Author Topic: Do LC power-on spikes usually matter?  (Read 1544 times)

0 Members and 1 Guest are viewing this topic.

Offline InfravioletTopic starter

  • Super Contributor
  • ***
  • Posts: 1098
  • Country: gb
Do LC power-on spikes usually matter?
« on: September 19, 2022, 06:30:14 pm »
I've just been reading this:
https://www.pololu.com/docs/pdf/0J16/destructive_LC_voltage_spikes.pdf
about LC spikes, particularly common when "large" MLCC caacitors are in use between Vcc and Gnd.

Are these problems common in circuits in practice or pretty rare really?

I've put an oscilloscope on some 5V circuit designs of mine (most designs having 2 to 4 10uF MLCC ceramic caps between Vcc and Gnd) and found spikes reaching about 6.8V for 10uS during power-up, is this sort of thing within the tolerances of most 5V chips*, or should future things I design add high ESR electrolytic caps or small resistors in series on the power rails (for those with low current consumption) to suppress these?

Thanks

*It is above the Abs Max on most datasheets, but it is also really short
 

Offline strawberry

  • Super Contributor
  • ***
  • !
  • Posts: 1199
  • Country: lv
Re: Do LC power-on spikes usually matter?
« Reply #1 on: September 19, 2022, 07:20:05 pm »
scope probe loop could be problem here and breadboard inductance as well
usually 5V chips are specified in datasheet to max voltage ~6V + some manufacturing margin
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16898
  • Country: us
  • DavidH
Re: Do LC power-on spikes usually matter?
« Reply #2 on: September 19, 2022, 08:35:35 pm »
It is a real problem, however as strawberry pointed out, make sure what you observe is not caused by your probing technique.  This is a good place for a fast low voltage differential probe, or two passive probes used in add and invert mode to make a differential probe.
 

Offline TimNJ

  • Super Contributor
  • ***
  • Posts: 1670
  • Country: us
 

Offline InfravioletTopic starter

  • Super Contributor
  • ***
  • Posts: 1098
  • Country: gb
Re: Do LC power-on spikes usually matter?
« Reply #4 on: September 20, 2022, 08:52:04 pm »
TimNJ: Thanks, that's a good link. Hadn't realised one could add an extra MLCC cap with a resistor in series with it as a parallel combination to the existing caps to fake the effect of a high ESR cap type while still having the convenience of MLCC form factors.
 

Offline BlueSmokedAgain

  • Contributor
  • Posts: 12
  • Country: gb
Re: Do LC power-on spikes usually matter?
« Reply #5 on: September 21, 2022, 12:12:34 pm »
Hi All, new to all of this so sorry if this is all incorrect! Attached is a schematic of my circuit, which has the problem described of going to 8v ish when turned on. I don't have scope, but its there long enough for a multimeter to detect it, before settling down to 5v again. C2+3 are jellybean ceramic caps, C1+4 are aluminium caps (digikey 732-8636-1-ND). Is this the above discussed issue or is there something else going on? Many thanks
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16898
  • Country: us
  • DavidH
Re: Do LC power-on spikes usually matter?
« Reply #6 on: September 21, 2022, 05:21:16 pm »
Hi All, new to all of this so sorry if this is all incorrect! Attached is a schematic of my circuit, which has the problem described of going to 8v ish when turned on. I don't have scope, but its there long enough for a multimeter to detect it, before settling down to 5v again. C2+3 are jellybean ceramic caps, C1+4 are aluminium caps (digikey 732-8636-1-ND). Is this the above discussed issue or is there something else going on? Many thanks

That should not be happening.  C2 and C3 are not doing anything useful and should be removed, but they are not causing the problem.

The regulator might be of a questionable design.  Who made it?
 
 

Offline BlueSmokedAgain

  • Contributor
  • Posts: 12
  • Country: gb
Re: Do LC power-on spikes usually matter?
« Reply #7 on: September 21, 2022, 07:39:42 pm »
Thanks for getting back to me. Regulator is by Mornsun America digikey part 2725-K7805M-1000R3-ND purchased from digikey. Data sheet attached or on digikeys website. It’s being powered from an ok but not amazing power supply, but I thought one of the reasons to use this is to filter out any potential supply problems. If those lager caps aren’t needed, would you use other values instead? Or just nothing further needed? The data sheet suggests 10uf input and 22uf on the output, This board is someone else’s design so I’m trying to work out if it’s me doing something wrong or if it needs a redesign
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16898
  • Country: us
  • DavidH
Re: Do LC power-on spikes usually matter?
« Reply #8 on: September 21, 2022, 08:17:26 pm »
Thanks for getting back to me. Regulator is by Mornsun America digikey part 2725-K7805M-1000R3-ND purchased from digikey. Data sheet attached or on digikeys website. It’s being powered from an ok but not amazing power supply, but I thought one of the reasons to use this is to filter out any potential supply problems. If those lager caps aren’t needed, would you use other values instead? Or just nothing further needed? The data sheet suggests 10uf input and 22uf on the output, This board is someone else’s design so I’m trying to work out if it’s me doing something wrong or if it needs a redesign

Oh, so it is not just a 7805 clone, but it is a switching power supply module intended to replace a linear 7805.  Well, it still should not be doing that.

Start by using the capacitor values and types recommended in the datasheet, at least for the output capacitor.

The datasheet suggests that the minimum load current is 100 milliamps.  Is that being met?  Maybe place a pair of 100 ohm 1/4 watt resistors across the output to see how that affects the startup voltage spike.


