How to solve decoupling issues

[J_M_B]

Hi all

I'm working on a design with a specific spike on the 3V3 rail caused by insufficient decoupling of a bluetooth BLE module.

I know I can solve the issue thru trial by trying difference values of capacitance but I would like to know if there is a proper technique.
I assume I should be able to look at the frequency and amplitude of the spike and calculate the value of capacitor which would reduce it to within spec, in this case 10mV pk-pk ripple.


i) Could anyone advise on proper technique to quickly calculating decoupling values for this type of issue?
ii)Am I correct to assum ceramic would always be the best option due to low ESR?


Any thoughts would be much appreciated

 

[KlausST]

Hi,

Yes, ceramic are low ESR. (seldom this may cause additional ringing)

To be able to calculate the capacitor you additionally need the current.

C = I * t / U

Where t = 1/f; I is the ripple current, U is the ripple voltage (both ripple values could be in Vpp, or VRMS or so, but both need to be the same)

****
Is it now 10mVpp or do you need 10mVpp?

Often it is not necessary to go that low. To optimize, you could generate a more clean Vcc (maybe for the analog part) by decoupling with an additional series resistor.
Then with the known R and the Cs you could calculate the resulting ripple voltage..
(or vice versa)

Hpe this helps

Klaus

 

[J_M_B]

Hi Klaus

Thanks very much for your message.

- I am aiming for 10mVpp, just because spec for bluetooth module requests it.

In an everyday situation when all you have is a scope shot of a spike of dropout on your voltage rail, how would you know the current?
Is it a case of making an assumption?

 

[FvM]

I would try 10 µF/6V3 ceramic capacitor.

 

[KlausST]

Hi,

I am aiming for 10mVpp, just because spec for bluetooth module requests it.

Are you sure? This is a very tight specification.

Klaus

 

[J_M_B]

Hi

Thanks for your message.
Ive already added an extra 10uF but Im still left with the spike in the screenshot.

The reason for the post though was more to do with a structured method for determining capacitance value.
So that anytime I have an issue like this I can do a quick calculation and get a capacitance value rather than the trial and error method I usually use.

Cheers
Jamie

 

[J_M_B]

Hi Klaus

Yes,

I was more interested in the method itself for quickly getting to the value of capacitance required, rather than this specific example.

Cheers
Jamie

 

[D.A.(Tony)Stewart]

If your probing is accurate ( e.g. no probe ground inductance effects above 10MHz or ground shift )

Examine the ripple closely for AC and DC content. If there is no positive spike, it would appear to be resonant free. In your case the ripple is -20mV mostly at 16kHz but also +10mV component perhaps near 1 MHz.

Reducing Ripple can be a simple matter of Ohm's law or knowing the impedance of source and loads and estimating the additional attenuation required.

An active regulator with feedback offers a near zero impedance at DC which rises as feedback frequency response decreases. A series switch introduces an ESR ( perhaps from RdsOn).

In general if you wish to attenuate ripple, what impedance ratio do you have between source and load. Is there a switched capacitance load? That is your soruce impedance?

I use a nomograph for quick lookup of impedance vs Capacitance and determine both load pulse current and source ESR.
at 16kHz, 100uF is ~ 100mΩ. so if step pulse current is 100mA, it would cause 10mV spike.

Does ripple come from Capacitance, regulator frequency response? or ESR for unknown load current? If unknown you can test it with a known Capacitor of known ESR and pulse charge it momentarily and capture a 1 shot event, or, design it by known values or measure current with a small shunt i.e. 10~100 mV.