darenm

Are You Ready for High Channel-Count MIMO?

Blog Post created by darenm Employee on Dec 11, 2017

As 3GPP moves from LTE to LTE-Advanced Pro to New Radio, MIMO continues to increase in strategic importance, adding implementation complexity and more specifically, channel count.  MIMO has moved from 2 streams to 8 streams with a small one-dimensional antenna, to “full-dimension” (FD-MIMO, 8 streams radiated through a 64-element array with beamforming) and soon, to Massive MIMO with hundreds of channels and array elements.  Are you ready to test 64, 128, or 256 channel devices?

 

The promise of higher-order MIMO and beamforming are clear:  they enable a superior end-user experience, providing higher data rates and a robust quality of service over a larger geographic area, without the need to add more highly-regulated frequency spectrum.  According to the Shannon-Hartley theorem, the three ways to increase the data capacity are:  increase the bandwidth; increase the number of channels; or increase the signal to noise ratio.

 

 

If public policy constrains you to a fixed bandwidth, then taking advantage of multipath propagation to deliver more channels of data to a user (MIMO) and sending signals in the same time/frequency resource block in two different directions (beamforming) are techniques to maximize the other two Shannon factors.  

 

There is always a price to be paid, however.  Namely, the expense of adding more physical baseband and RF channels, or antenna array elements, or both, and therefore the system implementation complexity and cost.   More subtly, there is also an increase in validation and test complexity.  “How will you quantify how well these 64-, 128-, or 256-channel radios work?” 

 

Reduced Channel Count for Test

 

It seems self-evident that if you have a 64-channel FD-MIMO system, then you need to buy a 64-channel tester.  However, is that really true for testing purposes?  Or can you design a test architecture that also takes into account cost and physical floor space?  

 

RF and millimeter-wave test equipment certainly make you think twice about these trade-offs.  Regardless of vendor, the cost per channel and larger form factors for millimeter-wave test equipment make adding wideband measurement channels a strategic decision, particularly for higher volume testing.  Moreover, one significant technical challenge is the calibration and alignment of 64 measurement channels to remove the system response of the measurement system itself. 

 

It turns out that if you have control of multiple channels of live signals, and are able to synchronize and trigger them repeatably, then you can use fewer live acquisition channels, and then algorithmically stitch them together into a larger, synthetic measurement of the true channel count.  For example, 64 RF or baseband channels for an FD-MIMO base station can be captured sequentially using a smaller array of 2, 4, or 8 measurement channels.  At least 2 channels are required for differential comparisons and alignments, but banks of 8 channels make a more optimal trade-off of measurement speed vs. equipment cost.  These architectures require precise switching, control, calibration, and triggering, and also introduce secondary considerations including post-processing and automation software, noise, flatness, timing skew, and other issues. 

 

Keysight has delivered systems like this, and calls this subsystem a “MIMO channel expander” that virtualizes a higher channel count measurement reference system--It is available as part of the S5020A MIMO reference solution. The result is that a reduced 8-channel system costs a fraction of a 64-channel system, fits in a single rack, can support tomorrow’s gigahertz bandwidths and 5G frequency bands, and successfully trades off slower acquisition time for huge savings in system cost and complexity. 

 

Calibrating Test System Flatness and Time Alignment

 

Multi-channel MIMO and beamforming systems have tight alignment and timing requirements to form and steer beams in different directions.  The primary sources of measurement uncertainty in these systems include the channel-to-channel variations in magnitude and phase (flatness), as well as time alignment (timing skew) due to cabling and instrument-to-instrument triggering (Figures 1 and 2).  

 

The S5020A MIMO reference solution includes software that addresses this challenge with calibration routines for the full channel count (due to the paths in the switch matrix) as well as each physical measurement channel in the test equipment.  Timing skew (primarily from cabling and signal distribution) is reduced to low picoseconds per channel; amplitude and phase variations are reduced to fractions of a dB and tenths of degrees across a full 1 GHz bandwidth at 28 GHz.

Figure 1 – Magnitude and phase of 8 raw measurement channels, before calibration

 Figure 1 – Magnitude and phase of 8 raw measurement channels, before calibration.

 

 Figure 2 – Magnitude and phase flatness of 8 corrected measurement channels

Figure 2 – Magnitude and phase flatness of 8 corrected measurement channels.

 

From this core set of conducted measurements, external radiation patterns can often be calculated or inferred from post-processing, reducing the need for chambers or nearfield probes (Figure 3). 

 

Figure 3 – Calculated radiation patterns post-processed from cabled, multi-channel measurements.   

Figure 3 – Calculated radiation patterns post-processed from cabled, multi-channel measurements.  Note the low residual sidelobe levels; this represents excellent measurement performance.

 

3D visualization and beamwidth predictions of the full 64-channel system are included in the MIMO software. With regards to these 3D beam visualizations, one practical observation is that uncorrected physical measurement system errors (such as gain, phase, and timing skew) increase an effective sidelobe “noise level” around the intended beam direction.  This lower “sidelobe dynamic range” masks the true sidelobe performance of the array-under-test.  When these residual channel-to-channel measurement errors are removed using calibration, the residual sidelobe “noise” can be reduced to 10-20 dB below the true sidelobes of the array-under-test (Figure 4).  Having sufficient baseband digitizer bandwidth and dynamic range, clean RF paths with low spurs and noise, and repeatable triggering/time alignment are critical to achieving this result.

 

Figure 4 – Visualized 64-element beamforming and sidelobes, based on conducted measurements, before and after calibration of the channel-to-channel flatness and RMS timing skew.

  

Figure 4 – Visualized 64-element beamforming and sidelobes, based on conducted measurements, before and after calibration of the channel-to-channel flatness and RMS timing skew.

 

Summary

 

The engineering effort required to establish the calibration and alignment of the measurement system turned out to be significant, relative to the lower complexity of previous generation single-channel and low-order MIMO systems at lower frequencies. Bundling this capability together with phase-coherent source and analyzer hardware delivered significant value to some of Keysight’s most advanced MIMO and beamforming customers.


There will always be measurement trade-offs of how many channels and what performance is necessary at specific phases of the product lifecycle: from early R&D prototyping to design validation vs. conformance testing vs. volume manufacturing.  In this case study, the key insight was that a user-controlled, repetitive stimulus and excellent raw performance allowed the same measurement to be iterated over a large channel count, with acceptable measurement results and dramatically lower cost.

 

Are you ready for 5G?  Keysight can help you handle the complexities of testing MIMO and beamforming systems.

 

Further info

Outcomes