EDN Network

Subscribe to EDN Network feed EDN Network
Voice of the Engineer
Updated: 31 min 17 sec ago

A closer look at Microsoft’s custom chip duo for AI, cloud workloads

Fri, 11/17/2023 - 15:31

Microsoft, which developed silicon for Xbox two decades ago and later co-designed chips for Surface devices, has unveiled two custom chips: Azure Maia for artificial intelligence (AI) servers and Azure Cobalt CPU for cloud workloads. It shows how Microsoft is architecting its cloud hardware stack and why custom silicon is crucial in this design journey.

These homegrown chips, tailored for AI and cloud workloads, aim to work hand-in-hand with software developed to unlock new capabilities and opportunities for Microsoft’s data center services. However, Microsoft has provided few technical details about these in-house chips.

Figure 1 The two custom chips aim to optimize cloud infrastructure for Azure data centers. Source: Microsoft

Below is a sneak peek of these custom chips designed to power Microsoft’s Azure data centers while enabling significant cost savings for the company and its cloud service users.

Maia 100 AI accelerator

Microsoft Azure Maia 100 is an AI accelerator specifically designed to run training and inference for large language models (LLMs) and generative image tools. It comprises 105 billion transistors and is manufactured on TSMC’s 5-nm node. In a nutshell, it aims to enable higher density for servers at higher efficiencies for cloud AI workloads.

Named after a bright blue star, Maia is part of Microsoft’s multi-billion partnership with OpenAI; the two companies are collaborating to jointly refine and test Maia on OpenAI models. Currently, it’s being tested on GPT 3.5 Turbo, the model that powers ChatGPT, Bing AI workloads, and GitHub Copilot.

Figure 2 Maia 100 paves the way for training more capable models and making those models cheaper. Source: Microsoft

Microsoft and rivals like Alphabet are currently grappling with the high cost of AI services, which according to some estimates, are 10 times greater than traditional services like search engines. Microsoft executives claim that by optimizing silicon for AI workloads on Azure, the company can overhaul its entire cloud server stack to optimize performance, power, and cost.

“We are rethinking the cloud infrastructure for the era of AI, and literally optimizing every layer of that infrastructure,” said Rani Borkar, head of Azure hardware systems and infrastructure at Microsoft. She told The Verge that Maia chips will nestle onto custom server boards, which will be placed within tailor-made racks that fit easily inside existing Microsoft data centers.

That’s how Microsoft aims to reimagine the entire stack and think through every layer of its data center footprint. However, Microsoft executives are quick to note that the development of Maia 100 won’t impact the existing partnerships with AI chipmakers like AMD and Nvidia for Azure cloud infrastructure.

Azure Cobalt 100 CPU

Microsoft’s second in-house chip, Azure Cobalt CPU, named after the blue pigment, seems to answer the Graviton in-house chips offered by its chief cloud rival, Amazon Web Services (AWS). The 128-core chip, built on an Arm Neoverse CSS design, is designed to power general cloud services on Azure. And, like Azure Maia 100, Cobalt CPU is manufactured on TSMC’s 5-nm node.

Microsoft, currently testing Cobalt CPU on workloads like Microsoft Teams and SQL server, claims a 40% performance boost compared to commercial Arm server chips during initial testing. “We made some very intentional design choices, including the ability to control performance and power consumption per core and on every single virtual machine,” Borkar said.

Figure 3 Cobalt CPU is seen as an internal cost saver and an answer to AWS-design custom chips. Source: Microsoft

Maia 100 AI accelerator and Cobalt 100 CPU will arrive in 2024 and be kept in-house. Microsoft hasn’t shared design specifications and performance benchmarks of these chips. However, their naming conventions show that the development of second-generation Maia and Cobalt custom chips might be in the works right now.

We are making the most efficient use of the transistors on the silicon, says Wes McCullough, Microsoft’s corporate VP of hardware product development. Now multiply those efficiency gains in servers across all our data centers, and it adds up to a pretty big number, he wrote on the company’s blog.

Related Content

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post A closer look at Microsoft’s custom chip duo for AI, cloud workloads appeared first on EDN.

SiC modules meet diverse power requirements

Thu, 11/16/2023 - 23:14

SemiQ’s 1200-V SiC MOSFETs can be copackaged with or without a 1200-V SiC Schottky barrier diode (SBD) in SOT-227 packages. The QSiC modules provide a breakdown voltage of >1400 V and low on-resistance shift over the full operating temperature range of -55°C to +175°C. They are offered in 20-mΩ, 40-mΩ, and 80-mΩ SiC MOSFET categories.

Target markets for the copackaged SiC modules include EV charging, on-board chargers, energy storage systems, solar and wind energy, and many other automotive, industrial, and medical power applications. All of the modules are tested at wafer-level gate burn-in to provide high-quality gate oxide with stable gate threshold voltage.

In addition to the burn-in test, which helps to stabilize the extrinsic failure rate, stress tests—such as gate stress, high-temperature reverse bias (HTRB) drain stress, and high humidity, high voltage, high temperature (H3TRB)— ensure requisite industrial-grade quality levels.

Follow the product page link below to access datasheets and to request samples or volume pricing.

QSiC module product page

SemiQ

Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post SiC modules meet diverse power requirements appeared first on EDN.

