Decoupling question for fluorescent ballast

Other Hardware General Electronics
1

Hi, the last question on this specific topic was posted a year ago, and related to sensor accuracy. So, I trust I am not treading on toes when I ask for help on this issue that causes undesirable resetting of my microcontroller.

I have done a bit of reading up on decoupling, am aware of the basic needs, and have applied them to my limited understanding.

I have done a google search on the topic, and understand that fluorescent ballasts generate a high frequency noise which needs to be dealt with.

As I understand, I need to calculate the frequency of the hum, and attach decoupling to compensate for this. According to my calcs, that results in a whopping 666MHz.
Is my selection of the decoupling capacitors fair? I have attached the spec of the electronic ballast (30W Arcadia) onto the image. Lambda = 0.45, A = 0.365 C alpha rho = G13
I strongly suspect it isn't.

I have attached 3 decoupling capacitors: 10mF, a 0.1mF and a 15pF on the powerline next to the input jack.

Attached is a block diagram of my connections.

I am using a DC Boarduino with the AtMega 328 chip.

I installed an optoisolator BOB to protect the microcontroller from the relay switching noise. In spite of this, I still observed my device resetting. Therefore, I decided to remove the relay and have the Flourescent ballast connected directly to the AC line. This resulted in the same problem occurring when switched on from the physical switch.

Yes having a fluorescent tube IS essential for the project.

The microcontroller and power supply are in separate locations: each with their own faraday cage.

I installed a 120 ohm resistor between the RST and 5V pin to prevent stray resetting. This just halted the program completely when the resetting occurred.

I do not have an oscilloscope. However, if my project does turn out successful, I might be rewarded with one.

Does anybody have an opinion on the power supply?

Should the Boarduino have a separate supply to mitigate the issue? I have a 9V 2A unregulated MYRRA PS lying around.
Would this mean connecting the grounds on both supplies together?

If you think of any part that I could do with to help, I will buy it from Farnell.

Any and all help will be appreciated.

Thank you

ardy_issue-2.png
ardy_issue-2.png477×505 23.5 KB

2

How did you calculate that "hum" value? Hum is usual the mains frequency. Anyway before you get to carried away with that I would check out the basics:

Have you got a pull-up resistor on your reset pin?
Does the board reset if you use an LED or incandescent lamp instead of the fluorescent?

A detailed schematic may help us...

3

Fluorescent lamp electronic ballasts are switching power supplies that usually operate in the ultrasonic frequency range, i.e. 20-40 kHz.
A high quality AC line noise filter MAY help. For more ideas, search for "electronic ballast line noise filter".

4

Decoupling tries to reduce noise across the whole frequency band, but fails at the
high frequencies due to self-resonance and stray inductance, and at the low frequency
end by the increasing impedance of a capacitor at lower freqnencies.

But noise injection is more complex than that - no amount of decoupling will prevent
electromagnetic induction, since any conductor subject to a varying magnetic flux shows
an induced voltage that is fundamental. Kirchoffs laws do not apply here.

Good decoupling reduces the effects of capacitive coupling by having a large charge
reservoir to keep the voltage from changing much, however to reduce induction you
must reduce the area of any loops in the circuit, since magnetic flux is dependent on
the magnetic field change and the area. Placing the circuit board away from the source
of interference, or in a metal box, will directly reduce the problem.

Twisted pairs are good for rejecting induction since the area between the conductors
constantly inverts in sense, cancelling out. It also has a similar rejection of differential
mode capacitive coupling. Good shielding will exclude rapidly changing fields from
entering the cable in the first place (eddy currents on the outside of the shield in
theory exactly cancel the magnetic field change). This effect works best at high frequencies
and with a quality screen (not lap-wound).

The fluorescent ballast is basically a step-up transformer in the ultrasonic frequency range,
induction is an important means of interference to be considered. The wiring from the
ballast to the tube will also be generating lots of high voltage (able to capacitively couple).

Use twisted pairs or shielded cable, don't run sensitive signal wires alongside noise sources,
shield your circuits in metal enclosures, consider RF filtering on the input and output cables...