EDN Network

Broadcom Inc. claims the industry’s first Wi-Fi 8 silicon solutions for the broadband wireless edge, including residential gateways, enterprise access points, and smart mobile clients. The company also announced the availability of its Wi-Fi 8 IP for license in IoT, automotive, and mobile device applications.

Designed for AI-era edge networks, the new Wi-Fi 8 chips include the BCM6718 for residential and operator access applications, the BCM43840 and BCM43820 for enterprise access applications, and the BCM43109 for edge wireless clients such as smartphones, laptops, tablets and automotive. These new chips also include a hardware-accelerated telemetry engine, targeting AI-driven network optimization. This engine collects real-time data on network performance, device behavior, and environmental conditions.

(Source: Broadcom Inc.)

The engine is a critical input for AI models and can be used by customers to train and run inference on the edge or in the cloud for use cases such as measuring and optimizing quality of experience (QoE), strengthening Wi-Fi network security and anomaly detection, and lowering the total cost of ownership through predictive maintenance and automated optimization, Broadcom said.

Wi-Fi 8 silicon chips

The BCM6718 residential Wi-Fi access point chip features advanced eco modes for up to 30% greater energy efficiency and third-generation digital pre-distortion, which reduces peak power by 25%. Other features include a four-stream Wi-Fi 8 radio, receiver sensitivity enhancements enabling faster uploads, BroadStream wireless telemetry engine for AI training/inference, and BroadStream intelligent packet scheduler to maximize QoE. It also provides full compliance to IEEE 802.11bn and WFA Wi-Fi 8 specifications.

The BCM43840 (four-stream Wi-Fi 8 radio) and BCM43820 (two-stream scanning and analytics Wi-Fi 8 radio) enterprise Wi-Fi access point chips also feature advanced eco modes and third-generation digital pre-distortion, a BroadStream wireless telemetry engine for AI training/inference, and full compliance to IEEE 802.11bn and WFA Wi-Fi 8 specifications. They also provide an advanced location tracking capability.

The highly-integrated BCM43109 dual-core Wi-Fi 8, high-bandwidth Bluetooth, and 802.15.4 combo chip is optimized for mobile handset applications. The combo chip offers non-primary channel access for latency reduction and improved low-density parity check coding to extend gigabit coverage. It also provides full compliance to IEEE 802.11bn and WFA Wi-Fi 8 specifications, along with 802.15.4 support including Thread V1.4 and Zigbee Pro, and Bluetooth 6.0 high data throughput and higher-bands support. Other key features include a two-stream Wi-Fi 8 radio with 320-MHz channel support, enhanced long range Wi-Fi, and sensing and secure ranging.

The Wi-Fi 8 silicon is currently sampling to select partners. The Wi-Fi IP is currently available for licensing, manufacture, and use in edge client devices.

The post Broadcom delivers Wi-Fi 8 chips for AI appeared first on EDN.

Microchip Technology Inc. expands its Switchtec PCIe family with its next-generation Switchtec Gen 6 PCIe fanout switches, supporting up to 160 lanes for high-density AI systems. Claiming the industry’s first PCIe Gen 6 switches manufactured using a 3-nm process, the Switchtec Gen 6 family features lower power consumption and advanced security features, including a hardware root of trust and secure boot with post-quantum-safe cryptography compliant with the Commercial National Security Algorithm Suite (CNSA) 2.0.

The PCIe 6.0 standard doubles the bandwidth of PCIe 5.0 to 64 GT/s per lane, making it suited for AI workloads and high-performance computing applications that need faster data transmission and lower latency. It also adds flow control unit (FLIT) mode, a lightweight forward-error-correction (FEC) system, and dynamic resource allocation, enabling more efficient and reliable data transfer, particularly for small packets in AI workloads.

As a high-performance interconnect, the Switchtec Gen 6 PCIe switches, Microchip’s third-generation PCIe switch, enable high-speed connectivity between CPUs, GPUs, SoCs, AI accelerators, and storage devices, reducing signal loss and maintaining the low latency required by AI fabrics, Microchip said.

Though there are no production CPUs with PCIe Gen 6 support on the market, Microchip wanted to make sure that they had all of the infrastructure components in advance of PCIe Gen 6 servers.

“This breakthrough is monumental for Microchip, establishing us once again as a leader in data center connectivity and broad infrastructure solutions,” said Brian McCarson, corporate vice president of Microchip’s data center solutions business unit.