8-bit MCUs exploit 32-bit development tools

Thu, 11/16/2023 - 23:14

BB5x 8-bit MCUs from Silicon Labs join the company’s PG2x family of 32-Bit MCUs in sharing a common development environment. The Simplicity Studio software is also the development platform for Silicon Labs’ portfolio of wireless SoCs, allowing developers to develop once and deploy in multiple product variations, regardless of whether they are connected or not. Developers do not have to learn two sets of tools and can cost-optimize their devices by selecting the part that best fits application needs.

Powered by a pipelined 8-bit C8051 core operating at 50 MHz, the BB5x family of microcontrollers generates up to 36% more compute power than other general-purpose 8-bit MCUs. They can be used in battery-operated power tools, handheld kitchen gadgets, and children’s toys, as well as industrial automation and LED/lighting control. The MCUs support a range of voltage options, from 1.8 V to 5.5 V, allowing them to operate for years on a coin-cell battery.

BB5x microcontrollers offer a choice of package sizes. BB50 variants come in 2×2-mm packages, while BB51 and BB52 devices come in 3×3-mm packages offering additional GPIOs and increased analog functionality.

The BB5x family is now generally available from Silicon Labs and its distribution partners. To aid evaluation of the microcontroller family, Silicon Labs offers the BB50 8-bit MCU Explorer Kit.

BB5x series product page

Silicon Labs 

Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post 8-bit MCUs exploit 32-bit development tools appeared first on EDN.

Earbud speaker performs ultrasonic modulation

Thu, 11/16/2023 - 23:14

Cypress, a solid-state MEMs speaker from xMEMS Labs, replaces legacy push-air sound reproduction with ultrasonic amplitude modulation transduction. Using this principle, Cypress turns ultrasonic air pulses into rich, bass-heavy, high-fidelity sound. This tiny MEMS speaker offers a high-quality, high-resolution alternative to the moving-coil speakers found in true wireless stereo (TWS) earbuds with active noise cancelling.

As an air pulse generator, Cypress comprises a modulator and demodulator. The modulator generates an amplitude-modulated ultrasonic wave (carrier) that faithfully follows the amplitude of the intended audio signal. The demodulator synchronously demodulates the ultrasonic wave, transferring the acoustic energy down to the baseband. Thus, producing the intended audible sound as a result. With its superior resolution in the time domain, xMEMS claims the Cypress speaker can more accurately reproduce advanced sound formats, including high resolution and spatial audio.

Housed in a 6.3×6.5×1.65-mm package (9-mm diagonal), Cypress is 40 times louder at low frequencies compared to xMEMS’ previous-generation speakers. It provides stronger, deeper bass that is consistent with the best 10-mm to 12-mm legacy coil speakers, including sound pressure levels of greater than 140 dB at frequencies as low as 20 Hz.

Full-function Cypress prototype silicon is now sampling to select early customers. Production-candidate samples of Cypress and the companion Alta controller/amplifier ASIC will sample in June 2024. Mass production is planned for late 2024.

Cypress product page

xMEMS Labs 

Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Earbud speaker performs ultrasonic modulation appeared first on EDN.

PCTEL expands embedded antenna lineup

Thu, 11/16/2023 - 23:14

PCTEL announced an embedded antenna platform for integrated radio deployments, providing off-the-shelf Wi-Fi solutions to a broader set of customers. The portfolio comprises compact, low-profile designs that provide wide coverage patterns in the 2.4-GHz, 5-GHz, and 6-GHz frequency bands. The antennas are intended for portable and network devices used across different vertical markets.

One of the devices, the EMB-910004, is a dual-band dipole antenna for 2.4-GHz and 5-GHz Wi-Fi applications. It mounts on a plastic support or directly on a nonmetallic surface in the host product with plastic screws and nuts or heat stakes. The EMB-910004 comes with an attached 6-in. micro-coax cable with a U.FL-type connector.

Other devices in the lineup include: the EMB-910001, a horizontally polarized monopole antenna for 5-GHz Wi-Fi; the EMB-910002 triband inverted-F antenna for 2.4-GHz, 5-GHz, and 6-GHz Wi-Fi; and the EMB-910003, a triband monopole antenna for 2.4-GHz, 5-GHz, and 6-GHz Wi-Fi.

To learn more about PCTEL’s embedded antenna platform, click here.

PCTEL

Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post PCTEL expands embedded antenna lineup appeared first on EDN.

Processor board augments endoscopic imaging

Thu, 11/16/2023 - 23:13

Omnivision’s OVMed OH0131 image signal processor (ISP) works with both reusable and disposable endoscopes connected to tablets or camera controllers. The OH0131 kernel employs image pre-processing algorithms supporting brightness, contrast, saturation, sharpness, hue, white balance, and gamma adjustment with advanced noise reduction.

The OH0131 ISP is compatible with all Omnivison medical image sensors with resolutions up to 2 Mpixels, including the OCHTA, OVM6946, OCHFA, OCHSA, and OCH2B imagers. It provides the company’s proprietary AntLinx CMOS chip-on-tip endoscopy imaging interface, as well as MIPI interfaces, for easy implementation.

The image signal processor serves as a kernel board for integration with any MIPI-input post-processing board. This combination of pre- and post-processing boards can be used to build custom handheld tablet consoles or camera control units.

“Our off-the-shelf OVMed OH0131 streamlines engineering for medical OEMs, enabling customers to get their products to market faster. The ability to purchase a medical-grade image sensor with wafer-level optics, cable, and ISP from one vendor also simplifies the supply chain,” said Richard Yang, senior staff marketing manager, Omnivison. “The OH0131 ISP is based on an ASIC system-on-chip for the highest image quality, reliability, and cost-effectiveness.”

