EDN was able to tour the Si Labs engineering facility in Austin during Embedded world North America. The headquarters were located a convenient 15 minute walk from the convention center, making it a pretty natural choice to explore. The tour mainly covered the analog and RF testing and validation processes with senior applications engineer Dan Nakoneczny and RF engineer Efrain Gaxiola-Sosa. Upon entering, it was readily apparent that the facility was nicely stocked with high-spec test equipment to conduct the range of tests required. The lab managed to host large windows that let in the Texas sun instead of the usual test lab experience—an “instrumentation cave” tucked away in a basement or some windowless-wing of a building. The unfortunate side effect of this was an increase in the necessary cooling required, causing a constant hum of white noise.
Analog testing
Dan Nakoneczny began by showing us the benches to test analog peripherals such as comparators, ADCs, voltage regulators, and oscillators. Generally, the team uses sockets to hold down the device under test (DUT) which could be any of Silicon Labs’ SoCs where the various current, voltage, and timing measurements were not massively impacted by the socket and they allow for one test setup between devices. “We can use the sockets for most of our tests but for other tests, like DC/DC converters, we have to solder the parts down to our boards,” explained Dan. The test bench included oscilloscopes, power supplies, function generators, and a binary counter, “with analog peripherals, you don’t have pins so you’ll have to rely on simulations and have a more indirect way of taking measurements of your system.”
For this lab in particular, the DUTs can range from the typical devices in production to prototypes, “we’re building platform devices and a lot of this IP will be used in the next device that will come out at the end of the year or in a couple of months, so the validation team is between the product engineering and design team trying to find the small parts per million bugs, or issues that a customer might find during high volume production ten years down the road. We can make changes now before it gets stamped into subsequent designs, where we might have to do 10 different revisions,” said Dan.
The automated test setup shown in Figure 1 includes a pick and place robotic arm that uses computer vision to grab and place DUTs in the socket before pressing down the socket and locking the device in place. All measurements go up to the cloud to Silicon Labs’ database where there are special tools used to visualize the data to, for instance, compare it with past devices.
Figure 1 Automated test setup that can be left for a weekend to test 20 to 50 parts.
RF validation
Receiver station
Efrain then guided us through the RF test stations scattered throughout the lab and began at the receiver test setups that sat within large Faraday cages that provided up to 135 dB of isolation to prevent any interference. The PCB presented in Figure 2 shows the Silicon Labs motherboards that are able to receive several daughter cards, “these are developed for each of our products so that the very same infrastructure, connectivity, and flexibility in our lab can be used across multiple platforms. It’s a little challenging to keep them updated all the time, but it makes our life easier,” said Efrain. The RF tests have the unique challenge of requiring soldered down DUTs so a proper test fixture is key and using the same ones across the various RF test stations and, as much as possible, across new production devices must be a challenge. There are specialized motherboards that can go into the oven for temperature testing from -40oC to 135oC. “We have a bunch of switches and so we can test serially, but we cannot test in parallel because our equipment has a single channel for receiving information.” Efrain stressed that the most critical parameter from this test was receiver sensitivity; the better the sensitivity, the more range the wireless signal had. These test setups are also largely automated and can be remotely logged into and controlled outside of the annual calibration required to ensure there are no test errors due to drift.
Figure 2 Setup for the receiver testing with power supply, a microwave switch system, signal generator, and PXI Express backplane chassis/modules.
In-band transmit station
The next stop was the test station for in-band transmissions, “we transmit in several protocols where OFDM modulation is one of the most complex. So we want to make sure we can transmit the high data rates and that it is good enough for the receiver to actually get this information.” Efrain reminded us that the quality of the transmit signals depends largely on its error vector magnitude (EVM), causing this to be one of the more critical parameters this station was meant to measure; however, the setup only measured within the ISM bands (e.g., 2.4 GHz and 5 GHz).
Figure 3 Test station for measuring in-band transmissions.
Transmitter out-of-band station
For out of band testing (Figure 4), test and validation engineers will take a look at the emissions on other bands including cellular, radar, etc. “Ideally you want to transmit on your channel at a particular frequency alone, but you’re going to have harmonics that exist in frequencies that are a multiple of the fundamental frequency,” explained Efrain, “these cannot be higher than what the FCC allows.” He expressed how the nonlinear nature of fast-switching transistors are often the culprit of this EMI.