 

Offline InfravioletTopic starter

  • Super Contributor
  • ***
  • Posts: 1098
  • Country: gb
Re: Do LC power-on spikes usually matter?
« Reply #9 on: September 22, 2022, 01:07:55 am »
BlueSmokedAgain: are you sure it got up to 8V? Because multimeter measurements of brief events and during any sort of changing voltage can be a bit weird. I've seen 8V or more read on a multimeter for half a second while turning on various circuits, but an oscilloscope check (mine is not a fancy scope, not differential probes or anything) showed nothing quite so big as that. Infact some circuits which, by the oscilloscope reading, don't seem to spike at all on power up ,things with big high ESR caps somewhere between their power rails that turn on slowly as the caps gradually charge,  still give brief mutimter readings of well above the voltage they actually get to.
 

Offline BlueSmokedAgain

  • Contributor
  • Posts: 12
  • Country: gb
Re: Do LC power-on spikes usually matter?
« Reply #10 on: September 22, 2022, 08:00:47 am »
Hi both, thank you for your replies. So I have checked the load and it’s minimum 130mA and I am still observing the 8v momentary read. I’m starting to think it’s possibly my meter, I’m going to ask around and see if I can get access to a oscilloscope and measure it on that :) I’ll also order some correct value capacitors as per the data sheet. I have some of the correct value in tantalum, but that’s not ceramic, is that ok for a test? 
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16898
  • Country: us
  • DavidH
Re: Do LC power-on spikes usually matter?
« Reply #11 on: September 22, 2022, 05:11:08 pm »
Hi both, thank you for your replies. So I have checked the load and it’s minimum 130mA and I am still observing the 8v momentary read. I’m starting to think it’s possibly my meter, I’m going to ask around and see if I can get access to a oscilloscope and measure it on that :) I’ll also order some correct value capacitors as per the data sheet. I have some of the correct value in tantalum, but that’s not ceramic, is that ok for a test?

Try the tantalum and see.  It would be unusual if it had problems with it.  Of course it should not be having problems with your current configuration.

I thought it unusual that the application notes recommend such a large ceramic capacitor, but modern designs tend to go that way.

I have never had a problem with overshoot when using peak-hold mode on any of my multimeters, but that should be checked.  If there is overshoot, then I would expect a multimeter to miss it.  An oscilloscope is the proper tool for this.
 

Offline InfravioletTopic starter

  • Super Contributor
  • ***
  • Posts: 1098
  • Country: gb
Re: Do LC power-on spikes usually matter?
« Reply #12 on: September 22, 2022, 10:12:34 pm »
To just sum up the answers to my orginal point in the question, you're saying that these power on spikes do present a serious enough hazard to the longevity of chips to be worth worrying about in most designs one may produce. And that solving them can be as simple as adding some extra MLCC caps but having these be in series with some resistors before they go to the power rails, hence faking the properties of a high ESR electrolytic cap whilst still having the form factor and longer lifetime of a ceramic?

The image below is NOT the right circuit, but finding some generic series and parallel diagrams looked the best way to state more clearly than in words what I meant by that last sentence:
https://electricalacademia.com/wp-content/uploads/2017/10/Figure-3-Series-Parallel-Circuit-2.gif
Consider only the circuit on the left
Ignore the absurdly high voltage battery, replace it with a low resistance power supply of Vcc, and with some amount of inductance in the wires from it to the area near the caps and chips
"R1" is where the existing MLCC caps in a circuit design are
"R2" would be a resistor of few ohms
"R3" would be some new extra MLCC caps
and the power rails towards the chip ofcourse would branch off from either end of "R1".

I'm guessing for 10uF to 100uF of MLCC's to compensate for you'd be looking a similar amount of MLCC's in series combinations, and if you had to put these a fair few cm away near the edge of an existing board design the trace lengths from them to the main MLCC's near the chips wouldn;t be long enough to cause more inductance sisues, plus they'd slow the rise of the power rails anyway so a sharp step wuld never reach the chip's deocupling MLCCs to cause any spiking.
« Last Edit: September 22, 2022, 10:28:29 pm by Infraviolet »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16898
  • Country: us
  • DavidH
Re: Do LC power-on spikes usually matter?
« Reply #13 on: September 22, 2022, 10:31:54 pm »
To just sum up the answers to my orginal point in the question, you're saying that these power on spikes do present a serious enough hazard to the longevity of chips to be worth worrying about in most designs one may produce. And that solving them can be as simple as adding some extra MLCC caps but having these be in series with some resistors before they go to the power rails, hence faking the properties of a high ESR electrolytic cap whilst still having the form factor and longer lifetime of a ceramic?

There are a few different things which can cause overshoot on the output of a regulator.

The frequency compensation of the regulator may be unhappy with the combination of load capacitance and ESR.  The higher ESR of a capacitor adds a zero into the frequency response, which translated means it adds phase lead inreasing stability.  Some regulators will outright oscillate if a large ceramic capacitor is used, and some newer devices designed for a "zero" ESR capacitor will perform poorly with a high ESR tantalum or electrolytic.

The inductance of a circuit combined with a shunt capacitance can cause destructive ringing.  This could be applicable to the input to the regulator if the input capacitance is small and the series inductance, like from a transformer, is high.  This also comes up if the output of the regulator is switched with a transistor producing a rapid rise time which can excite the parasitic inductance of the output circuit wiring.

The regulator may be designed poorly.  One trick switching regulators use is to make the output current of their transconductance error amplifier asymmetrical, so that during startup when the output capacitance is charging, the integration time constant of the error amplifier does not "wind up" causing overshoot.
 

Offline BlueSmokedAgain

  • Contributor
  • Posts: 12
  • Country: gb
Re: Do LC power-on spikes usually matter?
« Reply #14 on: September 23, 2022, 01:18:32 pm »
Hi all, thank you for your help with this, looking at it with a scope it doesn’t seem to overshoot, must be something to do with my multimeter, lesson learned and scope is on my wish list 😜 I will change the capacitors to the correct values also for completeness :)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf