LM5148-Q1: Internal regulator fails

Hi Tim,

Thanks for getting back to me.

I have a test board I made with a single LM5148 and all supporting components that allows me to change voltages, operating frequency, and configuration. When I set it up for 12V operation and with VIN between about 25 and 60V and connect an external 5V regulator to VCCX it works fine. Minimum current consumption on my bench supply is only 28mA. At that operating point there is no load on the output and the FETs are turning on intermittently to maintain output voltage, which is 12V +/- a few mV as shown by the display on a connected electronic load.

If I then apply a load everything works fine up to my design limit of 7A. Next I disconnect the external 5V regulator and tie the VCCX pin to ground through a zero ohm resistor. At that point the LM5148 ceases to function. Current draw is zero A. I thought at first that might be a consequence of starting up at no load, so I tried applying various loads before startup but that made no difference.

It acts as if the internal regulator has failed, but I don't really believe that. I've seen this same behavior on multiple parts.

Hi Tim,

I'm not sure what you mean by DAP -- there's no pin by that name in the datasheet. The issue does occur at lower VIN -- I did tests at 28V and it happened then. Even at 28V and 28mA that's 750mW which is excessive for a package that small. The thermal pad is tied to GND, but no thermal vias are in place. During tests, the power supply is unloaded, so all that current is going through the package. Is there any setup which would cause it to be that high? Also, it could be thermal shutdown, but it seems to happen very quickly if so. When I power it up without external VCCX I don't even get a blip on the output voltage -- it just settles at a few hundred mV. And by the way, these tests were all with my test board which has nothing on it but an LM5148 and supporting components. If you would like me to do other tests with that board, I'm happy to do so.

I think you've at least partially addressed my issue. I'm going to do further testing with an external VCCX supply. I don't really trust that the part will be reliable with a watt or more of baseline power consumption before the gate drivers even get going. But, I am curious if you have any other insights.

Kind Regards,

Brent

Yes, the DAP is connected. I can't measure Vin current on my present test board, but I just redid it and will have it back and ready to test in about 10 days. I put in a jumper to allow direct measurement of Vin current. I'll be in touch with the results. Meanwhile here's a zip file with the calculations from the TILM5148_Design.zip Designer and a schematic sheet showing my two designs. It's worthy of note that the 5V version seems to work fine even without the VCCX supply. As you can see, the two designs are identical except for passive values.

Sorry for the file link in the middle of the paragraph. There doesn't seem to be a file upload button so I dragged and dropped, which seemed to work.

Thanks - send on the complete quickstart calculator file as well.

Where would I find that file? I designed the supply using TI's online Power Designer.

I downloaded the file but ran into problems filling it out. It asked for a password and then locked up. I had to kill the process. In any event, how does that differ from your online Power Designer tool which I used for the original design?

Thanks Tim,

I'll fill it out when I get a moment. But, FYI, I don't use Microsoft Office -- I use LibreOffice instead and the previous version of the spreadsheet I downloaded was problematic.

Meanwhile, it's obvious that the LM5148 is not simply a linear circuit so I'm not confident the spreadsheet will be useful -- there's a processor or logic circuit in there that's handling initialization at least. That makes it impossible to figure out what's going on. Under what conditions does the part simply fail to start? Under what conditions is the internal linear power supply disabled even if VCCX is grounded? Are there any diagnostics that can be run to find out if the part is behaving properly? What are the tolerances for the resistor values that set various configuration options as some are not standard values? What is the timing of the initialization steps? It would be very useful if you could point me to an application note that has this kind of information. Otherwise I'm flying blind.

Thank you,

Brent

Brent,

This is an analog buck controller IC - please refer to the LM5148-Q1 data sheet.

Have you calculated the expected VCC current with your FETs? Please send a scheamtic.

Regards,

Tim

Hi Tim, I sent a schematic earlier, but I will resend the schematic of my latest test pcb. I'm expecting these to arrive by Monday next week. I have calculated the expected VCC current with my FETs. It's around 70-80mA, I don't have the exact calculations handy at the moment. I realize that's excessive for the internal LDO at my input voltage so I do plan to use an external supply. The note about the series resistor is a test I want to do to see if I can cut the heat dissipation in a small low cost wide voltage range LDO, but it's not relevant to the issues I've encountered so far and can be ignored.

Please note, that what I'm encountering is a failure of the LM5148 to start at all under as yet unclear circumstances. The FET current has nothing to do with it as I'm testing at no or very light load. The part is operating in pulse skipping mode and the average FET current is probably under a milliamp. I believe you when you say it's analog, but the data sheet makes it very clear that it's not linear. It can't be if it is sampling resistor values at startup and then using that information to change later operation. And that's the part I need to understand.

Kind Regards,
Brent

3443.schematic.pdf

Thanks Tim,

I try some of that. I don't want to change FETs though as these are both high current power supplies (10.6A for the 5V, 7A for the 12V) and I'm concerned about heat as the box is full of motor control FETs operating at 40A.

I'm expecting some new test boards in this weekend and I'll do further checks. Thinks for checking the schematic. I'll leave the issue open for now and report back on what I find.

Brent

Thanks, Brent.

Hi Tim,

I assembled a new test board with components set up for: spread spectrum, no diode emulation, internal compensation, 12V out and 300kHz. The board came up seemingly fine with output stable at 12V, no load. However, current draw was 134mA. More interestingly, the part powered up with current draw at 1mA, but then the power surged to 134mA after about 1 sec. At the same time, the FETs and inductor got very hot.

I tried several things: turning off spread spectrum, reducing frequency to 200kHz, and finally switching on diode emulation. Only the latter made any difference. Once diode emulation was enabled, everything ran cool and current draw at no load was  less than 1mA. However, the part still doesn't work properly. It shuts down at very low output current -- less than 2A. Changing the sense resistor value only made slight differences, going from about 1.5A at 5 milliOhms to 1.8A at zero (a solid copper shunt in place of the resistor). My other working 12V supply met spec of 7A with 5 milliOhms and otherwise identical setup.

The last thing I tried was changing from internal to external compensation using the values recommended by your spreadsheet. It made little difference -- the no load voltage was actually less stable and the supply still shut down at 1.8A.

My conclusion so far is that this is an incredibly frustrating part to work with. I've built around 15 of them of which only 5 work at any output V, only 2 at 12V and only 1 at 12V, 7A. I'm at the end of my rope and am ready to design it out. Any final suggestions?

Brent

Hi Tim,

To follow up on my previous reply -- I decided to do one more test this morning. I removed the 5148 from my test board after doing some temperature tests on my rework equipment with the goal of seeing how little heat I could apply to the part when hand assembling. Armed with that knowledge I attached the part to the thermal pad being careful to apply just enough heat to reflow the solder paste under the thermal pad. I previously had created a new footprint for the part optimized for hand soldering. So I then carefully applied solder to the pads making sure the solder flowed onto the wettable flanks.

This time it worked. I was able to run the part both in and out of diode emulation mode, quiescent current was around a milliamp and i could go all the way to my design limit of 7A.

My conclusion is that the part is rather fragile compared to others I have worked with. This especially shows up during development work when parts have to go through handling abuse as well as less than tightly controlled soldering temperatures during hand or bench top reflow oven soldering. It doesn't help that it's only available in a QFN which is very difficult to hand solder and rework. I avoid QFN's wherever possible and I think I'm going to double down on that. What a pain.

Anyway, thanks so much for your help. I think this resolves my issue. I did go ahead and design an LM5085 based supply as a backup using a new generation low RdsOn PFET and a FERD as the flyback diode. We'll see if that can handle the heat and I'll decide which design to go with based on tests.

Kind Regards,

Brent