IQ Noise Correction (IQ-NC)

Menu Path: Input > IQ Noise Correction

IQ-NC is currently under BETA support. This means that the programming API 1) Access Preamble Indicator, or 2) Application Programming Interface or measurement results may be improved in a future release.

IQ Noise Correction (89601EVMC), or IQ-NC, is a feature that can be used to compensate the acquired signal such that the measured EVM Error vector magnitude (EVM): A quality metric in digital communication systems. See the EVM metric in the Error Summary Table topic in each demodulator for more information on how EVM is calculated for that modulation format. will not include the contribution from the analysis hardware. This is done by acquiring multiple repetitions of the same signal and using averaging to separate the noise from the signal.

Correction Mode - selects between three modes of IQ noise correction to accommodate various measurement needs.

No Correction - the IQ-NC measurement is still performed so that the Repetition Info trace is valid, but the data passed to the demod is the uncorrected data from the first repetition.

Analyzer Noise - the signal is coherently averaged to suppress uncorrelated noise, then the uncorrelated noise vector is scaled such that the power represents only the noise from the DUT Device under Test: An acronym used to describe some type of electrical apparatus connected to test instrumentation. The apparatus can range from a single component to a complex subsystem such as a mobile phone, base station or MSC., and then recombined with the signal. This way, when overall EVM is computed, it will represent the DUT EVM without the contribution from the analyzer's uncorrelated noise.

Analyzer + DUT Noise - the signal is coherently averaged to suppress any noise that is uncorrelated with the signal. This way, when EVM is computed, it will tend to represent the distortion performance of the DUT (assuming the analyzer distortion is not a significant contributor).

Remove kTB Noise (only available when Analyzer Noise is the selected Correction Mode) - when unchecked (default), no compensation is applied to remove kTB noise and the user must specify the assumed temperature to use when computing the kTB noise. When checked, kTB noise compensation is applied and the user does not need to specify the Temperature (Kelvin).

There are three noise sources being considered for IQNC. One is the excess noise of the DUT above kTB (can be computed from the Noise Figure of the DUT), one is the thermal noise (kTB), and the other is the excess noise of the analyzer (from its Noise Figure). For now, we can call the combination of these noise sources, Total Noise.

As the power of the DUT signal is decreased, the signal drops closer and closer to the Total Noise, which is a reduction in SNR Signal-to-Noise Ratio and translates directly to an increase of EVM (EVM=1/SNR). But, since the analyzer noise is included in Total Noise, the EVM shown is higher than the true EVM of the DUT.

When the Remove kTB Noise parameter is unchecked (default), the input to the demod will be compensated for only the excess analyzer noise and leave the kTB noise. If you compare a regular EVM vs an IQNC EVM bathtub plot, the left-hand side of an EVM bathtub plot, where it slopes upwards at 1 dB EVM per 1 dB decrease in DUT power, will shift left, removing the analyzer's excess noise contribution to the EVM measurement.

When the Remove kTB Noise parameter is checked, the input to the demod will be compensated for the excess analyzer noise AND kTB. This will remove the slope upwards on the left of the bathtub curve. You can think of the resulting EVM as the EVM of the DUT as if kTB noise were not present at the DUT connector (which it always is). One thing to consider, however, is that as the kTB noise begins to dominate, the numerical accuracy required to perform the IQNC compensation tends towards infinity (infinite accuracy in measuring the actual noise present), so at some point, as you move left in the bathtub plot, the IQNC measurement will fail and a DATA? trace LED will appear on the traces indicating that the measurement cannot be performed.

Temperature (Kelvin) (only available when Remove kTB Noise is unchecked) - the user must specify the assumed temperature to use in the computation of kTB noise.

Default: 290 deg (corresponds to the -174 dBm deciBels referenced to a milliWatt: dB relative to 1 milliwatt dissipated in the nominal input impedance of the analyzer/Hz value typically used)

Enabled - enables IQ Noise Correction. If the Enabled checkbox is grayed out, hover over the "(Not Supported)" message to see a tool tip with possible reasons. See IQ-NC Tips and Assumptions below for more information.

Auto-configure from Measurement - When checked (default), the following parameters are auto-configured by the measurement:

Signal Mode
(when Signal Mode = Continuous) Repetition Period and the corresponding Auto-detect parameter

Note: Changing any of the auto-configured parameters will cause the Auto-configure from Measurement check box to be disabled.

Signal Mode - indicates the type of signal being corrected and influences the search for repetitions within the acquisition.

Bursted - (default) signal consists of bursts, with idle time between. The amount of idle time is allowed to vary when Auto-detect is checked. Otherwise it is expected to be constantly repeating with the specified Repetition Period.

Continuous - signal is transmitted continuously without gaps. If the repetition period is known, Auto-detect may be unchecked and the period specified. Otherwise, Auto-detect should be checked.