The OVMed OH0131 63×51-mm board is available now. IEC 60601 EMC and EMI pre-scan testing of the ISP is planned.

OVMed OH0131 product page

Omnivision

Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Processor board augments endoscopic imaging appeared first on EDN.

Protecting antennas: Part 2

Thu, 11/16/2023 - 17:20

Looking at the damage done to a protected antenna over time.

A while back, we looked at how an antenna structure was provided with a protective, cylindrical enclosure.

Taking another look via Google Maps, we find an image of the protected antenna in Baldwin, NY when it looked like Figure 1.

Figure 1 Protected antenna found in Baldin, NY that is in working order.

I drove past this location on October 26th, and this protected antenna now looks like Figure 2.

Figure 2 The same protected antenna in Figure 1 that is now damaged.

The larger diameter section of the protective shroud has fallen off. The antenna elements themselves and their cabling are now in view.

Presumably, this will soon be repaired but the mishap does offer a brief (we hope) view of the goodies.

John Dunn is an electronics consultant, and a graduate of The Polytechnic Institute of Brooklyn (BSEE) and of New York University (MSEE).

Related Content             

googletag.cmd.push(function() { googletag.display('div-gpt-ad-inread'); });
googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Protecting antennas: Part 2 appeared first on EDN.

Ultrasonic sensors find new applications in IoT

Thu, 11/16/2023 - 16:15

Ultrasonic sensors have been a staple in the sensing space for several decades due to their capabilities, flexibility, and low cost. As the rise of IoT finds its way into virtually every market and industry, ultrasonic sensors have continued to find new applications in the areas of smart office, industrial and healthcare.

As a quick refresher, ultrasonic sensors function by sending out ultrasonic “chirps,” or sound waves, in the range of 23 kHz to 40 kHz, well above the human audible range of 20 kHz. These “ultrasonic” chirps bounce from nearby objects and return to the sensor. By measuring the amount of time that it takes for this send-and-receive process to occur, an object’s distance from the sensor can be calculated.

Figure 1 The above diagram shows the basic operating principle of an ultrasonic sensor. Source: CUI Devices

Ultrasonic sensors offer several benefits, including very accurate object detection. And, because they are based on sound waves and not electromagnetic waves, the color and/or transparency of an object being detected has no effect on the readings, also known as effect of material. This means that in addition to general object detection, they can also be used for liquid-level sensing or to detect glass.

Ultrasonic sensors also do not produce or require light to operate, making them well suited for applications in variable light conditions. With high refresh rates (hundreds of chirps per second), relatively small footprints and lower cost compared with other technologies, such as photoelectric, laser and inductive sensors, it becomes easy to see why ultrasonic sensors are so established.

With these basic principles and benefits in mind, we will now take a look at some ultrasonic sensor application examples in the smart office, industrial and healthcare markets.

Smart office applications

Driven by the need for safety, advancements in offerings and surging demand for sensor-based networks, Allied Market Research projects the smart-office market to reach $90 billion by 2030. Due to increased energy efficiency to support industry and local regulations, ultrasonic sensors are playing an expanded role in automating various processes around the office.

Figure 2 Ultrasonic sensors can be used for HVAC and lighting controls to turn heating and cooling equipment and lighting systems on and off using people detection. Source: Shutterstock

A prime example of this is HVAC and lighting control. Ultrasonic sensors have been employed to detect populated rooms in the office throughout the day. HVAC systems can be programmed, using this data, to heat or cool rooms that are in use or turn off the systems at night and kick back on upon first arrival.

In a similar way, ultrasonic sensors can control automatic motion lighting as people enter and leave certain rooms or areas of the office. While simple in nature, the energy savings of cutting back on HVAC and lighting in unoccupied office space offers significant cost savings when powering large office buildings. To sense objects across areas this large, an ultrasonic transceiver is ideal due to its detection range of up to 15 meters and wide beam angle of 80°.

A few other smart office applications worth mentioning are touchless building entry and hygiene devices. Ultrasonic sensors have long been utilized in automatic door entries as well as touchless hygiene products, such as soap dispensers, faucets, paper towel dispensers and waste bin lids. Due to the Covid-19 pandemic, these commonly recognized applications have seen ever-increasing demand as public health and safety became crucial for local businesses and offices.

Industrial IoT applications

Industry 4.0 and its demand for real-time analytics has led to surging growth in the industrial IoT (IIoT) market as companies look to target enhancements in manufacturing and automation. One of the main goals of utilizing IIoT is targeting inefficiencies and problems sooner to save time and money and to support business intelligence.

Edge devices, which transmit data between local networks and the cloud, are being further integrated into industrial settings for optimization of process and output. These include optimized quality control, sustainability, asset tracking, gaining efficiencies and predictive maintenance.

When it comes to ultrasonic sensors, they can be used in manufacturing processes for automated process control. By detecting an object passing through the line, an ultrasonic sensor could in turn trigger certain parts of the manufacturing process. High-speed counting and box sorting are other potential use cases for ultrasonic sensors due to their high refresh rates.

In these types of applications in which precision and a close range of detection are required, an ultrasonic transceiver with its close detectable range, aluminum housing and IP68 rating to deal with the often-harsher environmental conditions of industrial systems, is a good option.

