Problems with Power supplies designed with Switcher Pro (TPS40054 & TPS54620)

What type of problems are you having with the TPS54620 circuits?

Testing things with the 54620s is a bit difficult due to the problems with the 12V input, but one thing is that although the datasheet states that the rising threshold on the ENABLE pin is 1.26V max, the supply does not start with 1.32V (measured) applied to it. If i force it to 3.3V  3 out of 4 supplies start and assert their power_good signal. The 1V2 supply however does not assert power_good. 

Let me state my questions about the 40054 a bit more explicit:

• I used the PSMN014-60 5658.PSMN014-60LS.pdf from NXP as top and bottom mosfets. Are these maybe not so well matched with the TPS40054?
• I changed the value of the coil (in SwitcherPro) to be bigger to improve efficiency of the circuit. Are there drawbacks to doing this?
• Does the TPS40054 require a minium load to properly maintain its output level?

I should probably also mention that the supplies are going to be used in an extended temperature application (-40 ... +75).

Any help is greatly appreciated.

Marijn

Thalesgroup.com

TPS54620 1.2V supply:

I have bypassed the 12V supply so I can start testing the 4 TPS54620 supplies. This on a non-modified proto with f_sw=182khz.

3 out of 4 are ok (if I put 3v3 on the EN inputs).

The 1V2 supply however shows oscillation on its output, and the powergood is continually asserted/deasserted.

I measured the level on the SS pin with a scope and it is stable 1.9V. So the device is not restarting due to thermal overload or UVLO.

The output level is a sinus between 0.8 and 1.8V with a frequency of around 7kHz.

What can be the cause that the other 3 supplies (which use the same layouts) are Ok, while this one oscillates?

SwitcherPro report: 3000.1V2.pdf

TPS40054 12V Supply:

With little load: the output goes up to 12V and then drops to 0V again, with a frequency of about 1 Hz. SS pin rises to 1.0V (in 50ms) and then drops to 0V again.

With 1.5A extra load: the output is steady at 10V. SS pin is at 1.9V.

So no clear oscillation. Trying a 40055 will take some more time, but will this help as there is no oscillation to be seen?

If it helps to measure other signals, I will gladly do so.

I can see a couple issues.  First is your clock circuit.  I recommend that you terminate the RT/CLK pin at each TPS54620 with a RT resistor for 200 kHz operation.  Each TPS54620 should have its own buffered clock input.  Try disconnecting you clock input and use an RT resistor to see if your 1.2 V rail comes up.

The compensation is a little conservative, but the circuit should be stable.  The inductor value in switcherpro is large (it calculates based on the output current, it should be ok, but we may want to try something smaller).

can I see some waveforms?

Disconnecting the clock from the RT pin is almost impossible as it is connected to a via very close to the pin. Besides that it would be very strange if the clock signal is the problem as both the 3.3V and 1.2V supply use the same clock, as you can see in the schematic 7462.supplies.pdf (bottom left is the 3.3V supply, bottom right is the 1.2V supply) and the 3.3V rail is ok.

Increasing the clock frequency can be done quite easily, would you like me to try that?

Of which points would you like to see waveforms? I only have an analog scope here, so I'll snap some photos of the screen if that's ok.

The TPS54620 is designed so that it should start up using the RT current to start the internal VCO and then transition over to the external clock whne the PLL locks onto the clock signal.   I cannot say for certain what mechanism may cause a problem, but I had to point out this design issue.  Can I assume the external clock is stable and running before the 1.2 V rail is enabled?

Take any waveforms that show your problem. Be sure to show the PH node, and take both wide time scale and zoomed in shots.  It would also be good if you can measure the inductor current with a current probe and loop of wire.  Also you could try measuring the loop to see if it is stable.  What is the load?  Is there a lot of capacitance that is not shown?

You information about the handling of clk mode surprises me. The data sheet states:

Synchronization (CLK mode)
An internal Phase Locked Loop (PLL) has been implemented to allow synchronization between 200kHz and
1600kHz, and to easily switch from RT mode to CLK mode.
To implement the synchronization feature, connect a square wave clock signal to the RT/CLK pin with a duty
cycle between 20% to 80%. The clock signal amplitude must transition lower than 0.8V and higher than 2.0V.
The start of the switching cycle is synchronized to the falling edge of RT/CLK pin.
In applications where both RT mode and CLK mode are needed, ....

Which led me to believe a resistor should not be necessary in clk mode.

You are right to assume the external clock is stable before the 1.2V rail is enabled. The 1.2V rail is enabled by the powergood of the 5V supply which is on the inverse of the same clock.

I can't measure the inductor current which the equipment I have right now. If still necessary I could try to arrange it at our headquarters on monday.

As I am more a digital designer, how do I go about measuring the loop ?

No large capacitance on 1.2V. Only 10uF and smaller at the FPGA.

Vout 500mV/cm 20uS/cm:

Vout 500mV/cm 50uS/cm:

Vout 500mV/cm 2mS/cm:

PH 100mV/cm 20uS/cm:

PH 100mV/cm 200uS/cm:

To be certain that the 180kHz clock was not the source of the problem, I increased the clk to just above 200kHz. The 1.2V supply symptoms stay exactly the same.

Concerning the TPS54620 designs:

I changed the loop compensation to a type III compensation and recalculated the supplies using the excel-sheet instead of switcher pro. Those 4 supplies are now stable with load-step behaviour as calculated with the excel-sheet.

One question remains though: Why doesn't switcher pro suggest a type III compensation, as from what I've read the only drawback might be one component extra?

Concerning the TPS40054/TPS40055 design:

As suggested I changed the TPS40054 to a 40055. This improved light load behaviour and the part is stable now.

There is however one issue remaining: the load-step response seems not good to me

Testing with 16V input, when a load-switch is enabled and capacitor inrush follows, the output voltage drops to  7.6V and it takes 160us (32 clock cycles) to reach 12V  again. What worries me is that I see no duty-cycle increase, it just stays at 66.85%. See scope images below.

Is this behavior as expected. If so I will need to add more output buffer capacitance.

Regards,

Marijn

Can someone from TI maybe also respond to my question involving the TPS40055? If I need to add more capacitance, than that is no problem, but I want to be sure there is no bigger issue hiding in that supply...

Regards,

Marijn

You probably will not get a response until Monday as it is a holiday in the US...