In the recent blog post, Radio420: Local oscillator reconfiguration latency when using SPI, we described the advantages generated by dynamic reconfiguration of RF local oscillators (LO). Dynamic reconfiguration can be used for frequency hopping or cognitive radio applications. However, the time required for reconfiguration, which causes link downtime, can vary.

On Nutaq’s Radio420X FPGA mezzanine card (FMC), the LO for the RF front-end (RFFE) can be modified with SPI write transactions triggered directly from the FPGA logic. This enables fast and dynamic reconfiguration of the RFFE, as the reconfiguration can be triggered as soon as the recently digitized data meets a certain criteria and because SPI transactions do not need to be handled by a microprocessor.

The SPI transaction time is not the only parameter that affects reconfiguration time. Once the SPI transactions have been completed, the RFFE LO has a certain amount of settling time.

The datasheet for the LMS6002D, the radio transceiver chip used on the Radio420X FMC, states that the oscillator settling time to 1 ppm is typically 20 µs. In this blog post, a few measurements are done on the Radio420X to see the behavior of the local oscillator following its reconfiguration.

TX local oscillator reconfiguration

In order to test the settling time of the LO in the transmission path, I started with the orthogonal frequency-division multiplexing (OFDM) reference design included in Nutaq’s ADP 6.5 software tools (see OFDM reference design: Moving into cognitive mode). The algorithm that generated the SPI transactions has been kept unchanged except for the part that changes the frequency when the channel seems occupied. Now, the LO is reconfigured when a given custom register has its value changed. The LO is reconfigured from a 433 MHz to 453 MHz frequency at the custom register rising edge. For these tests, the SPI clock is configured to 10 MHz.

The SPI start signal has been routed to a generic purpose input/output (GPIO) of the system. This makes it possible to monitor the LO reconfiguration state with an oscilloscope and have a time reference point for the analog signals.

An oscilloscope is used to simultaneously record the transmitted RF signal and the SPI start state. Figure 1 shows the oscilloscope acquisition triggered at the rising edge of the first SPI start pulse.

Figure 1: Oscilloscope acquisition of TX LO reconfiguration from 433 MHz to 453 MHz

Figure 1: Oscilloscope acquisition of TX LO reconfiguration from 433 MHz to 453 MHz

The TX signal of the Radio420X (blue) is generated from a baseband 1 MHz direct digital synthesis (DDS) signal modulated by the TX LO at 933 MHz. The twelve red SPI start pulses come from the six TX and the six RX registers that need to be written (see Radio420: Local oscillator reconfiguration latency when using SPI). The first marker shows that the TX SPI transactions start at 150.3 µs and the second marker shows the time at which the TX transactions are completed at 170.4 µs.

In the time domain, it’s hard to see the TX signal changing from a 934 MHz to a 954 MHz sinusoidal signal. Displaying it on a spectrogram plot makes the analysis easier.

Figure 2: Spectrogram of TX LO reconfiguration from 433 MHz to 453 MHz

Figure 2: Spectrogram of TX LO reconfiguration from 433 MHz to 453 MHz

The spectrogram in Figure 2 shows the spectral domain of the signal over time, using consecutive fast Fourier transforms (FFT) of the time domain signal. The markers show the beginning and end of the TX SPI transactions as well as the moment the TX frequency becomes stable.

The first SPI register (0x15) changes the frequency range of the LO, the next four (0x10, 0x11, 0x12 and 0x13) change the LO divider value, and the sixth register changes the oscillator capacitor. During the SPI transactions, the impact of the modification of the LO divider can be seen and, after the capacitor modification, a rapid modification can be seen followed by a settling time.

It takes approximately 48 µs to become stable after the beginning of the SPI transaction. This time could certainly be reduced if the SPI transactions are performed faster (for example, by increasing the SPI clock frequency). Furthermore, in this case, the initial and target frequencies are relatively close to each other, so the first SPI register (register 0x15) does not need to be updated (its value is unchanged).

Figures 3 and 4 show the reconfiguration of the TX LO from 400 MHz to 780 MHz. In this case, register 0x15 needs to be reconfigured since the target frequency is not in the same frequency range as the initial frequency.

Figure 3: Oscilloscope acquisition of TX LO reconfiguration from 400 MHz to 780 MHz

Figure 3: Oscilloscope acquisition of TX LO reconfiguration from 400 MHz to 780 MHz

Figure 4: Spectrogram of TX LO reconfiguration from 400 MHz to 780 MHz

Figure 4: Spectrogram of TX LO reconfiguration from 400 MHz to 780 MHz

This time, a major frequency changed happened after the frequency range modification. It took approximately 60 µs after the first SPI transaction before the TX LO frequency became stable.

Conclusion

The LMS6002D datasheet states that the settling time of the LO is typically 20 µs. After experimentation, the TX LO settling time varies depending on the initial and target frequencies. Also, the SPI transaction time need to be taken into account. Increasing the SPI transaction speed could reduce the overall reconfiguration time and, depending on the system requirement, all six register transactions may not be required.

In the next part of this blog series, the LO settling time for receiving path of the Radio420X will be analyzed.