Input/Output transience for LM5146-Q1

[AJrocks]

Hi there

I am facing an issue with the LM5146 buck. I am using it for a 58V to 12V 5A application.


When the input of "58V" is given to the buck through a two pin connector, a huge transience is observed at buck input as well as output. It can be seen in the image attached to this post. The yellow waveform represents buck output and the pink one represents its input.

The input transience ranges from 21.6V to 60.8V and the output transience ranges between -3 to 24.2V.

Kindly help us understand this phenomenon.

[T3sl4co1l]

Consider what you would measure if there is impedance between grounds.

Without layout, it's not clear if this is the problem; it is currently the most likely explanation.

BTW, "transient" is the correct translation.

Tim

[AJrocks]

Hi Tim

Thanks for the correction

Can you help me understand what you said a little better.

Do you mean that because of the impedance (R+L+C) between my input as well as output ground (which may be there because of how I may have connected them on the PCB) is causing this overshoot.

Kindly excuse my ignorance.

Regards
Abhijeet

[bdunham7]

Can you easily remove C6 through C10 temporarily and see if the transient is reduced or eliminated?  How does the circuit respond to rapid load changes?

[AJrocks]

Hi there

can you help me understand that Why do you think, removing the input capacitors may cause the output voltage spike to go away.

Regards
Abhijeet

[AJrocks]

I captured these new waveforms with a series resistor (4 Ohm) before the input power connector.

The input as well as output spikes appear to be reduced.

On plugging and unplugging multiple time I did not see a peak of more then 13 or 14V (I discharged the input capacitors before every plugin).

 

[T3sl4co1l]

I mean like this.  LS might be a mixture of resistance and inductance.  What would you expect to measure in such a case?  And indeed, given the waveform, can you solve for the values?



Tim

[AJrocks]

Hi Tim

1) So do you think this Ls is due to this huge loop formed by my input, buck and output Gnd as shown in the image (Buck Layout) attached below.

2) I was trying to analyze the circuit sent by you here's what I understood
     a) As soon as the input is connected, C1 acts as a short and LS acts as an open circuit So the battery voltage appears across the inductor, soon the inductor starts conducting and the
         capacitor's current should start to fall.
     b) Based on the derivation that I did the inductor current should be iL(t)=√(C/L)*Vs*sin(t/√(L*C)), kindly correct me if I am wrong.
     c) Here what would be a good estimation for Ls and what should I use for the value of t ?

3) I captured another waveform (attached to this post), but this time with a smaller time scale so that we can see the event more clearly.

Regards
Abhijeet

[bdunham7]

can you help me understand that Why do you think, removing the input capacitors may cause the output voltage spike to go away.

Tim and I are telling you roughly the same thing in slightly different ways.  The spike is most likely related to the surge in current you get in the input capacitors.  Removing them will confirm that this is true before you start tracking down how that pulse makes its way to the outputs.  One other thing to consider is how you are measuring the output of the power supply as the connection of the probes and ground clips can make the scope sensitive to certain EMI and may show you noise that really isn't present in the output.  Your faster timebase captures seem to indicate that this is not the case, but it is still something to keep in mind.

[T3sl4co1l]

I guess?

No idea what the rest of the layout is, or where the probes were connected.  Looks like there are hidden pours, and I guess internal pours/planes too?

Your derivation I think is about right, also it ends suddenly after a half wave, I think because the input capacitor is overcharged as a result, and current is prevented from reversal by the diode (which I guess is not jumpered out on your assembly).  Your derivation does not include resistance, which, it's not obvious how much resistance there should be here, probably just some milliohms.

Note that the impedance includes paths from the power source, and to the scope probes.  I don't know how you have everything wired up.  Trace the real network and solve for that.

Aside, interesting that it looks pretty linear (sinusoidal).  Ceramic capacitors generally lose value with increasing voltage, distorting the waveform.  It may simply be because we're seeing inductor voltage here, dunno.

As for values, C is probably around 3x4.7uF and L needs to be calculated from wiring lengths, or the waveform.  Evidently iL(t) = 0 after 18.4us, or (18.4us)/sqrt(LC) = pi.

Tim

[AJrocks]

1) I am attaching a more detailed snap of my layout (Its a 4 layer board with second layers as Gnd plane and the third one as power plane).
 
2) I am also attaching a snap of my test setup.

3) I am actually applying a Step input. This phenomenon will occur only once in our system when the board is connected to the battery through a connector, after that the board is always
    powered on. The issue is, during just this one time the LDO on the 12V line fails occasionally because of this output spike.

4) I am planning on adding a TVS on bucks output https://in.element14.com/littelfuse/p4sma16ca/tvs-diode-400w-13-6v-bidir-do/dp/3010366?COM=main-search%20CMPNULL
    for now as I have some time constraints. I still need to check if it works.

kindly let me know your thoughts.

Regards
Abhijeet

[T3sl4co1l]

Ah!  The black wire in your hand is the inductor.  Move the probe ground to the board -- don't probe across both supplies at their terminals, probe relative to the board.   Put the probe grounds in the same place.  They are not isolated from each other!

I presume the battery will be isolated (literally just a battery, no other connection), so an isolated supply is a good representation.

Speaking of which, I assume the power supply has an isolated output.  If it is grounded (through the power cord), that will act in parallel with the scope ground, and still give erroneous measurements, in that case I think acting as an inductor divider, so that the apparent output spike will be some fraction of the full amplitude instead, but still there.

I don't know how to explain the LDO failures.  Were those wired in a similar way?

Tim