While ultrasonic sensors are perhaps most often thought of for general object detection, as mentioned earlier, their ability to accurately detect translucent objects like water makes them well suited for liquid-level sensing applications. In this scenario, an ultrasonic sensor is mounted at the top of a chemical-holding tank to monitor fluid levels. As the ultrasonic sensor sends out its chirps, it can detect the liquid levels of the tank by measuring the amount of time it takes for the chirp to return.

As the liquid level decreases, the longer it takes for the chirp to return to the sensor, and vice versa, it gives an operator real-time data on tank fluid levels used for monitoring and reporting. Again, an ultrasonic sensor carrying an IP rating is an important consideration for this type of application.

IoT in healthcare applications

Medical IoT refers to technology that collects, analyzes, alerts, and saves medical information via many different platforms and devices, with an intention to streamline patient’s health and open more accessibility to doctors and other healthcare providers in the sharing and monitoring of a patient’s care. According to Market Research Future, the IoT healthcare market is projected to reach over $320 billion in the next five years.

Intra-office communication is highly important within a facility supplying medical procedures and patient monitoring. Ultrasonic sensors have been used in tandem with wearable communication devices to open communication lines when entering a room. This works on a similar principle as motion-activated lights, only in this case, the ultrasonic sensor detects motion and triggers the communication system.

Of course, when one hears the term “ultrasonic” in the medical field, it is easy to think of ultrasonic imaging via an ultrasound machine. Ultrasound machines create images by using both the echo time of the ultrasound wave and the Doppler shift of the reflected sound to determine the distance to the targeted internal organ and its movement.

While most ultrasonic sensors typically operate at frequencies between 23 and 40 kHz, this application benefits from high-frequency ultrasonic sensors that operate at 100 kHz and above. Although higher-frequency models have a shorter detectable range, they also have a smaller “blind zone,” or area closest to the ultrasonic sensor.

For example, CUI Devices’ CUSA-TR07-008-500-TH67 has a detection range as low as 3 cm, making it well matched for an ultrasound machine in which the device is extremely close to the object it is trying to detect.

Figure 3 Higher-frequency ultrasonic sensors have a shorter detectable range, making them suited for ultrasound machines. Source: Shutterstock

Final design considerations

When searching for ultrasonic sensors, they can be acquired as independent transmitters and receivers or as a combination of the two in a single unit, known as an ultrasonic transceiver. A designer will also need to decide between an analog or digital output as well as the beam angle.

Deciding on the appropriate beam angle will ultimately depend on the intended use of the end system. Wider beam angles are great for general object detection, where perhaps less precision is required, while narrow beam angles can avoid detecting false positives over longer distances. Other considerations include whether to select a more standard or high-frequency sensor and whether the intended operating environment might benefit from a sensor with an IP rating.

The diversity of applications for ultrasonic sensors can be attributed to their straightforward operating principle, reliability, and cost-effectiveness in a range of designs. As new markets continue to emerge in IoT, automation, robotics and more, ultrasonic sensors will remain a go-to solution when object detection and monitoring come into play.

Since joining CUI Devices in 2014, Rex Hallock has been a key member of the product management team. He has overseen various product lines over the years, including the expansion of CUI Devices’ thermal management group, which now features a range of DC fans, heat sinks, Peltier modules and thermal accessories.

Recommended Reading
Back to basics: An introduction to ultrasonic sensors

Related Content

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Ultrasonic sensors find new applications in IoT appeared first on EDN.

Simultaneously control gain on two independent channels with a single element pot

Wed, 11/15/2023 - 16:59

A question that arises frequently in audio and signal processing applications is this: How can I control gain on dual channel (e.g., stereo) inputs simultaneously with only a single knob? Of course, an obvious solution would be to simply use a dual element ganged pot. But ganged pots, particularly the precision multiturn variety, are relatively expensive specialty items.

This design idea offers an alternative. It avoids the liabilities of dual element ganged pots by controlling the gain of both channels with just one ordinary single element pot, R. Two implementations are illustrated. One using a quad op-amp (see Figure 1) suitable for both AC and DC signals, and another with four discrete transistors good for AC (e.g., 20Hz to 20kHz audio) only (Figure 2).

Both schemes hinge on a connection of R with its wiper terminal grounded. This creates two mechanically linked but electrically independent variable resistances, A and B.

A =W R and B = (1 – W) R

 W represents R’s wiper’s position, going from 0 to 1 as R is rotated from full counterclockwise (0) to full clockwise (1). R is the total element resistance.

Wow the engineering world with your unique design: Design Ideas Submission Guide

Figure 1 Op-amp solution requires four amplifiers comprising two Howland current pumps, a differential amplifier, and a boatload of precision resistors; it is also DC coupled.

In Figure 1, amplifier A2 and its surrounding resistors are connected to create a Howland current pump, injecting

i = Ain / 2k

 into the WR portion of the pot, to generate the Aout signal:

Aout = i W 2k = (Ain / 2k) W 2k = Ain W 2k / 2k = W Ain.

Simple enough. But what about the B channel? That starts our similarly, thanks to buffered Howland source A2 and A3, a current

i = Bin / 2k

 is injected into the (1 – W) R resistance to generate:

v = i(1 – W) 2k = (Bin / 2k )(1 – W) 2k = Bin(1 – W) 2k / 2k  = Bin(1 – W).

