1
\$\begingroup\$

To minimize regulator power dissipation in cases with a large input/output voltage differential, many linear regulator application notes suggest adding an input resistor to dissipate some power external to the regulator. An example from the LT3083 datasheet is shown below, where Rs is the added input resistor.

Since the resistor causes the regulator input voltage vary with load current, the output transient response of the regulator could be affected. In general (i.e. for all linear regulators, not the one pictured as an example of an input resistor on a regualtor), assuming that the resistor is sized to meet the regulator dropout requirements at maximum load, does adding an input resistor negatively impact the output transient response of a regulator? At what point, if any, does the impact become significant?

enter image description here

\$\endgroup\$
1
\$\begingroup\$

If the current exceeds the Iout*Rin > (Vin-Vout) drop on the input for a short time, a storage cap is required to compensate for this will work, as long as the energy dissipation for the input resistor to charge up the cap with power on is not too excessive above its Steady Power rating. Ec=1/2CVin^2 is the cap energy required to charge up from 0 V.

This assumes no output load on startup so Vin=Vin' worst case.

For T=RinCin if this time is < 0.1s it won't have much time to heat up.

A more precise optimum value could be done if there were data to show average and peak currents and voltage drops.

To fully understand your problem, we would need relevant PN's, layout, probing photo and scope traces.

The load current may be handled by the output cap for time durations of about 2% of V/I*Cout=T for 2% transient load regulation with rise times shorter than the LDO loop bandwidth and Tr= 0.35/f

The load regulation error is purely a function of the Rce divided by internal loop gain @ Imax using an equivalent R divider when linear but when VIO drops to zero, the loop gain drops to 1 so it just a low Rce value (at low Vce{sat}) in series with the input cap.

So a 1% load regulation means the regulator is approx 1% Zout of the Rload at Imax. If this Rce is still too high for the current spike thena bigger low ESR cap is needed. However this attenuates the LDO feedback error and if datasheet has limits on this for loop stability, this must be observed then an RLC filter may be a solution.

The same is true for transient load spikes except the ESR of the output cap becomes the source.

The datasheet shows an example of Cset=0.1uF and Cout=22uF ceramic giving an output deviation of 20mV @1Vout @ 0.1 to 1A step load or 2% transient error. Response time indicates ~10us which means the loop bandwidth f=0.35/T for T=10us = 35kHz and Req=T/C=10us/22uF = 0.5 Ohm ESR which may part;y due to cap ESR but steady state ESR from Vin'-Vout = 0.35V/3A@50'C ~ 0.12 ohms.

\$\endgroup\$
  • \$\begingroup\$ Tony, thanks for the detailed analysis for the LT3083. However, I was wondering about the impact of an input resistor in general. It seems that it is related to the loop bandwidth and gain of the regulator? \$\endgroup\$ – crj11 Mar 22 '18 at 14:45
  • \$\begingroup\$ Yes of course as I explained with Cap added to Vin' to improve this. More details on transient load are needed to verify stability and step response. It can be very nonlinear with pullup only and depends on load to pull down. \$\endgroup\$ – Sunnyskyguy EE75 Mar 22 '18 at 14:55

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.