TPS54540: Using an SPI potentiometer in the feedback path

Hi Nicholas,

Is the converter stable without using the microchip MCU? Normally we adjust the voltage based on this app note

http://www.ti.com/lit/an/slva861/slva861.pdf

You have a top and bottom feedback resistor and you can inject a current or make sure that the RSET is close to the FB pin

Thanks

-Arief

Hi Arief:

I am able to replace the digital pot with a mechanical SMT 10k pot. The top resistor is 75k and the bottom is 5.6k so I should have enough current through the feedback path for a stable voltage at FB. I can set the pot to give me, say, Vo=8v @ a load of around 8 mA. If I then apply the load (around 300 mA) Vo drops to about 6v and fluctuates about 300 mV either way. As it turns out, this happens with the mechanical pot at both extremes as well.

At this point, I can only assume that this problem has something to do with instability in the FB loop. For the TPS 54540, it is better to have the pot in parallel with the top or bottom resistor as opposed to series? The app note seems to suggest this as an alternate to injecting a low current source between the 2 resistors at the FB pin.

So, I have eliminated the pot from the circuit and I now have a typical 2 resistor divider for FB, with Rh = 75k and Rl = 11k. This gives me Vout=6.27v (the calculated Vout is 6.25v) With this configuration, Vout is still very unstable and drops to around 5.5v with a load of about 400 mA. At around 200mA, the output voltage actually increases to about 7.5v and fluctuates around that point. At load currents below 100mA, Vout looks to be OK.

The RT resistor is set to give a 1MHz switching frequency and we are using a 3.3uH inductor.

It looks as though the PCB layout shouldn't be a problem although I would like to have someone at TI review the artwork if that is possible. Thanks

I should also mention that the package center pad is properly connected to GND.

Attached is a short video of what I am seeing on the SW pin as well as on Vout. The PCB with the LEDs on it are the app for this part. The part is used to supply power for the LEDs. This board is one that is used in an electronic message sign.

As can be seen on the scope, the part appears to be changing between CCM and DCM operation in addition to the changing switching pulse width. There are also sporadic times in which the device stops switching altogether.

Are there any other avenues for support of this problem other than this forum?

My next step was to disconnect the supply from the load on the PCB so that I can connect resistive loads of different values from Vout to GND. I am finding that I cannot seem to draw more than about 150 mA from the supply before it goes unstable and Vout drops to well below it's set value so it seems that, for whatever reason, the supply cannot source anywhere near it's rated current of 5A. 

I have also opened a case for this problem on TI.com 

When can I expect a followup on this case, if at all? My problem at this point appears to be related to component values that do not agree with those determined by your Webench app. If this problem cannot be reviewed and resolved soon, I will have no choice other than to respin the board with a supply from another vendor.

FWIW, our EAU for this part will be around 30k / yr

I see that the 54540 is replaced with the 54540B. Will try that...

Sorry for the delay.

LM61460 and LM73605 might be an option.

Nevertheless, if you would like to continue with this device we will support it, but the above devices have recently RTMed and are synchronous, which can be argued to be easier to design-in.

Have you analyzed your compensation components and output network? Please share your schematic.

Do you have a secure area to which I can upload design files?

The compensation components are values that were determined using the Webench tool using our anticipated input / output voltages and output currents. The RT resistor is a 100k which sets the switching frequency to about 1 MHz.

The only major difference between our design and the Webench / datasheet example is the addition of a 10k digital potentiometer in series between the top and bottom resistors in the feedback network with the wiper connected to the FB pin. I can bypass the pot to create a simple and typical 2 resistor divider which did not solve the problem. 

Looking at the SW pin with a scope, it looks like the regulator fails whenever it is trying to work in CCM operation which, on average, seems to happen when I try to draw more than about 150 mA.

As mentioned above, I noticed that the LM54540B is the replacement for the 54540 but I get from your reply that this part is no longer supported either. The LM73605 looks like a good replacement and the LM61460 looks to still be in the preliminary stages. If the 54540 / 540B is no longer supported, I am probably better off revising the board to use one of the newer parts, especially if the 54540 doesn't work as intended.

Hopefully, the 54540B will work for us since it's a drop in replacement but I am curious as to it's EOL status?

It may make more sense for us to rev the board now for one of the new parts.

Thanks

FWIW, we tried the 54540B version of this component. Although "better" than the 540, it too becomes very unstable after a few minutes of operation when sourcing more than about 150 mA of current. A quick shot of cold spray restores operation until the part warms up again.

Apparently TI no longer supports either version of this part. Not surprising since it obviously doesn't work.

I will upload my design files if you provide me with a secure link. Otherwise, we are ready to move on and rev the board with a new component, either from TI or another manufacturer. Tech support is among my top priorities when selecting a new component which is currently steering me towards another choice.

I would greatly appreciate any kind of a response at this point, even if it is simply to close out this case.