RF Switching speed test

Introduction

Direct conversion receivers are popularin current market. It eliminates the IF filters with on-chip integrated lowpass filters for channel selectivity. Another advantage is high integration andlow BOM cost. In this receiver, a low-noise-amplifier (LNA) may be helpful toeffectively lower the overall cascaded noise figure of the whole system.

The RF2374, designed on a low-noise, GaAsHBT technology, is an excellent choice for WLAN/Wimax application. Thislow-cost, high third-order intercept point (IP3) LNA features a programmablebias, allowing the input IP3 and supply current to be optimized for specificapplications. The LNA provides up to +10dBm input IP3 while maintaining alow-noise figure of 1.6dB and a typical gain of 11dB at Wimax 3.6GHz. Moreover,the special bypass function saves current consumption and brings higher IP3 inthe case of good signaling state. Figure 1 shows the block diagram of RF2374.

Figure 1.Functional block diagram.

Since WLAN/Wimax is a TDD system, theLNA must have fast turn-on and turn-off times in order to fulfill systemrequirement. Typically, 10µs is an acceptable limit for 802.11b.Tougherspecification for Wimax system is due to higher data rate. Some Wimax solutioneven request less than 1us switching time.

Lab measurement indicates that theTurn-on time for the RF2374 is approximately 356.7ns (power ramping up fromless than -40dBm to ­-19.5dBm ±1dB). The Turn-off time is 261.7ns for 20dB dropfrom nominal output power. Both timings meet the system requirement of 10µs for802.11b, and meet Wimax specification with good margin.

Table 1. The RF2374 Switching Time Performance

TestProcedure

A standard RF2374 evaluation kit was used as shown in Figure2.

Figure 2. Standard EVB.

The 10pFcapacitor will degrade the IP3 by 3 or 4dB, while we can increase the Icc by 2or 3mA for compensation, and keep the NF almost the same.

See Figure3 for the test setup that was used for all measurements.

Figure 3. The RF2374 switching time measurement test setup.

The function generator AFG320 was set to deliver aTTL-compatible logic pulse at 3.0VPEAK in burst mode. The pulse-width was 62.5µs, asshown in Figure 4.

Figure 4. Trigger pulse applied to Vref in RF2374 EVB andspectrum analyzer.

The trigger was routed to both the Vref in the EVB, and theexternal trigger input on the spectrum analyzer. This arrangement allowed fordirect correlation between the on/off trigger pulse and measurement data startpoint. There was one thing needing attention. The time delay inside of thespectrum analyzer should be taken into account in this case. The method toremove this effect was to connect the spectrum analyzer RF input and signalsource output directly, and do time alignments.

There were two steps.The first step was to test the signal generator delay in pulse mode at lowfrequency, and the second one was to test the spectrum analyzer delay at 3.6GHz.

The calibration herewas based on the assumptions that signal generator had the same delay atdifferent frequency operation.

EVENT1 from thesignal generator indicated the pulse beginning point. Both EVENT1 and signalgenerator output were routed to channel 1 and channel 2 in an oscilloscope respectively(shown in figure 5).

Figure 5.the pulse signal(in yellow) with trigger(in red) from signal generator.

The 256nS would be used in spectrumanalyzer time alignment.

After knowing thesignal generator delay, the trigger from EVENT1 in signal generator would berouted to spectrum analyzer TRG IN, and signal output would be connected tospectrum analyzer input directly. ­­­Operation frequency was 3.6GHz.

There are threecontributions to the delay in spectrum analyzer.

1. The bandwidtheffect of the RBW filter. 2. The latency of the RBW filter. 3. The delay of the trigger distribution path.

All this effect should becompensated in the test system. Figure 6 showed the spectrum analyzer delay.

Figure 6.the delay of the spectrum analyzer

The spectrumanalyzer was configured to display power (dBm in 8MHz resolution bandwidth) vs.time, and the time scale was referenced from the trigger pulse with proper trigdelay which has been tested. This method produced the results shown in Figures7 and 8.

Figure 7. POUT vs. time (5dB/div, TRG: rising edge).

Figure8. POUT vs. time (5dB/div, TRG: falling edge).

Under low gainmode, the Vref has no effect on the RF performance. The gain of RF2374 will be around-3dB no matter Vref is on or off.

ps: 本文写于2008年上半年。参考了www.maxim-ic.com/an1191