Multichannel scopes enable MIMO RF testing

Multiple-input multiple-output offers the potential to increase data rates for a single user by using two or four streams of data transmitted with multiple antennas.

For example, peak data rates of up to 172.8 Mbps for a 64 QAM 2x2 downlink FDD MIMO and 326.4 Mbps for a 64 QAM 4x4 downlink FDD have been discussed for LTE.

However, these potential peak data rates can be impacted by the increased complexity of implementing two- or four-channel MIMO, relative to a single-input single output single antenna implementation.

Hardware design and implementation impairments such as antenna crosstalk and timing errors may reduce performance gains to be achieved by moving to multiple antenna techniques.

In addition, the complexities of implementing multiple antenna techniques can make it difficult to troubleshoot and debug hardware performance issues. Increasing the number of antennas and data streams from two to four for 4x4 MIMO further compounds the level of complexity in debugging issues.

This article will show the impact of antenna crosstalk impairments, phase noise and timing error on MIMO downlink system performance. Troubleshooting techniques with time-coherent multichannel oscilloscopes and 89600 vector signal analyser software will be discussed to help engineers gain an insight into error mechanisms affecting their hardware error vector magnitude performance and system-level RF transmitter performance budgets.

LTE will be used as the case study in this article, although the concepts presented can apply to other signal formats such as Mobile WiMAX.

LTE MIMO interleaves a known signal, known as a reference signal, throughout the frame in both frequency and time. It is fundamental in the process of recovering MIMO signals because it allows each receiving antenna to establish a reference of the signals from each transmitter.

Figure 1 shows how the individual symbols of the reference signal are allocated to subcarriers for a two-antenna downlink signal.

Error vector magnitude is a key system metric for RF transmitter performance. Comparing RS EVM to composite EVM can help engineers gain insight into transmitter hardware design impairments. It can also help to diagnose specific impairments such as antenna crosstalk, amplifier gain compression distortion, phase noise and other error mechanisms.The subcarrier allocation of the reference signals is illustrated as the y-axis (every six subcarriers), while the time interleaving is illustrated as the x-axis. Note that the reference signal alternates between antenna 0 and antenna 1, both in terms of the occupied subcarriers and time (symbols).

This case study will illustrate how RS EVM and composite EVM can be used to gain insight into the types of impairments that may impact the system performance error budget. The case study will also highlight the impact of transmit antenna timing error on reference signal orthogonality and will show how it should be considered when interpreting antenna crosstalk, constellation and EVM measurement results.

The four-channel MIMO test set-up used in the case study is shown on the left side of Figure 2, which consists of four Agilent signal generators with arbitrary waveform generators and the company’s four-channel Infiniium 90000A series oscilloscope.

A baseline four-channel MIMO measurement result using the vector signal analyser software with the multichannel wideband oscilloscope is shown on the right of Figure 2. The 16 QAM physical downlink shared channel constellations for two out of the four layers of the spatial multiplexed data are shown on the left (layers 2 and 3 are not shown here).Multichannel oscilloscopes are suitable for two- or four-channel MIMO measurements because they offer time-coherent multichannel inputs, wide bandwidths to measure RF modulated carriers and deeper memory to analyse multiple frames of data for demodulation with 89600 vector signal analysis software.

The RF spectrum is shown on the upper right of the VSA display and the error summary table is shown on the lower right of the VSA display.

Note that the residual composite EVM (lower right of VSA display) for the baseline test case is less than 0.8%, reflecting the clean constellation states shown for layers 0 and 1 (left side of VSA display).

A multichannel oscilloscope and the VSA software would typically be used on a two- or four-channel IF-RF transmitter/upconverter hardware DUT for MIMO testing. Since a DUT was not available for testing, the company’s SystemVue simulator was used to model a four-channel RF transmitter with simulated design impairments.

Each transmitter consists of IF/RF bandpass filters, mixer with LO and a power amplifier. LO phase noise is specified at a 10 kHz frequency offset and a 1 dB gain compression point is specified on the PAs.

A custom model subnetwork is used at the output of the transmitters to model antenna crosstalk. ESG sinks are then used to download the simulated IQ waveforms (with simulated design impairments) to the four ESGs shown in Figure 3.

After downloading the simulated waveforms to the ESG, the test set-up in Figure 1 was used to measure the resulting test signal. The resulting test signals being output from the ESGs are centred at 1.9 GHz. These are captured with the wideband multichannel oscilloscope and demodulated by the VSA software as shown in Figure 4.Note that the layer 0 and layer 1 constellations now show significant dispersion (layer 2 and 3 show similar dispersion, but are not shown here). At first glance, this looks similar to dispersion resulting from amplifier gain compression distortion or LO phase noise.

