NIWeek: Inside the Nokia/NI 5G system
This year I had a special interest in 5G communications. I have recently written about 5G, and why modular instruments are well positioned architecturally to address the challenges of 5G. Architecturally yes, but without the mmWave (millimeter wave) instruments needed for the highest frequency 5G microwave bands. I recently made an unequivocal prediction: “We will see modular microwave, and mmWave in particular, within 18 months. That is, by the end of 2016.” Perhaps I should have said 18 days.
Why mmWaves? This chart shown by Nokia at NIWeek shows the wide spectrum opportunities offered by mmWave frequencies (in yellow).
As part of the Day 2 Keynote presentation, Nokia and NI demonstrated a “5G” 10 Gb/sec wireless link operating at 73 GHz, architecturally based on PXI and LabView. For completeness, I should mention that Keysight Technologies had just announced a 5G channel sounding reference system, based on PXI and AXIe, a week earlier. In this article, however, I will focus on the Nokia/NI system.
You won’t find any press releases or data sheets about the NI products behind the Nokia system. They are yet to be released as generally available products. However, NI was more than happy to showcase the system and describe how it operates. Here’s the inside view of how the system is architected…
[continues…]
The 5G System
The key thing to recognise about this system is that it is an actual prototyping system. That is, it is not merely instrumentation to verify other 5G hardware, it is the actual prototype hardware used by Nokia.
The system is a 2×2 MIMO system. That is, it has two transmit antennas and two receive antennas. The photo below shows the receive antennas. You can clearly see that one horn is vertically polarized and the other is horizontal. The system operates at 73 GHz, with [covering] 2 GHz of spectrum. The wide spectrum availability is one of the attractive reasons of going to mmWave frequencies for 5G.
The business end of an emulated 5G receiver. The head (on top) has two mmWave channels including the antennas. Baseband processing is performed by the PXI chassis below it.
The transmitter has a similar head design, also with a vertical and horizontal polarized antenna. Each supports a 5 Gb/sec stream. Together, they achieve the 10 Gb/sec banner specification for 5G.
To see the actual 8-minute demonstration, go to this link, click the “Wednesday, August 5” tab above the video screen, and click “Nokia” on the right. Here, Dr. Amitava Ghosh from Nokia describes Nokia efforts in 5G. They had shown a 2.3 Gb/sec a year ago, and now demonstrated a 10 Gb/sec system, which they claimed as the fastest mmWave demo to date.
While the live demo on the stage spanned approximately 20 metres from the base station to the handset (both emulated by PXI and LabView FPGA), Nokia has shown the system capable of 200 metres, a key benchmark in the race to 5G. The system uses a single carrier of 2 GHz, and (to my untrained eye) appears to use a 16QAM modulation method.
Simple summary of the Nokia prototype. Data was transmitted by the emulated base station on the left to the emulated handset on the right at 10 Gb/sec.
To show that the system was actually working…
[continues…]
To show that the system was actually working, Nokia would cover one of the receiving antennas and watch the data rate drop. Besides prototyping the actual 5G link, the system also instrumented the performance, showing the QAM constellation diagrams for each path, as well as total data throughput.
What’s inside
On stage, I got a good look at the system. The base station (transmit) side had a single PXI chassis, driving the transmit head. The handset (receive) side had two PXI chassis processing the signals from the receive head. Why one chassis on transmit, and two on receive? This is due to the baseband processing for receiving being more complex than for transmitting, causing the receive side to require an extra chassis when the analogue modules were added.
The main receive-side chassis was essentially a LabView FPGA based supercomputer. It held 13 LabView FPGA modules, performing the baseband processing. Some were standard FlexRIO modules, others are custom modules that added direct links between the modules, bypassing PXI’s PCIe backplane. Jin Bains, VP of RF R&D at NI, explained that the massive amount of data processing and data transfers dictated [use of] direct links in some cases.
I asked Jin how much of the FPGA code was generated via LabView, and how much was hand coded. He responded that LabView generated well in excess of 90% of the code, all the measurement science, while some glue logic at the system level was done manually. He emphasised how key this was; that the measurement code was open and modifiable, and enabled the user to modify at will. Obviously, in this stage of 5G development, this is a useful prototyping feature.
ADCs (analog to digital data converters) sat alongside the FPGA modules. These are digitising data from the IF module, essentially generating two IQ data streams (one for each MIMO channel) that would be processed by the FPGA modules.
The IF modules sat in the second PXI chassis. See the photo below.
A second PXI chassis housed the IF modules. These modules communicated with the mmWave head on one side, and the data converter modules on the other.
There are a few things to take away from the photo. First of all, these look like soon-to-be-real products, complete with a product number: The NI PXIe-3620, described as a IF-LO Module, 2 GHz BW. I asked if they were available as a product, and received a reply “not yet”.
The big orange cables go to the mmWave head, and consist largely of power and digital control signals. The green cables are the Local Oscillator to the head. The IF signal returns from the head to the module. It will then be forwarded to the ADC units in the other chassis. Observant readers may notice that Tx and Rx connectors exist on the module. This allows the same module to work on the transmit side.
Now, let’s look at the head itself.
Image shows two stacked receive heads resting on top of a PXI chassis. That chassis houses the FPGA modules and A/D converters. Another chassis out of view of the photo houses the PXIe-3620 IF-LO module.
The head is a self-contained unit, getting power and the LO signal from the PXIe-3620, and delivering the IF back. It appears to be constructed as a single channel unit, and stacked for multiple channels. NI did not state what this may look like as a final product. The high attenuation per distance at these frequencies may make a head-based design a common implementation. The heads themselves were developed at NI’s Santa Clara (California) site, an organisation previously known as Phase Matrix.
That’s the system, so what does this all mean to 5G development?
[concludes next page…]
What this means for 5G
I had a fortuitous encounter with Dr. Truchard, National Instruments founder and CEO next to the keynote stage. I asked why he unveiled this system before it was commercially available. Basically, he said he wanted to get some of the early 5G business. He pointed out that the principal architecture, that of multiple LabView FPGA blades inserted into a high speed PXI backplane, was essentially a supercomputer, and was similar to other applications NI has enabled with the same architecture, even one the previous day. Readers can see a similar architecture deployed in Japan when researchers developed the first real-time 3D optical coherence tomography system, a medical application.
Truchard also emphasised the need for fast prototyping tools for 5G. The LabView architecture, where the embedded FPGA algorithms can be quickly modified by researchers at a high level as they learn about the system performance, delivers this quick prototyping environment. Indeed, Nokia claims that their speed jump from 2.3 Gb/sec a year ago to 10 Gb/sec today was largely enabled by the NI tools.
A word of caution to readers. While this toolset offers some impressive prototyping capabilities, it’s not actually on the market yet, and NI made no statement or commitment as to when it would be. Second, while it exhibited 2×2 MIMO, massive MIMO systems are expected to be much larger. Some phased arrays are aiming at 256 elements on the base station side. Using or emulating these arrays will be essential to beamforming algorithm design. Third, like 3G and 4G before it, 5G will require a flotilla of measurement solutions, not a single application. Physical, protocol, and network layers will all have to be defined, designed, and verified. Indeed, 5G may need more solutions due to the diversity of frequency bands and the exploding combinations of inter-RAT (handovers between networks).
So, while the Nokia/NI system has achieved some impressive benchmarks, it is just a start. But it is a start.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