This signal is then subtracted from Bin by difference amplifier A4 to generate Bout as:

Bout = Bin – Bin(1 – W) = Bin(1 – (1 – W)) = W Bin.

Trimmer Bnull is provided to fine tune Bout = 0 cancellation at W = 0.

Figure 2 implements essentially the same functionality, but with AC coupling (to allow for transistor DC bias networks) and old-fashioned discrete components. I liked drawing it up mainly to prove to myself that I still remember how to work out transistor linear amplifier biasing networks!

Figure 2 Discrete solution consists of four transistors comprising three current sources and a differencing stage and is AC coupled.

Q1 is a straightforward current source driving the top half of the pot to generate:

Aout = W Ain.

Q2 does the same thing at the bottom half of the pot, generating a voltage at the base of Q4:

Q4b = B(1 – W).

This is subtracted by Q4 from the signal generated at its emitter across R4 by Q3 to yield at the collector of Q4:

 Bout = B(1 – (1 – W)) = W Bin.

And look, Ma! No op-amps! Also, only one polarity power supply. 

However, you may have noticed that both Figure 2’s A and Bsignal pathways are inverting. In audio applications this is generally not a concern so long as the inversion is symmetrical as it is here. But if it would be problematic in your intended application, you better go with the op-amp solution.

 Stephen Woodward’s relationship with EDN’s DI column goes back quite a long way. Over 100 submissions have been accepted since his first contribution back in 1974.

 Related Content

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Simultaneously control gain on two independent channels with a single element pot appeared first on EDN.

Extending network-on-chip (NoC) technology to chiplets

Wed, 11/15/2023 - 09:43

A monolithic integrated circuit (IC) is one in which everything is implemented on a single silicon die, also called a chip. The maximum practical size for a die using extreme ultraviolet (EUV) lithographic process is around 25 mm x 25 mm = 625 mm2. Although it’s possible to build larger dice, their yields start to fall off rapidly. So, one solution for today’s multi-billion transistor devices is to disaggregate the design into multiple smaller dice mounted on a silicon interposer, presented in a single package. In this case, the smaller dice are referred to as chiplets or tiles, while the final device is known as a multi-die system.

There are multiple advantages associated with adopting a chiplet-based approach. These include increased yield, reduced die cost, and the ability to implement different functions on optimal process technologies. Also, there are increased flexibility and customization options because designers can pick and choose the appropriate chiplets for different applications. This method delivers increased scalability because more chiplets can address higher workload demands and reduced time to market by reusing existing chiplets in various combinations across different products.

A few companies, like Intel, have been using chiplet-based technologies for several years, but these companies are typically outliers who have total control over the entire design flow. The dream is for multi-die system developers to be able to acquire hard chiplet IP from multiple vendors in the same way that today’s SoC designers employ soft IP functions from third-party suppliers.

It’s widely assumed that chiplets will power designs of the future, but what do we mean by future? As little as a year ago, industry pundits were predicting a five-to-six-year timeline for widespread adoption. However, several companies have recently emerged from stealth mode with chiplet offerings, indicating that adopting chiplet technologies may come sooner than expected.

As usual, of course, there’s an elephant in the room. Many of tomorrow’s chiplets will surpass the size and complexity of today’s ICs—ASICs, ASSPs and SoCs. Furthermore, the majority of today’s ICs employ some form of network-on-chip (NoC), which may be viewed as an interconnect IP that spans the entire IC. How will these NoC-based chiplets communicate with each other?

D2D interconnect scenarios

It’s possible to identify a variety of chiplet-to-chiplet interconnect scenarios. Such interconnect is usually referred to as die-to-die (D2D) to avoid confusion with chip-to-chip (C2C) interconnect at the printed circuit board (PCB) level. First, consider some non-coherent D2D interconnect possibilities (Figure 1).

Figure 1 Here are three non-coherent interconnect examples. Source: Arteris

The simplest option involves only two chiplets with direct D2D connections, as illustrated in Figure 1a. A more sophisticated example involves a greater number of dice (Figure 1b), still with direct D2D connections and static mapping mode configuration at boot time. In the case of indirect D2D routing involving chiplet hopping (Figure 1c), there are two possibilities: static mapping mode configuration at boot time or dynamic mapping mode configuration at run time. All three examples in Figure 1 assume heterogeneous dice, but multiple homogeneous (identical) dice are also an option.

Next, consider some coherent D2D interconnect examples (Figure 2). In this case, in addition to any on-chiplet memory like processor and accelerator caches, we are also showing possible deployments of external memory (MEM) like DDR, represented by the larger gray rectangles. These memories, which are external to the multi-die system package, will require on-chiplet memory controller IPs, as shown by the smaller gray rectangles.

Figure 2 The above diagram shows three coherent interconnect examples. Source: Arteris

The simplest form of coherent interconnect is heterogeneous and asymmetric, as illustrated in Figure 2a. In this case, there is a clear host chiplet to which the external memory is connected. At the other end of the spectrum, we have a homogeneous and symmetric architecture (Figure 2c). In this case, every chiplet can talk to its own memory and all other chiplets’ memories. Obviously, this quickly becomes complex. Also, the designers need to be extremely careful with respect to any bottlenecks and latencies associated with D2D communications.

