Author Topic: Slew Rate (dI/dt) for SMPS Transient Response Testing  (Read 2201 times)

0 Members and 1 Guest are viewing this topic.

Offline TimNJTopic starter

  • Super Contributor
  • ***
  • Posts: 1720
  • Country: us
Slew Rate (dI/dt) for SMPS Transient Response Testing
« on: May 29, 2020, 02:35:13 pm »
Hi everyone,

For low-to-medium power switching power supplies (~5-500W), does anyone have an opinion on a reasonable slew rate (dI/dt) for worst-case transient response testing (AKA load slamming)? I've seen people recommend anywhere from 1A/us to 50A/us to 1000A/us. I assume that a higher slew-rate will be more likely to expose a marginal feedback loop, due to it being a more strenuous test, but I'm not sure what is reasonable for this type of power supply. I guess a point-of-load converter right next to a CPU needs to handle much higher instantaneous loading. I'm mainly working on power-brick an open-frame AC-DC offline designs.

Any ideas?

Thanks!
« Last Edit: May 29, 2020, 02:36:50 pm by TimNJ »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Slew Rate (dI/dt) for SMPS Transient Response Testing
« Reply #1 on: May 29, 2020, 03:04:07 pm »
Faster load switching is more of a test of the output capacitance and decoupling than the regulator's control loop and this especially applies to point of load regulators.
 
The following users thanked this post: TimNJ

Offline TimNJTopic starter

  • Super Contributor
  • ***
  • Posts: 1720
  • Country: us
Re: Slew Rate (dI/dt) for SMPS Transient Response Testing
« Reply #2 on: May 29, 2020, 03:34:29 pm »
Faster load switching is more of a test of the output capacitance and decoupling than the regulator's control loop and this especially applies to point of load regulators.

Ah yes, definitely a good point. I'm thinking there's a cross-over point where further increase in slew rate does not do much to test the control loop. Although, the ESR of the output capacitance and ESL/ESR of the load connection is still a valid thing to consider. Considering a power supply might have a loop bandwidth of ~10KHz, that puts the transient reaction time in the order of ~100us to even "recognize" a perturbation on the output. Assuming ~10A full power rating, then even with a slew rate of 1A/us, the load will fully transition in 10us, which is still 10x faster than the control loop's response time.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Slew Rate (dI/dt) for SMPS Transient Response Testing
« Reply #3 on: May 29, 2020, 07:38:56 pm »
Faster load switching is more of a test of the output capacitance and decoupling than the regulator's control loop and this especially applies to point of load regulators.

Ah yes, definitely a good point. I'm thinking there's a cross-over point where further increase in slew rate does not do much to test the control loop.

It is not difficult to switch a load fast enough to control the slower control loop.  Anything faster just tests the output capacitors.

Quote
Considering a power supply might have a loop bandwidth of ~10KHz, that puts the transient reaction time in the order of ~100us to even "recognize" a perturbation on the output.

Linear regulators optimized for high speed, which means they can use lower output capacitance, can respond in several microseconds.  Switching regulators have a more difficult time because of the added lag of their output inductor and may use hysteretic control which allows for relaxed frequency compensation for improved performance.

Something else to consider is the impedance of the output capacitor and transmission line impedance between the regulator and load.  The highest performance designs use AC termination at the source or load or both to suppress resonance between the transmission line and decoupling capacitors.  This is where certain combinations of output capacitors and decoupling capacitors cause excessive ringing.

Since this occurs between the capacitors along the transmission line and does not involve the regulator's response, faster load switching is required to reveal it.
 
The following users thanked this post: TimNJ

Offline TimNJTopic starter

  • Super Contributor
  • ***
  • Posts: 1720
  • Country: us
Re: Slew Rate (dI/dt) for SMPS Transient Response Testing
« Reply #4 on: June 01, 2020, 05:20:32 pm »
Faster load switching is more of a test of the output capacitance and decoupling than the regulator's control loop and this especially applies to point of load regulators.

Ah yes, definitely a good point. I'm thinking there's a cross-over point where further increase in slew rate does not do much to test the control loop.

It is not difficult to switch a load fast enough to control the slower control loop.  Anything faster just tests the output capacitors.

Quote
Considering a power supply might have a loop bandwidth of ~10KHz, that puts the transient reaction time in the order of ~100us to even "recognize" a perturbation on the output.

Something else to consider is the impedance of the output capacitor and transmission line impedance between the regulator and load.  The highest performance designs use AC termination at the source or load or both to suppress resonance between the transmission line and decoupling capacitors.  This is where certain combinations of output capacitors and decoupling capacitors cause excessive ringing.

Since this occurs between the capacitors along the transmission line and does not involve the regulator's response, faster load switching is required to reveal it.


What do you mean by an AC termination at the source or load? What types of components and where? Or are you talking about some sort of impedance matching or something?

Thanks!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Slew Rate (dI/dt) for SMPS Transient Response Testing
« Reply #5 on: June 01, 2020, 09:26:23 pm »
What do you mean by an AC termination at the source or load? What types of components and where? Or are you talking about some sort of impedance matching or something?

This problem occurs when the bulk output or bulk decoupling capacitors have the wrong ESR.  When the power distribution wiring has a very low ESR, as will be likely in high current applications, load switching transients may cause the high Q circuit formed by the inductance or transmission line delay in the distribution wiring and the decoupling capacitors to ring.  The bulk decoupling capacitors suppresses this if their ESR roughly matches the transmission line impedance because they form an AC parallel termination.

So the bulk AC decoupling capacitors not only lower the impedance at lower frequencies but also act as AC terminations.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf