Hello, I am currently working on bringing up a PCB we have designed with a LMR36015-Q1 being used to convert 36V to 5V.
Unfortunately, the converter is not working properly. When 36V is applied, the converter appears to be stuck in the short circuit hiccup condition. I can see ~20ms bursts where the convert attempts to regulate the voltage followed by 94ms pauses.
Yellow = VOUT, Green = SW
I have confirmed with a multimeter that there is no short at the output so something else must be causing the hiccuping.
On the startup, it appears as though the converter starts in PFM and then transitions into PWM mode as the output voltage increases. The converter hits its 400kHz switching frequency briefly; however, after this the switching frequency falls to ~100kHz. The output voltage stays constant at around 1.5V.
With the output voltage stuck at 1.5V, the feedback voltage divider only reaches around 0.3V, this is below the 0.4V short detection threshold and I believe then the reason why the converter stays in this hiccup mode.
Green = VOUT, Orange = FB
Here is my schematic:
All of the values are directly from a WEBENCH design, with the exception of an additional 4.7uF input capacitor I added because I was concerned about DC derating.
Here is the layout:
This is done on a 6 layer board with 2oz copper on the top. The feedback trace is routed on layer 5. Layer 2 and Layer 4 have ground pours under the converter.
A couple questions:
1. Is there any reason why the LMR36015-Q1 might be clamping the output voltage to 1.5V? Or is this something that another device in my circuit might be doing? There is not much connected to this converter, A 3.3V linear regulator, two I2C buffers, two 5.6V zener diodes. I think none of these are likely to clamp the voltage at 1.5V. Also, the load on the converter should be very small, on the order of 50-100mA or less on startup.
2. Could such an issue be cause by noise pickup on the FB trace? I am realizing now that while my FB trace runs under the PCB ground planes, it remerges directly under the switching inductor... To make matters worse, the recommended inductor I selected from WEBENCH is a Bourns SDR1105-330KL which is unshielded, something I previously overlooked. I also ran a ground pour under this unshielded inductor - I have seen this recommended against in some app notes I believe. I am see ~300mVpp of noise on the FB node. This seems like quite a lot to me.
Could noise on the FB node cause this kind of issue? Perhaps re-routing the trace and using a shielded inductor could improve this?
3. Could this be an issue with converter stability?
I would appreciate any advice that you might have on this issue. Please let me know if you need any more information.
Thank you very much,
David