Average Count - number of repetitions of the signal to acquire and use during IQ Noise Correction. Increased Average Count will improve the accuracy of the resulting EVM at the expense of speed and memory. You can increase the count and see if the EVM reduces significantly. At some point, there will be diminishing returns and the EVM should stabilize.

Default: 10

Signal Repetitions

Auto-detect - when checked (default), the exact period of the signal will be autodetected. This will not be as fast as when Auto-detect is unchecked and the exact period is specified in Repetition Period.

Time Tracking (only available when Auto-detect is unchecked) - when checked (default), Time Tracking searches for the repetitions and does correlation to perform fine time alignment before coherently averaging. When unchecked, the Repetition Period is assumed to be the exact repetition period and no fine time alignment is performed.

Phase tracking between repetitions is always performed.

Repetition Period - (Signal Mode = Continuous, or Bursted with Auto detect disabled) specifies the expected time between repetitions of the signal.

Default: 1 μs (32 samples)

Max Repetition Period - (Signal Mode = Bursted with Auto-detect enabled) specifies the maximum time between repetitions of the burst.

Default: 1 μs (32 samples)

The measurement will be faster when Auto-detect is disabled and the exact period is specified.

The second text box for (Max) Repetition Period provides an alternate way to set the length of a repetition in terms of samples. These samples samples are at the sample rate of Raw Main Time, which is usually the same as the Measurement Sample Rate (typically 1.28 * Span). You can find the Raw Main Time sample rate by putting a delta marker between two samples to measure the sample period and then inverting to get sample rate.

IQ-NC Tips and Assumptions

There are several assumptions that must be followed to ensure correct measurement results:

the signal is gap-free in both time and frequency. While it is fine that a signal is bursted, within the burst there should not be any "off" portions.
the signal is the same for each repetition
the noise profile is measured accurately at the current DUT power level. Some analysis hardware can internally terminate the input, but others can only attenuate. When enough attenuation is not present to reduce the signal sufficiently, the analyzer noise will be estimated as larger than it actually is and the EVM value may be over-compensated (lower than it should be). To check whether this may be happening, check the Noise Correction Spectrum trace for any signal leakage. If present, you can reduce the DUT's transmit level.
the signal's occupied bandwidth fills the VSA's Measurement Span (the DUT+analyzer noise is computed over the Measurement Span)

When the noise ratio (analyzer / (DUT+analyzer)) approaches 1.0, the variance of the IQ-NC measurement increases. This is typically the case for low DUT powers. In these cases, the measurement may start to fluctuate more or the measurement may detect this case and stop providing results. This is due to the need for increasingly better accuracy of the noise profile measurement as the ratio approaches 1.0.

IQ Noise Correction is only able to compensate the analysis hardware's uncorrelated noise component. Contribution of noise from other sources (correlated with the data, such as distortion) cannot be reduced using this method.

While it is possible to look at results such as IQ constellations and Error Vector Spectrum traces, these traces will still show the shape of the analyzer noise contribution, but scaled. The scaling adjusts the overall EVM metrics to reduce the contribution from the uncorrelated noise power of the analyzer.

To improve measurement speed, follow these recommendations:

For bursted signals, reduce the idle time between repetitions
Configure the Repetition Period and de-select Auto-detect. This will reduce the amount of data to acquire and also reduce the searching required to find the repetitions.

IQ Noise Correction can be used with any repeating signal and any measurement. However certain measurements provide some level of auto configuration of the IQ Noise Correction parameters. This auto configuration will take effect when the Auto-configure from Measurement parameter is enabled.

If IQ Noise Correction cannot be performed, the Search Time (or Main Time) trace will show an IQNC? trace LED. If you click on this LED, you can see the reason why IQ Noise Corrections cannot be performed. The typical case is that not enough repetitions could be found.

To use IQ Noise Corrections, first configure your measurement to successfully measure EVM. Then make sure the signal is repeating and configure the Average Count you wish to use. If the signal is a bursted signal, configure the Repetition Period parameter. Then you can click the Enabled check box to perform IQ Noise Corrections.

Cases where IQ Noise Corrections is not supported:

Logical I+jQ input channel operation
Not licensed (requires 89601EVMC license)
Power Spectrum measurement (SEM/ACP Adjacent Channel Power: The power from a modulated communications channel that leaks into an adjacent channel. This leakage is usually specified as a ratio to the power in the main channel, but is sometimes an absolute power./Power Spectrum)
Segmented Capture enabled
Noise profile is not available from the current hardware (or recording)
Other reasons supplied by the current measurement

To see whether your analysis hardware supports IQ Noise Corrections, see your hardware's About topic (under Measurement Platforms > Available Measurement Hardware) and look for the IQ Noise Correction section. If present, it will indicate which hardware models are supported.

When Signal Mode = Bursted, the output of IQ Noise Corrections (typically in a Search Time trace) will contain the corrected burst, but the points outside the burst will be set to a constant small value. This is due to several reasons, but basically to avoid compensating IQ samples that do not contain a signal.