LM5002 Switch Mode Regulator emits high frequency noise

[elskardio]

Hi Guys,

I'm working on a compact power supply for a little enclosure. To make the phantom power, I'm using a LM5002 to boost the +16V to +48V.

Everything measure fine and the 48V is pretty clean... but the LM5002 emits a loud high frequency noise!

Touching the IC with my finger damps it a little... could it be related to the PCB layout?

Any help would be appreciated

Thanks


Here's my schematic:

 

[Bo Deadly]

Do you mean you can hear it emitted from the PCB or the noise is in the audio the supply is powering?

Designing SMPS is difficult. You have to pay very close attention to layout and parts. There are people who specialize in nothing but designing SMPS supplies.

 

[JohnRoberts]

Its possible for passive components to vibrate but not common. Maybe try touching them with a pencil eraser while buzzing.

Years ago I inherited a dual output switching supply design that cleverly provided two different regulated outputs from one DC to DC switcher. Instead of a single coil it used a transformer. Sadly that transformer hummed and buzzed like it needed to get spanked. That transformer was already expensive and I couldn't find a better one hoping it would be quieter. I decided to take the gordian knot solution (look it up) and turned the dual output switcher into a single output switcher feeding a three terminal series pass regulator for the second output voltage.

That was less efficient but as quiet as a church mouse....

JR

 

[elskardio]

Do you mean you can hear it emitted from the PCB or the noise is in the audio the supply is powering?

Yes the noise is coming from the actual PCB. Like I mentioned in the first post, touching the LM5002 with my finger damps the noise a little. Also the heavier the load on the 48v, the noise gets louder.

 

[volker]

Then your inductor L3 is vibrating. Put a finger on it and see if it's dampened.

 

[ccaudle]

Probably L3, although a slight chance C10 or C12 could contribute if they are ceramic.
What devices did you use for L3, C10, and C10? Any chance L3 is saturating?

 

[JohnRoberts]

I repeat... use something like a pencil eraser to probe around for a vibrating component. Your finger will introduce capacitance that could change switching frequency or circuit behavior.

JR

 

[Doc Morgan]

I've seen this when multilayer ceramic caps are used to filter the output of a boost switcher. Some of them have some pretty weird piezoelectric characteristics. Also, some polypropylene capacitors will do this.

 

[JohnRoberts]

+1 on some ceramics having piezo effects, never experienced a polypro singing but over the years random singers of multiple dielectrics.

JR

 

[Ivan K.]

Sounds more like feedback analogous to a mic near a speaker. This is likely a layout issue, but try paralleling the input & output 10uF caps with both a 1nF & 100pF caps to ground. Make sure the input and output are on opposite ends of the PCB. You could also try a low ESR cap like a 1 uF MLC X7R ceramic to ground at in & out.
Best,
Ivan

 

[Bo Deadly]

Either an inductor is vibrating or a capacitor is emitting noise. You have to pick parts very carefully. If you use inductors not designed for SMPS, they might actually vibrate because the magnetic fields are literally applying force on the wires. If you used MLCCs, they can emit noise if they're not suface mounted in the correct way.

Your design is also not isolated which is not good for ground noise.

Buy off-the-shelf SMPS wherever possible. There are 48V MeanWell supplies that are not too big. Like APC-16-350. But for a Phantom power supply you would have to load it some to keep the output from modulating.

I'm not sure I could make a good SMPS without 2-3 iterations of boards.

 

[Matt Syson]

For what it is worth 40 years ago I had an instance of acoustic feedback where the mic preamp and speaker drive amp in the meterhood (which was feeding a speaker in another room) would 'squeal' with the ceramic cap 'zobel network' of the little power amp picked up by the mic in the meterhood..
Chokes are, as previously said, prime candidates for physical noise.