However, the peak EVM is high (43%), so error vector spectrum (EVM vs subcarrier) and error vector time (EVM vs symbol) were evaluated to investigate the composite EVM result. This revealed a symbol-to-symbol variation with the reference signal, so the downlink profile on the VSA was modified to only show the reference signal, as shown in Figure 5.

To investigate this more closely, look at the MIMO information table in Figure 6.The RS EVM time shows that one pair of antennas is showing worse behaviour. (Reference signals are transmitted on consecutive timeslots between antenna 0/1, then antennas 2/3. The RS EVM values are computed across subcarriers and the lines connect successive averages.)

Row 1: tx1/rx0, tx2rx0, and tx3/rx0 - or crosstalk from transmit antennas 1-3 on receive antenna 0.The table is quite useful in showing the effects of antenna crosstalk, showing:

• Row 2: Crosstalk from transmit antennas 0, 2, and 3 on receive antenna 1.
• Row 3: Crosstalk from transmit antennas 0, 1, and 3 on receive antenna 2.
• Row 4: Crosstalk from transmit antennas 0-2 on receive antenna 3.

The individual RS EVM values are relatively low even though there is crosstalk between the channels. As discussed earlier and illustrated in Figure 1, MIMO reference signals should be orthogonal in both time and frequency, so RS EVM is typically not impacted by antenna crosstalk, unlike the composite EVM, which is impacted by antenna crosstalk.

However, examination of the RS timing values in the MIMO information table shows timing errors between the antenna channels ranging from 2.3 to 3 µs (tx1/rx1, tx2/rx2, tx3/rx3).

This is problematic, because timing errors which approach or exceed the LTE cyclic prefix duration (4.69 µs) can result in a loss of RS orthogonality. A loss can affect measurement accuracy, such as the crosstalk values displayed in the MIMO table, PDSCH constellation and EVM results.

Consider the impact of timing error on antenna crosstalk measurement results. The reference signals from the different transmit antennas remain orthogonal as long as the delay between the channels is considerably less than the cyclic prefix duration.

However, if this condition is not met, orthogonality can be broken and can be perceived as crosstalk between the channels.

Referring back to antenna port 0 in Figure 1, the presence of signal power on R₁ subcarrier location indicates crosstalk. Timing errors, or delays between channels, can cause the R₁ subcarrier location to contain power from the previous symbol, which can be interpreted by the VSA as crosstalk between the channels and result in a false value of crosstalk reported.

To examine the timing errors reported by the MIMO table, the oscilloscope is used to measure the timing error between the antenna channels as shown in Figure 7.The timing error between the ESG being used to generate the antenna 0 signal and the ESG being used to generated the antenna 1 signal shows an error of about 2.35 µs, which correlates to the RS Timing errors reported by the MIMO table.

The antenna 1, antenna 2 and antenna 3 ESGs are triggered from the antenna 0 ESG. Once the timing error was measured with the oscilloscope, the pattern trigger delay for antennas 1-3 ESGs were adjusted to resolve the timing error.

The resulting MIMO information table in Figure 8 shows the timing errors are now within 134 ns (now only 2.8% of the cyclic prefix), ensuring orthogonality between the RS signals. The antenna crosstalk values now being displayed correctly reflect the modelled antenna crosstalk in Figure 3.

Four-channel MIMO measurements introduce a multitude of testing challenges, and make troubleshooting and debugging more difficult. This article examined transmit antenna timing error, which can impact LTE MIMO reference signal orthogonality, and thus impact measurement results such as antenna crosstalk, constellation and EVM. Multichannel wideband oscilloscopes are suitable for two- or four-channel MIMO measurements, as well as to help diagnose potential timing errors between transmit antenna channels.For example, antenna crosstalk may not impact the RS EVM value but can impact the composite EVM. Other RF transmitter impairments, on the other hand, such as phase noise and PA gain compression can negatively impact both RS EVM and composite EVM.Figure 9 shows the composite EVM result is now 4.1%, which is significantly lower than the previously 12.5% reported when the RS orthogonality condition was not being met.

Using the VSA software with the wideband multichannel oscilloscope enables the engineer to measure and analyse MIMO signals from a number of different perspectives: time domain, frequency domain and modulation domain.

This can help the engineer to diagnose and isolate hardware performance issues. Comparing RS EVM to composite EVM can help the engineer to understand the contributions of various error mechanisms (eg, phase noise, antenna crosstalk, PA gain compression) of the RF transmitter’s EVM error budget.

By Greg Jue, Agilent Technologies