\$
    – The Photon
    Oct 6, 2020 at 1:48
  • \$\begingroup\$ Things like step response error, ripple and thermal regulation error are much more difficult to maintain train within 2% on SMPS than linear regulators with a 1st order control loop. The reasons are too many to list. \$\endgroup\$ Oct 6, 2020 at 2:08
  • \$\begingroup\$ Good points. What you pointed out is for the 0.5% error on Page 6. This number is even lower than 1.5% error specified for Vfb(Vref) along, which is on the same page. If you look at Page 24, an error of 1.5% is given for a real example. Using CircuitEXP with exclusion of the resistance error, the part still has 3.9% error. Considering the resistance tolerance, the total error is more than 5%. \$\endgroup\$
    – X J
    Oct 6, 2020 at 2:09
  • \$\begingroup\$ The main sources of steady-state inaccuracy are the reference tolerance and the divider resistor tolerance. Finite gain can contribute a very small inaccuracy but it's usually negligible in a properly compensated converter. You can see the overall tolerance due to Vref and divider here: syncbuck.com \$\endgroup\$
    – John D
    Oct 6, 2020 at 3:28


Your Answer

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