PXB Theory of Operation

The PXB signal processing chain starts with an input waveform. This can come from an external source or an internal source. It has some RMS value and some PAR (peak to average ratio), and the goal in using the PXB is to maximize the dynamic range of that signal as it flows through the system.

If the waveform originates from an external device such as an MXA, the RF input signal is down converted to baseband and digitized. The RMS value of the digital waveform coming out of the digital port is dependent upon the settings of the MXA. If the MXA is operating in the low end of its dynamic range, the digital waveform that comes out of it will have low dynamic range also. See Optimizing the MXA for detailed information on optimizing dynamic range in the MXA.

The digital signal from the MXA also has a native sample rate that differs from the 200 MHz internal rate of the PXB. So the first step in the PXB is to upsample each input source to 200 MHz. Additionally, MXA’s with the B25 option provide uncorrected IF data from the digital port, rather than corrected IQ data. The PXB recognizes these devices and applies IF to IQ conversion, as well as corrections.

The next step is to qualify the samples that will be fed to the input power meter. Many signals are bursted in nature, with duty cycles less than 50%. Normally we don’t want to count the “off time” when we’re measuring the RMS value of the waveform.The PXB has a number of ways to accomplish this, but the most obvious one is the threshold setting. This simply ignores any samples that don’t meet the threshold level. Those samples still flow through the system, but they are not measured by the input power meter. Note that though this power meter is used to help set the final output power correctly, it is operating entirely in the digital domain, measuring digital voltages.

We need to pause at this point and examine an alternate source of input, the digital baseband generator built into the PXB. It can generate either ARB or Realtime waveforms at virtually any sample rate up to 200 MHz, depending upon the software options purchased. These waveforms usually have built-in reporting of the RMS of the waveform, so the qualification step and power measurement are usually not necessary, unless the customer supplies his own waveforms. Internally generated waveforms also get upsampled to the common 200 MHz rate, so that they can be summed with other (different) waveforms, and to simplify the downstream signal processing. One difference from external inputs is that internally generated waveforms often have markers associated with them, to control external hardware like ALC or Burst modulator. These markers cannot be resampled along with the data samples, so they are tagged and flow along a parallel path, to be re-united with the original data at the end of the processing chain.

Both external and internal waveforms can be scaled, using the runtime scaling feature. This allows relative power adjustments between two or more waveforms. Once the final RMS of the input waveform is known, this value is passed forward to the fader. Here, the PXB examines the settings and computes a backoff adjustment to the RMS value. The long-term statistical effects of the fader are well understood, and the PXB can compute exactly how much the PAR will be increased. The RMS value is scaled downward accordingly, to prevent clipping while maximizing dynamic range.

This post-fading RMS is passed forward to the AWGN block, where the signal to noise ratio can be set. There is a single noise generator in each summing node, so the SNR of multiple signals entering the node are related. As noise power changes for one signal, it changes for all of them. At this point we are in still in time domain, summing voltage samples. The noise voltage can be higher or lower than the signal voltage, thus setting the SNR. What is important is the sum of all the voltages. This is what determines the total output power of the final RF signal. This sum is allowed to decrease after power calibration, but not increase. To change SNR while the PXB is running, you must calibrate with the highest total sum, then afterward change SNR so that the total sum decreases. Taken another way, after power calibration, you can use the AWGN block to decrease signal power, or decrease noise power, but you cannot increase either one.

Adding noise changes both the PAR and the RMS of the combined waveform. Any additional backoff required due to multiple signals being summed together is computed and applied. Normally the combined signal is now ready to be sent out over LVDS to the RF upconverter. In rare instances, the peaks of the signal can still cause overflow errors inside the RF upconverter, due to filter overshoot that occurs there. For this reason, a final runtime scaling setting is present on the PXB IO board, to scale the waveform down even further to avoid such errors.

The waveform and its associated markers flow across the LVDS cable to the RF upconverter. The PXB needs to initiate a power search in the upconverter, so that the final RF power is correct. But power search will not succeed if the RMS value of the waveform is less than 100 mV, which it often is after fading, summing, and noise addition. The PXB recognizes this condition when it occurs, and instead of sending the true RMS of the waveform for the power search, it sends in a value of 100 mV. Once power search completes, the PXB computes an amplitude offset to compensate for the power difference between 100 mV and the true RMS of the waveform, and applies this offset to the power in the RF upconverter.

The output power is a function of both LO power and waveform RMS power. Since the output LO power can be set independently of the input LO power there is no “insertion loss” when using the PXB as an RF to RF fader. Power search optimizes the combination of waveform power, LO power, and fixes the RF gain just below the RF gain compression point when the signal peaks occur. If you change the contribution of the RMS waveform power after power calibration, the output RF power will change proportionally. If the total RMS of the waveform goes down by as much as 30 dB, there is no problem as the RF gain stage can handle this. But if the total RMS of the waveform goes up (as it will do if you increase noise power or signal power), the RF gain amplifier goes into compression and clipping, and this is reported as an error.

You might think it would be easier to use the ALC (automatic level control) rather than power search on the RF upconverter. This seldom works with the PXB because the signals almost always include fading and/or summing. The ALC cancels out the amplitude variations caused by fading; it also cancels out any output power variations caused by changes in the input power (undesirable during power control testing). Even changing SNR with ALC on will cause both the signal power and the noise power to change, which is usually undesirable as well. So ALC off is usually the best choice.