Of particular interest to me is that, while I was attending the world’s first automotive-focused chiplet event, which was held in Leuven, Belgium, there was talk of having a special NoC chiplet that provides all the other chiplets with access to a shared memory while also acting as a kind of arbiter (Figure 2b). The idea would be to have this chiplet, shown as Die X in the figure, act as a hub. The other chiplets are competing for access to the central shared memory, and it’s necessary to regulate the cache coherency. This scenario allows designers to build intelligence into the hub.

A deeper D2D interconnect dive

Let’s look a little deeper into the D2D interconnect (Figure 3). We will start with the NoCs employed on the chiplets themselves. Various NoC technologies are available to designers. For example, the Advanced Microcontroller Bus Architecture (AMBA) from Arm embraces the non-coherent Advanced eXtensible Interface (AXI) protocol and the Coherent Hub Interface (CHI) protocol.

Figure 3 The die-to-die (D2D) interconnect example highlights NoC (left). Source: Arteris

Assuming the designer is using a NoC protocol like AXI or CHI—or NoC IP that can generate and receive AXI or CHI traffic—then any outbound traffic will have to be packed into some streaming interface format like CXS. The packed data is then passed to a link layer controller and associated PHY.

The physical layer will be implemented using something like Bunch of Wires (BoW), Universal Chiplet Interconnect Express (UCIe), or Synopsys eXtra Short Reach (XSR). Similarly, inbound traffic will be passed through the associated PHY and link layer and unpacked into AXI or CHI.

Early days, multiple options

It’s important to note that we are still in the early days of this technology, and people are still figuring out the various ways in which everything and everyone might play together. For example, since chiplets may employ IP blocks from various third-party vendors—and since each IP block may employ its own data width, clock frequency and interconnect protocol—it may be that the NoC needs to accommodate multiple standard protocols that have been defined and adopted by the industry, such as OCP, APB, AHB, AXI, CHI, STBus and DTL.

To address this issue, chiplet designers may turn to the non-coherent and coherent interconnect IPs because both of these NoCs support a wide range of protocols.

If chiplet designers choose to use interconnect IP, they may implement the pack/unpack IP themselves and acquire the link layer and PHY IP from a third-party vendor. Alternatively, it may be that the pack/unpack IP is bundled with the link layer and PHY IP. Yet another alternative is that the pack/unpack IP is provided as a module by the NoC vendor.

Irrespective of the nitty-gritty details, it’s becoming obvious that chiplets and multi-die systems are the wave of the future in electronic design due to their myriad of advantages with respect to cost, yield, flexibility, scalability, and customization. Just when we thought things couldn’t get even more exciting… they did!

Frank Schirrmeister, VP solutions and business development at Arteris, leads activities in the automotive, data center, 5G/6G communications, mobile, aerospace and data center industry verticals. Before Arteris, Frank held various senior leadership positions at Cadence Design Systems, Synopsys and Imperas, focusing on product marketing and management, solutions, strategic ecosystem partner initiatives and customer engagement.

Related Content

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Extending network-on-chip (NoC) technology to chiplets appeared first on EDN.

Photocell makes true-zero output of the op-amp

Tue, 11/14/2023 - 15:56

While choosing an op-amp buffer for a new high-resolution single-supply DAC, a source of negative supply was considered because the buffer op-amp had to provide true zero voltage on its output.  

Wow the engineering world with your unique design: Design Ideas Submission Guide

For instance, a typical rail-to-rail output op-amp can’t provide true zero voltage, it can guarantee at least several mV on its output, while a high-resolution DAC can have resolution in the of tens microvolts. The application required a true zero output, hence the problem.

For sure, some negative supply source was needed to increase the “headroom” around zero. (I use the term “headroom” because we are dealing not with an upper, positive supply, but with lower one. A better word would be “footroom”.)    

There was an intention to use the Cuk configuration circuit again, like the old circuit in EDN, but with an output voltage of about -1 V only and a low—less than 2 mA—output current.

While exploring alternatives, the idea arose to use a photocell instead of any ordinary voltage converter. It resulted in the circuit in Figure 1.

Figure 1 Using photocells instead of a voltage converter to help provide a true zero volage on the output of an op-amp buffer for a high-resolution single-supply DAC.  

The solution has comparable dimensions with the circuit based on the Cuk configuration, albeit lower efficiency.  But since the superfluous power doesn’t excess 0.1 W, this may be of no importance.

Such a solution has important advantages:  

  • It’s far simpler.
  • It produces very low electrical noise—a fact of great significance when you are dealing with low analog signals. (In this circuit the output noise was less than 1 mV even without output capacitor C1.)  
  • Any over-voltages on its output are excluded (while the Cuk converter can produce such over-voltage if any problems with feedback occurs).
  • The perfect level of isolation should also be noted, albeit it is not important in our case.

Since the external outlines of the gadget are determined by the photocell, the tiny photocell AM-1417 (of Toshiba) was used. Its dimensions are only 34 x 14 x 2 mm, and 4 sections it has—hence 4 LEDs, one for each section—produce about 3 V without any load.

The 4 LEDs are quite ordinary ones of bright red family (L-513HURC, 1800 mcd in 15° angle) because Si photocell has its maximal efficiency in this area of spectrum.

Reds are also preferable for +5 V supply since their low forward voltage allows to double the efficiency very simply, by stacking them in pairs with the same current through both.  

The circuit produces 490…520 mV on 2k load @ 20 mA current through the LED. This is more than enough for several micropower op-amps such as the AD8603/AD8607.  

