TPS65270 Loop Compensation external component values

Resolved.

Hi Eric,

Got it, thanks for your prompt reply. I'll change to 82pF and check the ramp-up curve. Will get back to you if I have further questions.

Thanks again:

Tom

Hi Eric,

I made the startup measurements with your suggested Cff = 82pF at C44 and with the maximum 5x220uF at the output and got the following ramp-up curve:

CH1 (yellow): 3.3M_EN (pin#3 of U1)

CH2 (red): inductor (output) current

CH3 (blue): output voltage 3.3M

This was done with a PCIe cellular module which didn't work with 1.6msec soft-start time (there was several stairs or plateaus but the output voltage was unable to reach the required 3.3V and remained at about 1.6V). By changing the the soft-start time to 16msec (C39=100nF) the problem dissappeared.

There are two plateaus on the above curve, one for each current peak. The other modules I tried have just a single current peak during startup and there is a just a single plateau on that curves. Just please confirm that the above startup curve looks like as expected and there will be no startup stability issues. I'll run a long term startup stability test as well in the heat chamber.

Thanks:

Tom

Hi Eric,

This is the transient response with C44=82pF:

And this is the response with C44=2200pF:

The C44=82pF is fine but the C44=2200pF version looks more robust to me. What do you think? Is there any pitfall using the 2200pF for C44?

Thanks:

Tom

Hi Eric,

I posted two transient response curves yesterday one with C44=82pF and another with C44=2200pF. Could you please give me a direct contact to you (phone, Skype or mail)?

Thanks:

Tom 

I played with TINA and used the "TPS65270 TINA-TI Average Reference Design" model with the passive values I have in the schematic.

I run AC Analysis with different values of the feedforward capacitance. It was a long time ago I learnt control theory, but based on the simulation curves I think I have to go with Cff=2.2nF and not with 82pF (C44 in the original schematic, C15 in the simulation model). Increasing the value to 10nF decreased the phase margin, therefore the value of 10nF  for C44/C15 looks too much. I think the right value is 2.2nF and this matches with my time domain oscilloscope shots as well (see earlier).

I played with the value of C11 of the simulation model (representing C6 in the schematic) as well, and see minor improvment going up from 82pF to 470pF as indicated in the schematics.

The Amplitude/Phase curves are as follows (VOUT1=Green, VOUT2=Red):

Based on the curves above, I'm planning to proceed with the following values:

C11 (model) / C6 (schematic) = 470pF

C15 (model) / C44 (schematic) = 2.2nF

Please comment.

Moreover, I would like to understand the theory behind how to select the right value for the feedforward capacitor. According to the datasheet: „Make sure that the zero frequency (Fzff) is smaller than soft start equivalent frequency”. In my case, with the 16msec soft-start time (C39=100nF) 1/Tss = 62.5Hz which means that Fzff=1/(2*PI*C44*R22) has to be less than 62.5Hz. With the recommended 82pF at C44, Fzff is around 48kHz which is far from the 62.5Hz value. I'm sure I missunderstand something, please advise.

Hi Eric,

I'm not sure how to measure the loop response curve or AC frequency response. Theoretically, I have to open the feedback path and apply sine input at the opened end of the loop. This means that I have to lift C44 and R22 and apply sine input at the lifted and connected ends of the passives. However, my understanding is that the TPS65270 includes a current feedback as well which doesn't play in this case. Therefore, I have the very simple question: how to make the suggested AC or loop response test or measurement in case of the TPS65270?

Thanks:

Tom

Hi Eric,

Please have a look at the attached oscilloscope images. They both show the same event but with different time scale (100usec and 100msec). The event is a GSM module transient current which made the 3.3M cellular rail to collapse (RED: transient current of the attached GSM module, YELLOW: 3.3M supply rail).

This is what I think about the issue. The TPS65270 regulator features a cycle-by-cycle over current protection. What happens is that the module’s transient current is over the protection current limit of the regulator, and the internal power MOSFET is immediately turned OFF and will be turned ON again at the next switching cycle. If the overcurrent condition still exists during four consecutive clock cycles (6.4usec) the regulator simply shuts down. This is exactly what happens in our case. After shutting down, the regulator waits for four soft-start cycles (about 64msec with C39=100nF) and starts to recover with a normal soft-start. If the overcurrent codition is over, the 3.3M rail ramps-up without problem. This is what you can see on the 0.1sec resolution scope image.

Do you think my analysis is correct? Please advise.

Thanks:

Tom

Hi Eric,

Thanks for your reply. This is a cellular modem supply and most of the cellular modems feature not more than 2.5A-3.0A transient peak load. Unfortunately, there are some modems which can load the supply with much higher transient current. Besides increasing the output capacitance, I also plan to increase the value of the inductor to 10uH. In this way I can lower the peak inductor current and the over current protection will be less sensitive.

Regards:

Tom