Good morning,

I am curious as what the community's opinion on this topic are! I recently designed a prototype for a usb camera. The design has the AVDD and DVDD on the image sensor sharing the same voltage regulator (2.8V). I had a 0.1uF ceramic decoupling capacitor (201 package) located as close to the power pins on the image sensor as possible. The camera stream looked pretty good with just a little bit of noise in the usb camera stream, but not too noticeable. Took the prototype to a test center for radiated emissions testing. Passed with flying colors.

Then there was time to see if I could figure out a way to filter out the little bit of noise in the image. All I did was change the decoupling capacitor from 0.1uf to 4.7uF (same package), direct substitute. In theory, I assume that since the noise in the image completely gone after this change, that it would not make the radiated emissions worse than when it had the noise with the 0.1uF right?

Just wondering what the opinions are.

 

If you look at an impedance v. frequency gap for a MLCC capacitor, it goes down to a minimum and then back up again. The frequency of the minimum is roughly inversely proportional to the capacitance.
So when you increase the capacitance values, you decrease the amount of attenuation at high frequencies.
You might have reduced the noise in the 100kHz region, but increased it in the 20MHz region.

 

When it comes to decoupling, multiple capacitors offer superior performance. The large value device provides supply stability when the current demand changes. The smaller devices provide progressively lower impedance to high frequency noise. It is not uncommon for there to be 3 capacitors in parallel, each with a different value. The action of the three devices is markedly different than a single device with the equivalent capacitance.

 

I would do three things.

1) Instead of sharing the DVDD voltage I would feed the AVDD from DVDD through a 100Ω series resistor.
2) Instead of choosing between 0.1μF and 4.7μF capacitors I would leave both across AVDD and AGND. They filter out noise at different frequencies.
3) If you get the opportunity to redo the PCB layout, pay attention to the GND connection. Ideally, you want separate ground planes, AGND under analog components and DGND under digital components. Connect the two ground planes where the power and ground enter the board.

 

A ferrite bead in AVdd might be useful as well.

 

MrChips has obviously dealt with a similar problem. Put a resistor or inductor (or ferrite bead) between the capacitors that will improve the attenuation, which is the whole point of the exercise.

 

Below is some info that you can refer to and modify for what you need.
Using Toroidal Coil and capacitor LC Filter for the surge and interference of uC 5V and GND.