Meanwhile, Wireless industry has developed ad hoc specifications for PIM mostly based on field tests and trial and error. A maximum PIM level of -110 dBm in the receive band is found to be desirable in vast many cases, especially when BTS interfaces are involved. Many component vendors aim for -115 dBm or better depending on the type of component. Given the standard test level of 43 dBm per tone (CW), this -110 dBm absolute power level corresponds to a -153 dBc of relative PIM level. Note that it may be hard to relate this analytically to an ideal state of the art base station receive sensitivity of around -120 dBm range since these are defined for specific bandwidths and for CDMA/OFDM type signals that resemble anything but CW.

Fig. 1. IM3 vs power for two different devices compared with ideal at 1900 MHz. These are two different passive devices with rather high PIM. None follow the 3:1 slope expected of IM3 products. Typical PIM IM3 rises at a rate 1.5-2.5:1. Noise floor is near -140 dBm.

One of the drawbacks of fixed power testing per IEC is it its unsuitability to lower power applications. Designing and manufacturing a passive component or assembly for a 2 x 20 W test when let us say the application is only 2 W, results in a costly and heavier solution than required. In the absence of a standard and accepted guideline, some users are inclined to stick with the 2 x 20 W test requirement which results in significant cost, size and weight increase for the product. Even when the test requirement is pulled back to 2 x 2 W for example, which is somewhat of a low power test standard, then the false expectation is for the PIM performance to be now 30 dB better, around -140 dBm, based on traditional IP3 analysis as shown in Fig. 1. This level is about -173 dBc from the test tone, beyond the reliable measurement limit of most of the state of the art PIM test instruments. It gets even worse if the PIM of specific interest is 5th or 7th order and one tries to apply text book power laws, 5:1 and 7:1. Then the requirement exceeds what is physically measurable. In Fig. 2 reducing power down to 1W per tone for example, the actual IM5 measured may be about 20 dB higher than what a simple 5:1 extrapolation would predict. It is possible to lower PIM for a component to such low levels by careful design, numerous iterations and very well controlled manufacturing process, but very unlikely to obtain reasonable yield for commercial applications, in addition to being unnecessary for proper system operation. So there is an obvious need to re-think the PIM specifications as the base stations get smaller and power outputs get lower.

Some of the confusion can directly be traced to the convention of specifying PIM as a relative measure in dBc rather than an absolute power level in dBm. This has become a de facto standard method of PIM measure. In reality and practice, a radio receiver responds to power not dBc. Sensitivity of the radio is not affected by PIM or 

Fig. 2. IM5 for PIM plotted as a function test tone power for two different DUTs at 700 MHz. IM5s do not follow 5:1 slope either. IM3 for one of the devices also included for reference.

other low-level interference if the absolute levels are much below the sensitivity of the receiver. Since power levels involved may be substantially lower than 40 W for many new wireless infrastructure applications, one could then in theory adjust the required dBc value based on 2 x 20 W measurement and establish a new requirement in dBc appropriate for the actual power levels.

Unfortunately this would be straightforward only if PIM behaved according to a known power law. It does not, as shown in Figs. 1 and 2. So while higher power level components and systems are tested with a standard level of 20 W tones, which has proven to be a good figure of merit, lower power system specs still remain open to interpretation and requires better knowledge of specific operating conditions, receiver sensitivity requirements and PIM test limits. Otherwise, trying to extrapolate from 2 x 20 W tests to lower power levels, without considering the actual absolute power levels involved and required, will result in unrealistic component and subsystem PIM requirements which will drive the costs up and yields down. 

Specifically for example, for a low power system or assembly that cannot be characterized for PIM by using 2 x 20 W tones for physical reasons, then one may need to find out what exactly are the corresponding specifications for lower power, say 2 x 2 W tones. 

As explained, before, one cannot simply extrapolate -153 dBc to -173 dBc based on a theoretical 3rd order nonlinearity behavior. Third order IMD generated by 2 x 20 W when PIM is -153 dBc down, turns out to be -110 dBm level. In reality this is the accepted PIM performance for standard 2 x 20 W test. Question is, what is the dBc requirement for low power levels that would generate -110 dBm distortion levels. For a 2 W test this turns out to be -143 dBc. 

If we expect on the other hand an improvement in IMD of 30 dB, corresponding to the 10 dB drop in tone power as one would expect in standard 3rd order behavior, than we would be looking for -140 dBm (-173 dBc) PIM, which happens to be the noise floor of most of the respectable PIM testers available today. In fact, absolute PIM measurements much less than -130 dBm or so tend to be highly unrepeatable in general. 

In addition, PIM very rarely follows 3:1 slope vs. power for 3rd order components as shown before. Considering low end of the typical slope of 1.5-2.5:1, one should then expect a PIM level of -125 dBm approximately when test power is dropped to 2 W from 20 W. For a very “high” slope of 1:2.5, the PIM level would drop to -135 dBm. This is a far cry from -140 dBm expected from basic theory. It seems for a 2 W test, -117 dBm (-150 dBc) seems to be a very reasonable benchmark and where -110 dBm PIM is found to be transparent to system operation, then surely -117 dBm, a 7 dB improvement, should not be a cause for any concern. In practice, there will be many situations where the small cell or remote used may have power levels higher than 2+2 W, or there may be power combining. Even considering the worst case scenario of IMD increase with power, -150 dBc guarantees low PIM up to 8 W of total power. 

One word of caution is that though devices and subsystems may be designed for low-power performance, when in real applications high power macros and small cells are co-located and used together, there is always a risk of low-power elements being connected to high power levels by accident. Given the complexity of managing numerous bands from numerous carriers from their numerous base stations, this is not an unlikely scenario. It would be a good design practice to have such low-power systems at least be able to tolerate high power levels of 30-60 W even with poor PIM performance but without damage. 

Table: Maximum PIM values expected in dBm for various PIM test power levels and dBc specs

Conclusion
We have shown, by analysis and mostly by measured data that PIM specs traditionally used to qualify low-PIM components and assemblies need to be reconsidered and modified to achieve at least the same level of PIM performance as the standard 2 x 20 W test guarantees, namely -110 dBm. In fact, a 2 x 2 W test for a low power component and system with -143 dBc level PIM performance provides the same low PIM measure as the 2 x 20 W test for a higher power system. In view of this a -150 dBc of PIM for a 2 x 2 W test may be adopted for low power components, which provides a 7 dB improvement over the -110 dBm accepted level. Such a spec also guarantees low PIM operation up to 8 W of total power.