The output voltage of the photocell can be varied by changing the current through LEDs.

The photocell is a current—not voltage—source, so the capacitor C1 is required to reduce the output impedance of the circuit. Diode D1 enables a path for sinking current and protects this electrolytic capacitor if the negative voltage for some reason disappears.

As I mentioned, the output power is quite enough for a precision micropower op-amp, such as the AD8603, for example.  If you need more power, you can use higher current through LEDs, more efficient pair LED/photocell, or simply connect more such circuits in parallel.

Peter Demchenko studied math at the University of Vilnius and has worked in software development.

Related Content

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Photocell makes true-zero output of the op-amp appeared first on EDN.

Dissecting a malcontent (and moist?) microwave oven

Mon, 11/13/2023 - 18:43

Back in February 2015, my wife and I purchased Whirlpool’s WMC30516AB 1.6 cubic foot microwave oven from Amazon:

It lasted a bit more than four years. Toward the end, various segments of the four-digit, seven-segment per-digit display would flake out, then return, and random buttons on the front panel would also stop working only to revive later. Eventually, the resurrections ceased and, although the remainder of the microwave oven presumably still functioned fine, there no longer was any way to control it (notably via the no-longer-responsive “Start” button).

So, we hauled it to the dump and replaced it in July 2019 with…another Whirlpool (what can I say, my wife likes the brand, or should I say liked it), the familial WMC30516HV, this time for more than twice the price of its predecessor at Home Depot:

Stop me if you’ve heard this before: a bit more than four years later, and around a month ago, the replacement unit failed in exactly the same way. And if you search online for reviews of either model, you’ll find that plenty of other owners’ units have suffered the same fate.

We’ve subsequently purchased a Panasonic 2.2 cubic foot NN-SN966S in like-new claimed condition from Amazon’s Warehouse sub-site:

I’m happy to report that it was as advertised; I don’t think it had even been taken out of the box by the original purchaser (list price: twice what we paid) prior to being returned for resale. But given that we had two related-family Whirlpools die exactly the same way at near-exactly the same time, the engineer in me suspected a fundamental design flaw somewhere. So before taking this one to the dump…you guessed it…I decided to take it apart, in part because more generally I’ve always been curious about what’s inside one of these things.

My upfront suspicion was that moisture accumulation resulting from poor ventilation flow through the unit’s interior while in use was backflowing into the electronics area of the system and eventually causing something(s) on the PCB to fail. Part of this, I admit, might be our “fault”. I’m such a tightwad that if my wife doesn’t finish a Starbucks drink, I toss it in the fridge and heat it up and drink it the next morning as part of my daily caffeine intake. Further to that point, brewed coffee that we don’t finish goes into a carafe for me to later reheat and consume, too. Then there’s corn in the cob, soups, and plenty of other moisture-rich food and drinks that regularly find their way to the microwave for cooking and otherwise heating up…

But I’m not willing to shoulder all, or even most, to be blunt, of the blame. For one thing, I don’t think our usage pattern is all that atypical. And for further evidence of a potential fundamental design flaw, check out this case study example review of the WMC30516AB, complete with submitted photo and titled “Major steaming problem, and no help at all from Whirlpool”, that I found on Amazon in the process of finalizing this writeup:

This Microwave had a major problem with steaming up, even with small cook loads like a few slices of microwave bacon. This soon led to obvious streaking and spotting on the inside of the (non-cleanable) viewing window. Whirlpool customer service was the worst part. They insisted that this steaming situation is “normal performance,” though I’ve never seen another microwave steam like this. I requested a service call to evaluate the Microwave but Whirlpool refused.

Come to think of it, I’d noticed seemingly excessive interior condensation accumulation with our two units, too.

Enough preparatory chatter; let’s get to the teardown. Here’s our patient (for the bulk of this project, I’ve moved from my usual office desk to the workbench directly below it in the furnace room for perhaps-obvious available-space reasons, although the lighting’s not as stellar there):

Note that, unlike with my new Panasonic, there are no outgoing airflow vents (either passive or active) on the left side. Hold that thought:

Speaking of airflow, the back’s where the bulk of the action is:

Air is forced into the unit by a fan behind the ventilation hole mesh at left. It flows through the electronics, from there passively transitioning (theoretically, at least) into the main cooking cavity of the microwave oven, then again passively out vent holes on the opposite upper side and back corner of the interior (upper right side and back right corner, from this photo’s perspective). And how does the air exit the microwave oven? Through those passive vents you see at top and on the right edge in the photo, all at the back and on the right (again, from this rearward perspective, at least) half of the unit, mostly making a 90° turn in the process, again, versus directly out the opposite side with the Panasonic approach.

Before continuing, a couple of close-up sticker shots:

Now for the right side, those aren’t actual air vents, by the way, only cosmetic metal “trenches”:

Top:

And finally, the bottom:

Note that there are functional passive air vents here, too, but their locations are curious. They’re predominantly on the air-outflow half of the microwave oven, but since the air will be heated (albeit moist, therefore heavier than when it entered) at this point, and since hot air rises, not falls, I question just how functional they really are.