The out-of-band station is used for pre-compliance testing before sending their part off to an accredited test lab for full compliance testing. “Our equipment allows us to transmit and analyze some of this data (conducted emissions), so the output goes to the switch, the switch multiplexes the signal from the chip being tested, and this goes to the port of the spectrum analyzer where we can do several operations,” Efrain stated. An oscilloscope can be used in the place of the spectrum analyzer as well to perform other measurements. The power supplies within the setup must be quiet and clean to remove any unnecessary inference from the test instruments themselves. There are also battery emulators within the setup since many of Silicon Labs’ devices function with batteries.
Efrain continued, “We are sending a signal with a given power say, 1 mW or 0 dBm, where we can go up to 20 dBm. We want to transmit at the highest power possible where one of the key figures is the output power of our power amplifiers; however, if we reach high output powers and we do not pass FCC or ETSI requirements, we cannot sell.” In this station the power of the fundamental is isolated and a notch filter is used to remove it and look at what is appearing at the harmonic frequencies. “If we leave the fundamental there, some energy will leak and the measurement we perform won’t be as accurate,” explained Efrain.
Figure 4 Test rack for conducted emissions testing.
Radiated emissions testing
This test setup, naturally, will not perform radiated emissions testing. The Austin facility did house a small chamber for this designed by ETS-Lindgren with a robotic arm used to adjust the DUT for testing at various orientations. This is also used for pre-compliance testing.
Receiver out-of-band emissions
At this point, we enter yet another Faraday cage, this one much larger to see how Silicon Labs tests how the receivers of their SoCs perform with interference at different bands. “We have specialized equipment to emulate a real RF environment so we test a particular set of signals that could potentially interfere with our DUT, and we want to make sure they don’t.” The setup shown in Figure 5, hosts a lot of switches so that the engineers can test at all the bands/channels of interest.
Figure 5 Test station to measure how out-of-band interference could impact the receivers on the DUT.
Load-pull stations
The load-pull stations in Figure 6 were a newer test that the validation lab used to make sure that the power amplifiers (PA) were delivering the maximum power efficiency. Efrain explained how fabrication could slightly adjust the load behavior of the DUT from being that more ideal ~50 ohms to something more reactive or capacitive, “in these two stations we are pulling the load that the PA is going to see. The change in impedance will mean that the power we are delivering is not the same and we need to identify what conditions will make our power amplifiers not behave properly and bring that back to our design.” The goal of the test was to build a robust product that meet customer expectations, “You can say you promise a certain performance only under ideal conditions, but can you control the output power and do a feedback loop to make sure that what you say is happening all the time?”
Figure 6 Load-pull stations used to find the optimal load impedance at the chip pin for maximum power transfer and PA efficiency.
Radio regression test system
The small shielded enclosures found in Figure 7 are a benchtop solution for isolation (~80 dB) used by the PHY MAC team to conduct the battery of tests necessary. There are four of these boxes carrying test fixtures with 5 different DUTs, all connected to the Keithley’s S46 microwave switch system, configured as a 2:28 multiplexer (MUX). “The team does validation at the PHY and MAC level to identify what we need to change or fix, and to make sure we don’t break anything if we make changes to firmware,” said Efrain, “when you’re working with multiple radio protocols in a single hardware platform, you need to reconfigure your radio to support these different protocols.” The test is also used to emulate the fixes that Silicon Labs develops for customers that face issues in the field, “once those issues are fixed, they’ll come here and hopefully they won’t break anything else.” The regression stations run 24/7 with daily reports on testing.
Figure 7 Radio regression testing with shielded enclosures to test PHY and MAC protocols of the various radio models used in Silicon Labs SoCs.
Aalyia Shaukat, associate editor at EDN, has worked in the engineering publishing industry for nearly a decade with published works in EE journals and other trade publications. She holds a BSEE from Rochester Institute of Technology.
Related Content
- Combating noise and interference in oscilloscopes and digitizers
- EMC: How to write a good benchtop test report
- Not all oscilloscopes are made equal: Why ADC and low noise floor matter
- How to control your impulses—part 2
- DIY RTD for a DMM
- Why deep memory matters
The post An engineer’s playground: A tour of Silicon Labs’ labs appeared first on EDN.