|
Bolt It Down Tight
| by William D. Kimmel, P.E. |
| and Daryl D. Gerke, P.E. |
|
If you work the military EMC projects, you are no doubt familiar with the 2.5 milliohm resistance requirement commonly specified for bonding impedance. We don’t know how the actual number was arrived at, but there is no doubt that it is an important parameter: we have learned from experience that you can save lots of time by testing resistance with a milliohm meter before commencing your EMI tests. Basically, if your resistances are not very low, you have a screw loose somewhere, and you can waste lots of time running it town.
Accordingly, we have adopted a policy of "bolt it down tight" and checking it to make sure. If you have everything bolted down tight, you are likely to have to go back and find which bolt is still loose.
Filters and Ground Impedance
The issue of ground impedance has broader implications, one of which we will explore in this article, that of filter effectiveness. There are several interesting situations that arise as a result of high contact impedance. The phenomenon has been documented in power line filters, but the situation shows up elsewhere as well.
A basic problem with schematics is that they only identify the ideal component along with an assumed perfect interconnect. But, especially with EMI, we need to incorporate those parasitic elements into the design or it won’t work as intended. A little more sophisticated model will include series inductance in capacitors and shunt capacitance for inductors, both of which limit the high frequency performance of filters. Even inductance in ground paths is often considered. But the one that is not always recognized is ground bond impedance.
Figure 1 shows the basic phenomena. The ideal filter is shown with some series impedance in the common ground path, shown as Zg. As you can see, if Zg is large enough, the shunt path to ground is blocked, and the capacitors shunt the current around the series inductor. In this situation, the attenuation is driven largely by the impedance ratio of the first capacitor and ground – the series inductor and the second capacitor, C2, having second order impact.
So how high does the ground impedance have to be before this problem becomes significant? The answer is, surprisingly low. Let’s look at a couple of common cases, the power supply filter and the filter connector.
Power Supply Filters
When trouble-shooting power supply emission problems, we repeatedly find on-board power supply filtering composed of an array of series ferrites and common mode chokes interspersed with shunt capacitors. The problem is, the ground impedance limits the performance of the filter. Basically, the filter performance is determined by the ratio of the input capacitor to the ground impedance, no matter how many additional filter stages are employed. We have solved more power supply emission problems by removing capacitors than by inserting extra capacitors.
Filter Connectors
The problem is amplified in the filter connector. Let’s do a comparison of three common filters under various conditions. The simplest filter connector is the C filter, which runs a capacitor from pin to ground usually with a capacitor at each pin.
Figure 2 shows a simple C, L and Pi filter with ground impedance inserted. Let’s assess the effect of ground impedance with each filter type. Assume a 50 ohm source impedance and a high impedance load. Further, assume a 50 ohm impedance for the series element and a 0.5 ohm impedance for the shunt capacitors – both would be typical values for 100MHz problems.
If we start by assuming the bond impedance is zero, we see that the C filter attenuation is primarily driven by the ratio of the 50 series resistance to the shunt capacitance, which gives us 100x or 40dB. The L filter has an additional series element, so the attenuation is 200x or 46dB, not much improvement in this example. The Pi filter gives us much better performance, about 80dB.
Now let’s insert a 5 milliohm ground impedance. This impedance is not enough to significantly alter the performance of any of the three filters, but the Pi filter performance is ultimately limited by the ground impedance. When the ground impedance becomes significant, the attenuation becomes primarily the ratio of the input impedance to the ground impedance, a condition that will be discussed next.
Now let’s consider what really happens in a connector – you may have a number of filter capacitors, each dumping its current into the ground impedance. The good news with a multi-pin connector is that the effect of filter capacitors tends to be additive. The bad news is that all of the currents shunted through the filter capacitor get funneled through the resistive ground. In the worst case, with the cable currents being fully common mode, the current is multiplied by n, the number of pins, and the result is that the impedance in the ground connection is multiplied by the number of filter pins. Thus, if we have a 100 pin filter connector, the 5 milliohm ground impedance gets multiplied by 100, or 0.5 ohm. Now, this impedance becomes a significant factor. The C filter attenuation now falls from 40dB to 34dB, the L filter attenuation falls from 46dB to 40dB and the Pi filter attenuation also falls from 80dB to about 40dB, as well. So the Pi filter performance really degrades when the ground impedance is factored in.
Finally, let’s depart from the specified 50 ohm input impedance, and look at the real world. In any practical case, the impedance of a wire will swing from a maximum of about 100 ohms down to less than an ohm at resonance conditions – for simplicity, let’s assume 0.5 ohm. At resonance, the C filter attenuation is negligible, the Pi filter attenuation falls to about 6dB, and the L filter attenuation is still about 34dB. Thus, in the real world application, the L filter works much better than the Pi filter, by the simple expedient of removing one capacitor.
So, you see that low ground impedance is vital to the performance of a Pi filter or C filter. That is why we need to have a 0 ohm impedance path for the connector ground.
The Alternate
So what can be done to improve the performance of a Pi filter? Simply put, you need to remove the common impedance between C1 and C2. Basically, this means you ground the two capacitors with their own grounding path. In military electronics, it is common to insert part of the filter in a "dog house" immediately behind the connector. Note that you have a C filter in the connector, a series filter element in the dog house, and a C filter array terminated to ground.
Well, we don’t have a high impedance in this path, but we do have another factor, which we call the impedance multiplier. Figure 1 assumes the ground is used for one filter, but what about a 100 pin connector?
Why a Pi Filter?
Frankly, we are a bit puzzled by this. The filter connector houses test filters with 50 ohms in and 50 ohms out, and the performance of a Pi filter is impressive under those test conditions. It is easier to get a high attenuation with shunt capacitors than it is with series inductors. But how to they perform in the real world?
Once we leave the protection of a 50 ohm resistive load, the Pi filter doesn’t fare so well. The impedance of the input may fall below an ohm at resonance, and the impedance of the load is almost never 50 ohms.
Summary
Ground bond impedance degrades filter performance, in power supplies and especially in filter connectors, where even milliohm resistances can be a problem. This condition is especially prevalent in Pi filters. If you select a Pi filter, you better make sure that the ground impedance is as close to zero as you can get. Otherwise, you may get a nasty surprise when you get on the test floor.
So do yourself a favor and bolt it down tight.
|
