Radiated Susceptibility from Structure Resonance in a PCB Upgrade

This case study describes an EMI troubleshooting investigation into a product that developed radiated susceptibility issues after retrofitting a new daughterboard IC. The unit, originally reliable, began to experience tone changes when exposed to radiated fields around 1.2–1.4 GHz following the upgrade. Detailed analysis revealed that structure resonance caused by the daughterboard’s layout and return paths was leading to strong radiated coupling at GHz frequencies, requiring specialised measurement and mitigation techniques.

Retrofitting a new IC to replace an older technology in an existing product often is not as easy as it sounds. Problems such as failing to function reliably or causing EMI issues could occur. Often such a fault is discovered after the original IC is no longer available in the market. 

In the following example, a company is caught by surprise when they found a new IC daughter board they designed and mounted on the original IC pinout location has suffered from radiated immunity issues. Since the product is a sound product, tone issues were found when the unit was exposed to radiated emissions around 1.2GHz to 1.4GHz. The original IC is no longer available after more than 10 years of service time. The new IC is based on ARM architecture and is powered by 3.3V. 

Identify the resonant structure 

Both the original IC and the small daughter board that replaces it are shown in Figure 1 (a) and (b), respectively. The backside of the daughter board shows that the pin legs are also bent 90 degrees on the connection points, which inevitably introduces inductance in the connection between the two boards.

The daughterboard (with circa 374 mm² area, 5 mm distance to the main board) forms a capacitance of approximately 1 pF (based on the simple equation C=ε₀ε_RA/h). The long trace from the daughter board to the main board can easily have a 10nH inductance. The self-resonance of this board is then estimated to be 1.5GHz, indicating the system is more likely to have immunity issues at this frequency range.

Measurement 

Among the techniques that could be used to measure the structure resonances, the method introduced here is perhaps the easiest and cheapest set-up. A network analyzer or a spectrum analyzer with a tracking generator can do the job. A directional coupler (such as a Mini-Circuits ZFDC-20-5+ or ZFDC-10-5+) is needed to work with a spectrum analyzer, whereas a network analyzer eliminates the need of a directional coupler (since it can determine return loss). 

The basic idea is that a small loop radiates RF energy depending on its loop area. Since the area in this case is small, it will only radiate efficiently above 3 GHz or so. Below this frequency, most of the RF energy bounces back due to reflection. The directional coupler directs some of the reflected energy, so one expects to see a flat curve on the tracking generator result. If the loop is close to a resonant structure, however, the RF energy could be absorbed by that resonance structure, causing a dip in the spectrum analyzer tracking generation plot results.

Figure 2 shows the test set-up (with spectrum analyzer, directional coupler, and a shielded magnetic loop). A shielded loop is supposed to perform better for this kind of test as it avoids capacitive coupling of the loop itself. But, in reality, both shielded and non-shielded loops prove to work equally well most of the time. 

As one probes the field probe around the daughter board, dips in the tracking generator results can be seen in the frequency range that causes immunity issues (see Figure 3). 

Alternatively, if a spectrum analyzer is unavailable, one can use a signal source that is capable of producing the signals in the frequency range of interest (in this case frequencies up to 2.1GHz) and a high bandwidth oscilloscope (at least 1GHz in this case), together with two square magnetic field loops. The set-up is shown in Figure 4, where one magnetic loop is used to generate noise across a wide frequency on the daughter board, and the other magnetic loop is placed alongside the 0V trace on the mainboard. When the RF noise hits the resonant frequency, one can see a sudden jump in amplitude of signals measured in the oscilloscope. The jump in amplitude indicates at resonant frequency, the ‘ground bounce’ between the two boards becomes more obvious. The magnetic field loop basically picks up the noise via mutual inductive coupling between the trace and one side of the loop. One caution in using this method is to be aware of the close field coupling between the two magnetic field probes. 

Problems, Challenges and Fix

One of the issues with the structure resonance in an immunity test is that once the RF energy is absorbed by the structure, RF current starts to flow on the traces of the PCBs. One particular issue is the ‘ground bounce’ caused by common-mode current flowing between the 0V of the daughter board and the main board. In fact, it was believed this was the main issue causing the tone change problem in this product, since the driver stage for the speaker is rather weak and there’s no ground plane on the main board design. 

The challenge here is that at such high frequency, measuring the “ground bounce” is extremely difficult, especially if high frequency precision measurement equipment is not available. Trying to measure this using a FET probe might not be good enough as 1pF capacitance (which often comes as the intrinsic capacitance of a FET probe) at 1.3GHz has an impedance of 122 ohm. Using a typical passive probe is a no go, because 10pF (often the intrinsic capacitance of a high impedance passive probe) at 1.3GHz is only 12 ohm. In fact, using a passive probe, even with the shortest lead, provides a perfect path between the two 0V of the boards under test. Therefore, the immunity issue “magically” disappeared, an experience that could be very frustrating for even experienced engineers.

On the product side, the constraints are: 

1. The main board was designed many years ago, it is only a 2-layer PCB without ground/power plane. The design cannot be changed for this revision.

2. The manufacturers have already manufactured tens of thousands of the daughter boards. Getting rid of them will be costly, also not in good spirit for the environment.

3. The options to fix the issue needs to be cost-effective, which is often the requirement for volume manufacturers. 

Mitigation methods that were introduced in previously referenced literature were applied on this design. The result and comments are presented in Table 1. None of the methods seemed to work perfectly well, and the methods listed in the table need intensive manual work in the assembly process. 

Table 1 Some options for fixing the problem 

An effective way, as it turned out, was to insert a flexible EMI absorber sheet in between the PCBs (as shown in Figure 5). 16 pins need to be punched in order to insert the daughter board (which some manufactures can do at an additional cost). One might wonder whether simply putting the sheet on top of the daughter board will do the job. The answer is no, as the resonance is really caused between the PCBs. Since there are many absorber sheets around, the selection should focus on the attenuation at the frequency range of interest and depth of these materials (often thicker sheets perform better). 

Another approach, which is quite similar to the method discussed above, is to use a connector EMI suppression plate. Fair-Rite happened to have a part with the same pin layout as the original IC (also suitable height). Therefore, it can be considered as an alternative method; two plates are needed to achieve the required attenuation: see Figure 5(c).