Understanding and Correctly Predicting Critical Metrics for Wireless RF Links

Understanding and correctly predicting cellular, radar, or satellite RF link performance early in the design cycle has become a key element in product success. The requirements of today’s complex, high performance wireless devices are driving designers to assess critical measurements—noise figure (NF), 1dB gain compression (P1dB), third order intermodulation distortion versus output power (IM3dBc), and signal-to-noise ratio (SNR)—long before manufacturing begins. Traditional modeling methods such as rules of thumb and spreadsheet calculations (Friis equations) give limited insight on the full performance of an RF link in next-generation wireless products. This white paper highlights the advantages of using specialized RF system simulation software to accurately predict critical metrics for wireless RF links.

Figure 1: Traditional use of the spreadsheet as a system tool.

Simulation Software—A Novel Approach
Traditionally designers have used spreadsheets (Figure 1) to do calculations such as cascaded noise figure, P1db, compression point, and/or third order intercept point of an RF link. The advantages of using a spreadsheet are two-fold: data entry is simple and spreadsheet software is readily available. As wireless devices become more and more pervasive and complex, the limitations of spreadsheets become more apparent. In other words, spreadsheet responses are based on standard equations and therefore do not typically account for mismatch between components or noise at image frequency. In addition, spreadsheets do not normally support data files such as S2p, spur tables, etc., nor do they support yield analysis or optimization—techniques that are becoming increasingly important in order to produce high performance devices at a competitive price.

Figure 2: System simulation software tools start from a spreadsheet interface and automatically generate a system diagram.

A more modern approach is to use a tool such as AWR’s Visual System Simulator™ (VSS) software to determine system specifications (Figure 2). This tool is built specifically to exceed the capabilities of the traditional spreadsheet method and offer optimization features such as budget analysis and spur analysis. With this approach, designers can start from a spreadsheet interface to define the components (mixer, amplifier, etc.), whether file- or circuit-based, go on to define the circuit measurements such as cascaded noise figure and cascaded P1db, and also automatically generate a system diagram. This method provides far greater insight into what is happening. RF behavioral, circuit-based, and file-based models are available in VSS that account for voltage standing wave ratio (VSWR) effects and frequency dependence, as well as support yield analysis and optimization.

Figure 3: RF link and LO path of a mixer design. The spreadsheet and simulation software show significantly different cascaded noise figures (red boxes).

Example 1 – Accounting for PSD at the Input of a mixer
The first example in this white paper looks at a typical design that accounts for noise power spectral density (PSD) at the input of a mixer.

As Figure 3 shows, the RF link starts with the continuous wave (CW) source, followed by the amplifier, the filter, the attenuator, and finally the conversion. Looking at the low order (LO) path, there is a CW tone, an attenuator, a model representing a cable, the amplifier, and then it goes directly into the LO. When the analysis is run in the VSS software, a significant difference in the cascaded NF can be noted between the traditional spreadsheet, which shows 4.64dB, and the specialized software, which shows 11.45dB.

What’s happening to this link that is causing the discrepancy? The math was done correctly and entered into the spreadsheet, so the expectation is that the NF should be 4.6dB. The design was built and the analysis done, but the VSS simulation does not return anything close to 4.6dB. Why? The VSS tool is much more sophisticated than a spreadsheet and has lots of capabilities and measurements, so the next step would be to try doing a further analysis of the LO link.

Figure 4: The noise density analysis of the PSD at the beginning of the link is -138.6dBm/Hz due to thermal noise.

The analysis done here is to make the PSD measurement at the very beginning of the LO link. It can be seen in Figure 4 that the PSD at the input of the mixer’s LO is -138.6dBm/Hz. This is due to the fact that there is an amplifier prior to the LO input of the mixer. When there is gain and NF in the amplifier, what does that do to thermal noise? The NF goes up, as shown in Figure 4. Typical spreadsheet equations do not account for this, so while designers can follow the book and do everything right, the VSS software tool provides greater insight on the link.

Figure 5: Upon placing a filter after the amplifier, the noise density at the input of the mixer goes down to 174dBm/Hz and the software gives the expected measurement of 4.63dB.

What’s the solution to this problem? Place a filter after the amplifier and, as shown in Figure 5, the noise density at the input of the mixer goes down to 174dBm/Hz and the software gives the expected measurement of 4.63dB.

Figure 6: The behavioral filter is replaced with an actual circuit implementation.

Example 2 – Accounting for Reflections
The second example in this white paper shows how the behavioral filter can be replaced with an actual circuit implementation (Figure 6). The analysis is done and the S-parameters examined, specifically the S11 and S21.