 

[JohnRoberts]

Either an inductor is vibrating or a capacitor is emitting noise. You have to pick parts very carefully. If you use inductors not designed for SMPS, they might actually vibrate because the magnetic fields are literally applying force on the wires. If you used MLCCs, they can emit noise if they're not suface mounted in the correct way.

the transformer (dual voltage switcher design) I encountered was clearly designed for use in a switching supply and still sang like an angry bird... it clearly was not designed for use in a quiet room. I looked and couldn't find an alternate source, so changed to a single voltage switcher design using a dead quiet off the shelf inductor.

JR

Your design is also not isolated which is not good for ground noise.

Buy off-the-shelf SMPS wherever possible. There are 48V MeanWell supplies that are not too big. Like APC-16-350. But for a Phantom power supply you would have to load it some to keep the output from modulating.

I'm not sure I could make a good SMPS without 2-3 iterations of boards.

 

[elskardio]

Thanks guys for all the replies!

Turns out the LM5002 didn't like the regulated 16V at the input. (according to the datasheet, it's the minimum input voltage for this circuit)

The noise is gone when I'm feeding the circuit with an unregulated 19V.

Your design is also not isolated which is not good for ground noise.

Could a simple common choke work here? Any suggestions?

 

[JohnRoberts]

Thanks guys for all the replies!

Turns out the LM5002 didn't like the regulated 16V at the input. (according to the datasheet, it's the minimum input voltage for this circuit)

The noise is gone when I'm feeding the circuit with an unregulated 19V.



Could a simple common choke work here? Any suggestions?

it may have expected an unregulated capacitor input not active regulated.

JR

 

[Matador]

Turns out the LM5002 didn't like the regulated 16V at the input. (according to the datasheet, it's the minimum input voltage for this circuit)

It is super difficult to say with certainty that this is the root cause: keep in mind the control/feedback design of these chips is incredibly complicated, and almost everything ends up mattering. It's possible with a higher source voltage, that peak inductor current fell, or the time spent by the internal MOSFET driver in conduction changed, which cause the frequency to switch outside of the resonance of your design, etc, etc. Designing with these chips requires a lot of experimentation and tuning.

 

[JohnRoberts]

It is super difficult to say with certainty that this is the root cause: keep in mind the control/feedback design of these chips is incredibly complicated, and almost everything ends up mattering. It's possible with a higher source voltage, that peak inductor current fell, or the time spent by the internal MOSFET driver in conduction changed, which cause the frequency to switch outside of the resonance of your design, etc, etc. Designing with these chips requires a lot of experimentation and tuning.

It would be easy enough to test my hypothesis by adding a simple RC between the 16V regulated output and switcher input.

One could also glance at the chip application note and see what they suggest.

JR

 

[ccaudle]

It would be easy enough to test my hypothesis by adding a simple RC between the 16V regulated output and switcher input.

That is shown in the original post, 10uF/100 Ohm/10uF CRC at the input.

One could also glance at the chip application note and see what they suggest.

The OP schematic is straight from the TI datasheet, although the datasheet doesn't show an RC filter on the input.

It is possible that the RC filter is causing ripple on the input and that is dropping the input voltage below what keeps the switching operating at the expected duty cycle. Or the ripple is interacting with the switching operation to cause ripples at sub-harmonics of the switching frequency.
Without scope shots from the operating circuit it is not really possible to tell what was happening originally.

 

[Matador]

That seems plausible! I've done quite a few SMPS designs meant to utilize a regular 12V regulated wall-wart, but they all had minimum voltages down below 10V (and none were particularly high current).

Maybe you could supply just the IC pin 2 from the RC filter, and place the input to the series inductor L3 directly from the wall supply?

 

[elskardio]

It is possible that the RC filter is causing ripple on the input and that is dropping the input voltage below what keeps the switching operating at the expected duty cycle.

Thanks for pointing that out. Removing the RC filter keeps the LM5002 happy even with the 16V at the input.

I'm still tweaking parts of the circuit while looking at the scope. I'll post my findings in a few days