Back to the front; let’s now pop open the door:

The inside of the door is conventional for a microwave oven, as far as I can tell from my limited, elementary experience with these devices (and shielded, of course, for obvious reasons):

Again, the airflow direction through the interior is right to left from this front-view perspective. That metal plate on the right side is a cover for the waveguide, called a mica plate:

A couple more sticker closeups before continuing:

Before diving in, I decided to satisfy my curiosity and see if the microwave oven’s several-week sojourn sitting downstairs unplugged and awaiting “surgery” had left it reborn, as had happened (temporarily, at least) in the past. Nope:

The “8” front control panel button still worked, so you can tell which segments (of which digit) failed:

But many of the other numerical and functional buttons remained non-responsive…again, including the all-important “start”. Oh well.

Onward. You may have already noticed the large Torx head screw at the bottom of the right side of the unit, and the four additional ones around the edges of the back side. Let’s get those off:

With them removed, the unified right-top-left panel slides right off the back:

From boring-to-exciting (IMHO) order, here’s the now-exposed left side:

Top side:

Complete with warning (hah!) sticker closeup:

And the right side, where all the electronics action happens:

Your eye will likely be immediately drawn to the cavity magnetron at the center, behind which (not shown) is the aforementioned waveguide:

That metal shroud to the right draws in ambient air from the outside to keep it cool. Speaking of which, this doodad perched above it:

is, I’m assuming, a temperature sensor to ascertain whether the magnetron is overheating due to, for example, using the microwave oven with nothing inside it or with metallic contents.

Below the magnetron is a hefty transformer:

And to its right is an equally formidable capacitor:

In the upper right of the earlier overview shot is a small PCB:

Presumably, particularly given the diminutive size of the onboard fuse, it does AC/DC conversion for only a subset of the entire system circuitry.

And at far right is the fan:

Now let’s move to the left side of that earlier overview shot. First off, here’s the light bulb, which shines through the passive air inflow vents to illuminate the interior:

To its left and below are three components whose purpose wasn’t immediately obvious to me:

until I purposelessly pressed the latch to open the microwave door and noticed that they’d also transformed in response:

These are, I believe, triple-redundancy switches intended to ensure that the magnetron only operates when the door is closed.

Last, but not least, let’s look at the main system PCB at far left, which is the upfront intended showcase (not to mention the presumed Achille’s Heel) of this project:

Here’s a slightly tighter zoom-in:

First step: unhook the various bits of cabling connecting it to the rest of the system:

Two screws are immediately visible along the left edge. But removing them:

didn’t free the PCB from its captivity:

Looking again, I found another one hidden among the connectors, capacitors, and such on the right side of the PCB:

That’s more like it:

Left behind, among other things, is the oddly-varying-contact-color ribbon cable that originally routed between the PCB and the front control panel:

And here’s one more wiring mention; referring back to earlier airflow comments, I at first thought that the two wires heading underneath might be going to a ventilation fan, intended to pull cool air into the body cavity from the outside to the underside:

But after pulling off the bottom panel to expose the otherwise unmemorable under-interior to view, I realized that they were instead connected to (duh on me) the glass turntable motor:

Now back upstairs, where the lighting’s better, for the rest of the PCB analysis:

Let’s stick with this latter side for the first closeup shot set. Here’s that faulty-segments display:

and the exposed portions of this side of the PCB, dominated by solder points and traces:

Did you notice, though, what looks like one corner of an IC sticking out from under the display, further exposed after slipping off the surrounding gasket?

Let’s see what some side views reveal:

Yep, there’s definitely a large lead count chip underneath. Fortunately, by unclipping two of the plastic “legs” from the bracket surrounding the display, I was able to swing it out of the way, revealing both its underside and the remainder of this side of the PCB:

The glossy finish atop the IC makes it very difficult to read (far from photograph) the product markings, so you’ll need to take me at my word that it’s a M9S8AC16CG microcontroller, containing an 8-bit S08 CPU, 16 KBytes of flash memory and 512 bytes of SRAM, and still with its original Freescale Semiconductor vendor logo stamp atop it (the company, therefore product line, were later merged into NXP Semiconductors).

Let’s now flip the PCB over to its other side, starting with some side views. Check out, for example, that circular “beep” piezo transducer near the middle:

And, wrapping up, a couple of full-on closeups, starting with the top half:

The two ICs you see at left are an I-core AiP24C02 2 Kbit EEPROM (what an EEPROM is doing in a microwave oven is beyond me, unless it’s used for operating dataset fine-tuning on the assembly line, or something like that) and, below it, an unknown-supplier LM358 dual 30V 700-kHz op amp.

Now for the other (lower) half:

The clutch of ICs in the lower right corner comprises two chips oriented 180° relative to each other and strangely stamped:

1730, preceded by an upside-down 7 in a larger font size
817 C
WK

and below and to the left of them, Power Integrations’ LNK364 AC/DC converter.

No obvious failure culprit emerges from my visual inspection of the PCB; see anything, readers? It kills me that the likely moisture- or heat-induced (another potential side effect of poor ventilation, of course) failure of a single inexpensive component on this board is likely what caused the demise of the entire expensive microwave oven, but that’s our “modern disposable society” for you, I guess…Sadly, even if I could fix it, I’d be reluctant to pass it on to someone else without a plethora of upfront qualifiers, because it’d likely only be a matter of time before the unit died again, due to its innate shortcomings.

With that, I turn it over to you for your thoughts in the comments!

Brian Dipert is the Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.

 Related Content

googletag.cmd.push(function() { googletag.display('div-gpt-ad-native'); }); -->

The post Dissecting a malcontent (and moist?) microwave oven appeared first on EDN.

Pages