Figure 7: The filter’s response changes with the inductance value, and the resulting noise figure of the link fluctuates. Changes in S11 and S21 result in cascaded measurement changes.

Going down to the circuit level, what happens if a particular component level is switched or if the inductance of one of the inductors in this filter is tuned? As the S11 response in the reflection is changed, once again the impact on NF can be seen (Figure 7). The filter’s response changes with inductance value, and the resulting NF of the link fluctuates as well. Changes in S11 and S21 result in changes in cascaded measurements.

Figure 8: The desired measurement of the third order intermodulation product is IM3dBc.

Two cases have now been demonstrated where the VSS simulation tool has differentiated from the typical spreadsheet. In the first case, the PSD running through the LO path causes the NF to go up due to the higher PSD in thermal, and in the second case, tuning or optimizing on the inductance value and changing the S11 changes the NF.

The final example (Figure 8) is a little more complicated. The goal here is to measure the ratio of a third order intermodulation product to the carrier IM3dBc.

Figure 9: Spreadsheet calculation for IM3dBc versus VSS measurement.

First, the typical spreadsheet process is used to define the components, and then the system is built in the software using a budget analysis tool such as VSS RFB™ (RF Budget Analysis) (Figure 9).

Once again, it can be seen that VSS does not agree with the standard spreadsheet measurement. The spreadsheet calculation is 39.577dBm and the VSS calculation is 30.845dBm (29.9dBm – 0.945dBm). Why do the two methods differ?

Figure 10: The third order intermodulation product is measured at the output of the second mixer and compared to the spreadsheet.

Let’s look more closely. The third order intermodulation product is measured at the output of the second mixer and compared to the spreadsheet (Figure 10). In the software, the intermodulation product is -92dBm, and the spreadsheet is calculating -97dBm. Only the ratio of that tone to the third order intermodulation product is being measured. To get to -97dBm, the previous value for the third order intermodulation is -101dBm, and -105dBm has been voltage combined to get to -97dBm. The software is calculating -92dBm—what is going on?

Figure 11: The RFI signal heritage window reveals a spur that falls on the IM3 product of -99dBm in addition to the expected -105dBm.

AWR’s VSS software offers a spur analysis tool called RFI™, short for RF Inspector, which enables users to understand the contributions to a particular spur and what causes the spur to be at that particular value—indicating there is something contributing to that third order intermodulation product to make it higher than expected, something that is folding into the third order intermodulation frequency (Figure 11).

Note there are no frequencies labeled, but there is a combination of Tones A, B, C in the index of one on the first LO, the combination of Tone A minus B minus C plus the LO is causing that value of -105dBm to be -99dBm. That is what the software calculated. Something folded over into that third order intermodulation product.

Figure 12: RFI reveals the output of the filter following the first mixer, Tones C, A, and B.

So, the voltage combined -95.5dBm + -101dBm equals -92dBm. What is coming into that mixer? At this point, RFI reveals that Tones C, A, and B are combined with the LO (Figure 12), and that value fell onto the third order modulation product. So how can the combination of Tones C, A and B in the LO be removed?

Figure 13: The VSS RFI spur analysis tool suggests placing a filter to eliminate unwanted Tones A and C.

The mathematics in the spreadsheet can’t tell us what to do, but VSS’s spur analysis does. It suggests that a filter be placed to eliminate unwanted tones (Figure 13).

A filter can be placed at that point to eliminate any of the Tones—A, B, or C. The LO can’t be eliminated, but once one of the tones is suppressed, the software gives exactly what the spreadsheet said: -101dBm from the previous and -106dBm from VSS. (Note: if designers are worried, they can dive into -106dBm to see what contributed to that to make a difference from the -105dBm.)

Figure 14: Spectrum plots showing the output after the additional filter is added, the output after the second mixer, and the final output of 30.47dBm minus -9.32, equaling 39.79dBm.

So once again the software was able to find the problem and offer a solution. A filter is added, two of the tones are reduced, the measurement is made, and the results are now reconciled (Figure 14).

Conclusion
Using modern software tools to determine system specifications provides much more value than using a spreadsheet as the only method. The software provides for a richer set of models, real word effects accounted for in the calculations, optimization and yield analysis, as well as in-depth spur analysis. Software tools such as VSS give designers much more accuracy and automation in determining system requirements than working exclusively with and maintaining legacy spreadsheets.

Tags: RF, Visual System Simulator, Wireless RF

Category: Visual System Simulator

This entry was posted on Thursday, February 14th, 2013 at 10:41 am. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.