TPS548D22: schematic review

Louis Lo 

1) I would not recommend a 0-ohm resistor between the DRGND pin and the exposed thermal pad.  The resistor and resulting traces is likely to add more inductance and noise to DRGND and BP, which is bypassed to it, when the drivers turn on and off and need to conduct large currents through the DRGND.  I would recommand making the DRGND a direct connection to the thermal pad.

2) The compensation design with R x 3 ramp, 400nH inductor and 650kHz switching frequency is setting the output impedance at about 3mOhms.  A 100% dynamic load current of 35A will result in about 110mV drop (3.2%) on Vout.  If that level of transient response is not needed, other FSEL ramps may be selected.  The transient drop will increase inversely with the time constants provided in table 7 of the datasheet on page 25.

3) The stability of the loop as currently designed depends heavily on the ESR of the 120μF capacitors.  If the ESR of these capacitors is more than a few milliohms, the loop will likely be unstable.  You have a number of potential options:

Increase the ceramic capacitance so that the loop is stable without the electrolytic capacitors. This would require an increase of about 5x the capactiance.

Select a larger D-CAP3 compensation Ramp with a lower time-constant from Page 7

Change the reference voltage programmed by VSEL and update the VOUT to RSP divider to increased the division from Vout to RSP

A combination of two or more of the above.

If you need the tight transient requirements above (110mV drop on 35A dynamic load change) I would recommend increasing Vout.

If you do not need the tight transient requirements above, selecting a lower ramp multiplier using FSEL and a lower reference voltage VSEL in combination would provide the best path to stabilizing the current output capacitor bank.

Select R x 1 Ramp with R608 = 22.1kΩ

Select Vref  = 1.1V with R607 = 110kΩ

Combined, these will increase the dynamic output impedance to 11mΩ for a 388mV drop on a 35A dynamic load change, or a 5% drop (165mV) on a 14.8A dynamic load.  If that is sufficient dynamic performance for this rail, those changes would be the easiest to implement.

4) I would recommend connecting R611 (resistor to VOSNS) to the junction of R614 / R615 and C614 rather than directly to the output so that the VOSNS input can also benefit from the R614 /. C614 filter, and the loop injection for a network analyzer can be placed across R614 and include both VOSNS and RSP.