TPS74401: Loop compensation for TPS74401

Hello,

Technically this design is stable, however if you wish to enhance the stability with a feed forward capacitor, can you please try C1983 = 10nF?  That is the value that I'm calculating (or in that range) to make a difference in your frequency band.  The lower values you have tried would not make a big impact unless you change (increase) your resistors.  And post back the test data in case we need to adjust further.

Thanks,

- Stephen

HI Stephen,

Changed the Cff to 10nF per your recommendation. Below is the bode plot at 0A (1st waveform) and 1A load (2nd waveform). At 1A load seems there is a significant phase boost happens at ~1Mhz frequency. Is this expected?

And with 10nF feedforward cap, is there any drawback? Since normally this cap is at pF level from what i have seen. Please kindly let me know. Thanks,

Hello,

The feedforward capacitor can be any size, technically, as it is used for both stability and noise reduction purposes.  10 nF is fine, and so would higher values if that met your design goals.  Before using large values of feed forward capacitors, you will want to understand the Pro's and Con's of using this capacitor which are outlined in the following document.

https://www.ti.com/lit/an/sbva042/sbva042.pdf

I calculated the pole / zero pair based on the first plot, which I guess was at no load.  My apologies on this, as I thought the second plot was the measurement after you had added 2.2nF Cff.  

If you wish to see an improvement around the second plot = 1A load, then you will need to shift this towards 60 kHz. 

For a Cff = 2.6nF:

Zero = 1 / (2*pi*1.13k*2.6nF) = 54.2 kHz

Pole = 1 / (2*pi*(1.13k//4.53k)*2.6nF) = 67.7 kHz

One more comment: It is possible that this device has an internal fast feedback loop, as many modern linear regulators have this feature.  As the TPS744 is a much older device, it may or may not have this design feature.  A fast feedback loop takes Vout and feeds it directly to the control of the pass element, which provides much higher bandwidth to load transient responses.  However when you take a bode plot by breaking the slow feedback loop, you may not actually capture an accurate bode plot of the system as the fast feedback loop may still be intact.  A better method is to use a load transient response to assess the phase margin.  The frequency of the ringing of the load transient will be at the most unstable frequency in the system, and the number of rings will give you the phase margin.  If your load transient test gives you very similar results to the bode plot, then you can likely trust the bode plot results.  If the load transient contradicts the bode plot, then the load transient test should be trusted.

https://www.ti.com/lit/an/slva381b/slva381b.pdf

Let me know how these tests go and if you need any additional support.

Thanks,

- Stephen

Hi Stephen,

Thanks for the details.

For the original waveforms attached: The 1st one is for 0A with default Cff, and the 2nd one is for 1A with default Cff.

Can this part be confirmed with internal fast feedback loop?

And below is the transient waveforms.

0A<->0.5A, C1983 default 100p

0.5A<->1A, C1983 default 100p

0A<->0.5A, C1983 change to 10nF

0.5A<->1A, C1983 change to 10nF

0A<->0.5A, C1983 default 100p

0.5A<->1A, C1983 default 100p

0A<->0.5A, C1983 change to 10nF

0.5A<->1A, C1983 change to 10nF