Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

dc/dc resonance at 76MHz

Status
Not open for further replies.

buenos

Advanced Member level 3
Joined
Oct 24, 2005
Messages
960
Helped
40
Reputation
82
Reaction score
24
Trophy points
1,298
Location
Florida, USA
Activity points
9,116
Hi

I have a DC/DC converter (sync.rect., 1.1V@13A supply) switching at 600kHz, producting 40mVpp switching noise.
This would not be a problem, but I also can see a 76MHz oscillation sitting on it at 150mVpp at every turn-on and turn-off.
How can I supress this 76MHz resonance?
 

That noise is at wich bus of circuit ?

+++
 

it is nothing to do with any of the digital buses.
When the power switches turn on or off they resonate, and this gets out to the output of the DCDC converter. The biggest problem with this resonance is the amplitude of it, too much noise on the power rail. This resonance creates bigger noise than the normal switching noise of the regulator.
 

go through these...

**broken link removed**

**broken link removed**
 

You probably want an RC snubber across the power switch.
 

Any capacitance at all on the switch node, will ring with
the power inductor. So will the inductor's self-resonance
(winding C,L). You need very short current loops, and
a set of good high frequency caps on the output to
return AC currents to the source instead of the load.
 

hi

thanks.

i think my layout is not so bad, i have designed almost a hundred DCDC converters before and I never had this high-voltage ringing problem. the connections on the layout are not tracks but copper rectangles (lenght:width =< 0.1).

I have tried to filter the noise at the output with small ceramic caps (3.3nF/0402/f_res=75MHz/300mR) and also with extra 100nF caps. Also tried at the switching node without a series resistor with both cap values. they didnt have any effect. I think this is what the analogitech article was suggesting.

i tried to connect a 100nF cap between the pins of the inductor, then the 76MHz oscillation at the DCDC output was increased. Im using a wide flat vishay inductor, probably flat-wire type inside the ferrite housing. It is like if the inductor already had a parasitic capacitance large enough to couple the hi-freq noise from the switching node to the output.
 

Filtering at the output of the converter often doesn't do much for this kind of "noise," since it is radiated everywhere and can bypass even a well designed filter. You want to eliminate it at its source, which is why I suggested a snubber (which is different than a filter).
 

what values do i need for the snubber?

is it an RC series?
is the resistor necessary? -i have tried caps at 3.3nF and 100nF between sw.node and gnd without any effect.

is it from switching node to gnd?
 

is the resistor necessary?
Yes, you'll find an optimal resistor value which achieves maximum oscillation dampening, which is related to the characteristic impedance of the unwanted resonant cicruit.

I think, it would be helpful however to identify the involved resonant circuit structure more exactly. Assuming you have a synchronous buck converter, the commutation between VIN respectively high side switch and low side switch is the critical point. The inductance of the loop formed by low side switch - input capacitor and high side switch together with the switch capacitance makes most likely the resonant circuit. Check if the input capcitors have sufficient low inductance (respectively, if you have enough capacitors connected in parallel) and if the said loop actually spans a minimal area.

Filtering at the output of the converter often doesn't do much for this kind of "noise," since it is radiated everywhere and can bypass even a well designed filter.
I think, filtering is always reasonable. According to the said commutation loop structure, interferences will mostly spread through the Vin node and the ground plane rather than the output node, which is already filtered by the output LC circuit. Radiated emission can be a problem, but not from the core components of a physical small converter. Observing radiated emissions is another word for the first filter level has been already cut out.
 
Last edited:

I think, filtering is always reasonable. According to the said commutation loop structure, interferences will mostly spread through the Vin node and the ground plane rather than the output node, which is already filtered by the output LC circuit. Radiated emission can be a problem, but not from the core components of a physical small converter. Observing radiated emissions is another word for the first filter level has been already cut out.
I've made switching supplies where the transient spikes would show up on practically every node in the circuit, even those that do not share a ground or supply. I could just probe the ground of the circuit and pick it up clearly. Short of shielding, there's no way to eliminate that sort of signal except by eliminating it at the source. This is generally with relatively high power converters though (10KW 400V boost operating at 50KHz), so it's an extreme case.
 

my supply is a 12V to 1.1V 600kHz 13A synchronous rectifier supply 14.3W. I is not that small.
so, would the RC snubber always eliminate/significantly-reduce this noise?
what is the R, C values, and how to calcualte the effect?

It would be very hard to get the parasitic values from the design.
main components:
- c_out: 2x 470uF/10mR + 1x 1u/ceramic/0402 + the decouling at the graphics processor
- c_in: 3x 10uF/ceramic/1206/25V
- L: vishay IFLP4040 0.5uH/0.88mR/32A
- FETs: STL150N3LLH5, 1x upper, 1x lower
 

so, would the RC snubber always eliminate/significantly-reduce this noise?
Reduce, yes. Unlikely eliminate, significantly reduce, possibly. You have to try.
what is the R, C values, and how to calcualte the effect?
R equal to characteristic impedance of the unwanted resonant circuit as a first estimation, Zc >= R. The latter is a compromise between effective snubber action and losses. As said, you need to identify the resonant circuit. If you have no clue about it, just try the RC values.

Based on the transistor output capacitance numbers, a characteristic impedance of 1.5 to 2 ohm can be expected.

P.S.: Sorry, I meant Zc <= R not Zc >= R.
 
Last edited:

2 Ohms, OK.
thanks.
What should be the capacitor value?
I have no clue about the resonant circuit.

The question is still there: is it a series RC between the switching node and the ground?
 

What should be the capacitor value?
Zc <= R. resonant circuit C is about 1 nF, I would try a snubber C of 1.5 to 4.7 nF
I have no clue about the resonant circuit.
I suggest the previously described current loop. The calculation results in a rather small L of about 4nH, I believe that the "layout is not so bad".
The question is still there: is it a series RC between the switching node and the ground?
Yes. Or split across the high and low side switch. Even a snubber across the input capacitors can have some effect (an the advantage of not causing switching losses).
 

ok, thanks again.

i have tried 10R + 3.3nF from sw to gnd.
it has reduced the ripple from 150mVpp to 40mVpp, measured at the DC/DC output
This should be good enough, but close to the processor I still measure 80mVpp. Strange. The noise seems higher there than at the DC/DC output. The ripple happens twice every 1.7us, so it is surely not SSN but the power supply switch on and switch off events.
 

A 10 ohm resistor should work with a smaller capacitor as well with reduced losses.

Are you sure that the increased noise level near the processor isn't caused by different probe ground connection? If it's real, there may be a problem of ground and/or power plane resonances.
 

The higher amplidue seems to be at the face of the ringing, the first pulse seems to be stronger at the graphics processor. If it was a resonance then multiple pulses would be at large amplitude. The ringing starts with the largest amplitude pulse then decays.
It might be probe grounding, I will re-test. But with the same grounding basically moving the signal probe by 5-10 mms from the inductor to one of the decoupling capacitors the noise seems to increse. I have pushed the probe ground to the GND-pin of the mentioned decoupling capacitor.
 

I fear, it's difficult to understand the observations without knowing the complete power and ground layout and the capacitor types and positions.
 

It's not surprising that you see the signal differently when probing different places. If you have a probe with a spring tip (one where you use the ground sleeve on the probe tip and connect that directly to circuit ground) use that and you should get more consistent measurement. And on the snubber, make sure you keep its loop inductance as low as possible. Keep playing around with different combinations of values; often there are certain values that work extremely well. You'll know you have it when the waveform looks critically damp (only one overshoot and pretty much no ringing afterword).
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top