Can we remove this noise using polymer capacitors or by other means?
As a random hack without applying any science I inserted some polymer capacitors at random into the PDN to observe any effects.
I chose the capacitors with the most convenient packaging. The ones that allows direct soldering replacement without the need for any adapter boards.
No perceptible difference was noticed on the video output or oscilloscope traces.
Applying some science into it shows that there would be no major improvements.
Why? The factor controlling the interference pattern is the high frequency components of the unwanted noise.
As frequencies rise the impedance from the capacitance decreases which is desired there is a limit where the inductance of the components and PCB retard the ability for current to flow.
The insertion of the polymer caps provided larger and more reservoirs of electric charge but the surrounding inductances created impedance that prevent current flow into and out of the polymer caps.
What happens that upon power up the polymer caps fill up with electric charge but once filled they are unable to respond to the instantaneous power demands.
Inductance, Inductance and more Inductance!
Inductance is the property of inhibiting the change of electrical current. It's everywhere!
In capacitors, in resistors, in semiconductors and in wires.
It's like retail sale where you can "Buy a capacitor and get one or two bonus resistors and inductors for free!"
The same is applicable to capacitance - "Buy a resistor or inductor and get one or two bonus capacitors for free!"
The values of these bonuses are usually minute but when dealing with high frequencies they are a big deal.
This is where the construction and physics of polymer capacitors comes into play.
A capacitor to high frequencies is often modelled like this.
Some models are even more complex with a resistor in parallel to the capacitive element to represent leakage current amongst other things.
(Some are even more complex with an inductor in series with the resistor parallel to the capacitive element).
The low ESR (Equivalent Series Resistance) of the polymer capacitor reduces the restriction of current flow into and out the capacitor.
If you read the datasheet of the polymer capacitors it seems that the ESR is inversely proportional to the capacitance value.
This may or may not take in the ESL (Equivalent Series Inductance) as a consideration.
For a complete design all of the components are surrounded in a sea of inductors (and capacitors but I'll just deal with the inductors for now)
Even a simple wire appears like this
A plane is like this but as a 2 dimensional matrix.
Breaking up a big problem into smaller manageable bits
Just like most complex problems breaking into smaller manageable bits is good strategy.
The first thing to do is to break up the circuit into noise domains with each domain consisting of a dynamic power drawing element and the power rails that power it.
The second thing to do is to rank the dynamism of each domain.
The third thing is to rank the power draw of each domain.
The fourth thing to do is analyse the intereconnection layout and arrange the power paths so that they minimally interact with each other..
For the microbee the 13.5MHz dot clock oscillator is the most dynamic element and this is addressed first.
The efficacy of the coulomb bucket is determined by the frequencies and inductances present.
Because this is a vintage design with a DIP package high inductances. The inductance signficantly due to the physical geometry from the IC; From the IC substrate through the bonding wire, to the lead frame and out the pin and it then subsequent connection to the PCB.
This value is about 13.5nH (nanoHenries). Secondary are the PCB traces and vias which further increase this number.
With these considerations in mid the capacitor value uses was 10nF (nanoFarads)
This ends up being a generally used value.
Another source of unwanted inductance is the capacitor package type. Due the their leads, construction and size; through hole capacitors have higher inductances than surface mounted counterparts.
An issue is that as the frequency rises the usable capacitances become smaller thus can hold less charge. It's a diminishing return.
This short video shows what happened after applying some basic principles to noise control.
The PCB now looks like a total mess but the wiggly line interference from the first article has now gone leaving just a fixed interference pattern in the chrominance signal.
Figure 2. The video after some noise reduction.
As you can see it is much better although not perfect.