TPS62916: Stability analysis with double feedback (second LC filter)

Lionel U

Prodigy 95 points

Part Number: TPS62916

Tool/software:

Hi TI support team

I've got a question during the design with the TPS62916.

I'm currently making simulation with SIMplis.

I'm wondering how to check Phase and Gain margin with a second LC filtering.

Indeed the TPS62916 have a multi feeback path (connection to VO and FB).

How can we simulate and check the 2 gain before doing a prototype ?

If I placed a Bode probe up to the resistor divider (classical approach), the loop gain is wrong (Fc=11kHz) as the fast lane is not taken into account.

And then the question is : how to measure it with a Bode100 for instance ?

A transient response will give a good estimation about the phase margin but we will be blind on the gain margin. Which is dangerous with a 2nde LC gain peaking.

Thanks for your support

Best Regards.
lionel

2 months ago

0 Tahar Allag 2 months ago

TI__Mastermind 32020 points

Hello,

Britton will take a look at this further

Injection the noise into the feedback network and taking body is the correct method. LC need to be included in the loop. I highly recommend doing transient after the Bode is taken just ton confirm the stability of the device. also look at the voltage ripple at steady state.

Thanks!

Tahar

0 Britton Jones 2 months ago in reply to Tahar Allag

TI__Expert 7676 points

Hi Lionel,

The 2nd filter must be removed in order to capture loop response both on the real device and in SIMPLIS. This is because the break point for loop response measurement is inside the die; the VO, FB, and injection point must all be taken directly after the inductor. Please refer to section 1.2.6 of the EVM user guide.

That said, we have done stability analysis on 2nd filter designs and can guarantee the following stable 2nd filter designs:

Best regards,

Britton

0 Lionel U 2 months ago in reply to Britton Jones

Prodigy 95 points

Hello Britton

Thanks for the feedback.

For us it's a pity do not a full qualification of the entire schematics. Maybe it will be a great feature to have in the SIMplis model an access where the fast lane and low lane feedback mix to get a realistic loop gain.

Did you also think about doing loop gain of the fast lane and low lane separately and mix them as following ?

https://www.edn.com/analyzing-power-supply-stability-in-a-multiloop-system/#google_vignette.

There is also the NISM method (if the feedback path can not be broken) but what I read from the web, this method is tricky to get accurate results.

Best Regards
Lionel

0 Britton Jones 2 months ago in reply to Lionel U

TI__Expert 7676 points

Hi Lionel,

I appreciate this feedback, I will share it with my team. For the short term, we have done full validation on the BOM's outlined in the datasheet and can guarantee stability for these conditions.

Best,
Britton

Power management

Power management forum

TPS62916: Stability analysis with double feedback (second LC filter)