Parasitic oscillations are one of the four typical causes for emissions in EMI/EMC problems. Try to reduce the gain or break the feedback, and the problem could be solved at low cost.
When failing in radiated or conducted emissions, the origin of the undesired emissions is usually one of four possibilities:
- main frequency (or harmonics) from oscillators, clocks, or switching circuits in the design,
- non-linear combinations (mixing effects) of several frequencies used in your circuit,
- ringing in underdamped parasitic resonant circuits, and,
- parasitic oscillations
A parasitic oscillator is an undesired oscillator. RF amplifier designers know a lot about this subject when trying to design stable amplifiers (remember “Amplifiers will, oscillators won’t”).
Why can a circuit oscillate?
To create an oscillation, you need two essential conditions: gain and feedback (Figure 1).
At the frequency where you have gain and the appropriated feedback, the circuit will oscillate.
Condition for oscillation is OPEN LOOP GAIN (A·F) equal or greater than one:
|A·F|=1 and Phase(A·F)=0°
So, when designing a circuit, many blocks, analog or digital, low or high in power, can offer gain at some frequencies, if a parasitic or undesired feedback is created (i.e., with the layout of the circuit), an unexpected high-frequency signal will be created.
The signal can then be injected in mains or power system (conducted emissions) or radiated by cables or any other metal structure.
That is because the layout of input and output pins in components or circuits must be avoided.
To discover the parasitic oscillator, we can use a set of near-field probes. Then, the gain and feedback must be located typically using the schematic and PCB layout of the circuit.
The solution is usually to reduce gain with ferrites or resistors in the gain path (Figure 2 left) or opening the loop to remove feedback (Figure 2 right).
Recently, I was involved in a radiated problem at 3 meters (Figure 3). Emissions in the VHF range (close to 155MHz) were related to a frequency not related to those in the nominal design (clock, harmonics, etc.).
Using the near-field probe, I identified the area of the PCB where the signal had higher amplitudes, which corresponded to the hot area.
The area was related to a small DC/DC converter switching in the kHz range. The radiated signal was not a harmonic of the converter. No ringing in that frequency was found in the circuit, so a parasitic oscillation was considered as a possible culprit for this problem.
In Figure 4 (left), the schematic of the DC/DC converter is included.
A +24Vdc level is converted to +5V with a TPS5410D IC. The layout of input and output signals to the converter is marked in color.
Note the proximity between traces red and green for the +24V (input) and +5V (output) to the converter. The proximity was considered as a possibility for an undesired I/O feedback.
The parasitic feedback was confirmed with a small knife, cutting the +5V trace and connecting it using a separated wire far from the +24V signal.
As you can see in Figure 5, the oscillation was removed, and the radiated test was passed.
Note other options as a small series ferrite with IC output could be tried as an alternative trying to reduce the undesired gain of our oscillator.
My final advice: Be careful with the layout of input and output signals to any IC or electronic circuit to avoid parasitic oscillations.