Offering full PCIe Gen 6 compliance, which includes FLIT, FET, 64-Gbits/s PAM4 signaling, deferrable memory, and 14-bit tag, the Switchtec Gen 6 PCIe switches feature 160 lanes, 20 ports, and 10 stacks with each port featuring hot- and surprise-plug controllers. Also available are 144-lane variants. These switches support non-transparent bridging to connect and isolate multiple host domains and multicast for one-to-many data distribution within a single domain. They are suited for high-performance compute, cloud computing, and hyperscale data centers.

(Source: Microchip Technology Inc.)

Multicast support is a key feature of the next-generation switch. Not all switch providers have multicast capability, McCarson said.

“Without multicast, if a CPU needs to communicate to two drives because you want to have backup storage, it has to cast to one drive and then cast to the second drive,” McCarson said. “With multicast, you can send a signal once and have it cast to multiple drives.

“Or if the GPU and CPU have to communicate but you need to have all of your GPUs networked together, the CPU can communicate to an entire bank of GPUs or vice versa if you’re operating through a switch with multicast capability,” he added. “Think about the power savings from not having a GPU or CPU do the same thing multiple times day in, day out.”

McCarson said customers are interested in PCIe Gen 6 because they can double the data rate, but when they look at the benefits of multicast, it could be even bigger than doubling the data rates in terms of efficient utilization of their CPU and GPU assets.

Other features include advanced error containment and comprehensive diagnostics and debug capabilities, several I/O interfaces, and an integrated MIPS processor with bifurcation options at x8 and x16. Input and output reference clocks are based on PCIe stacks with four input clocks per stack.

Higher performance

The Switchtec Gen 6 product delivers on performance in signal integrity, advanced security, and power consumption.

PCIe 6.0 uses PAM4 signaling, which enables the doubling of the data rate, but it can also reduce the signal-to-noise ratio, causing signal integrity issues. “Signal integrity is one of the key factors when you’re running this higher data rate,” said Tam Do, technical engineer, product marketing for Microchip’s Data Center Solutions business unit

Signal loss, or insertion loss, set by the PCIe 6 spec is 32 dB. The new switch meets the spec thanks in part to its SerDes design and Microchip’s recommended layout of the pinout and package, according to Do.

In addition, Microchip added post-quantum cryptography to the new chip, which is not part of the PCIe standard, to meet customer requirements for a higher level of security, Do said.

The PCIe switch also offers lower power consumption, thanks to the 3-nm process, than competing PCIe Gen 6 devices built on older technology nodes.

Development tools include Microchip’s ChipLink diagnostic tools, which provide debug, diagnostics, configuration, and analysis through an intuitive graphical user interface. ChipLink connects via in-band PCIe or sideband signals such as UART, TWI, and EJTAG. Also available is the PM61160-KIT Switchtec Gen 6 PCIe switch evaluation kit with multiple interfaces.

Switchtec Gen 6 PCIe switches (x8 and x16 bifurcation) and an evaluation kit are available for sampling to qualified customers. A low-lane-count version with 64 and 48 lanes with x2, x4, x8, x16 bifurcation for storage and general enterprise use cases will also be available in the second quarter of 2026.

The post Microchip launches PCIe Gen 6 switches appeared first on EDN.

Analog topologies abound for converting current to voltage, voltage to current, voltage to frequency, and frequency to voltage, among other conversions.

Figure 1 joins the flock while singing a somewhat different tune. This current, voltage, and power (IVW) DC power converter multiplies current by voltage to sense wattage. Here’s how it gets off the ground.

Figure 1 The “I*V = W” converter comprises voltage-to-frequency conversion (U1ab & A1a) with frequency (F) of 2000 * Vload, followed by frequency-to-voltage conversion (U1c & A1b) with Vw = Iload * F / 20000 = (Iload * Vload) / 10 = Watts / 10 where Vload < 33 V and Iload < 1.5 A.

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

The basic topology of the IVW converter comprises a voltage-to-frequency converter  (VFC) cascaded with a frequency-to-voltage converter (FVC). U1ab and A1a, combined with the surrounding discretes (Q1, Q2, Q3, etc.), make a VFC similar to the one described in this previous Design Idea, “Voltage inverter design idea transmogrifies into a 1MHz VFC”

The U1ab, A1a, C2, etc., VFC forms an inverting charge pump feedback loop that actively balances the 1 µA/V current through R2. Each cycle of the VFC deposits a charge of 5v * C2, or 500 picocoulombs (pC), onto integrator capacitor C3 to produce an F of 2 kHz * Vload (= 1 µA / 500 pC) for the control signal input of the FVC switch U1c. 

The other input to the U1c FVC is the -100 mV/A current-sense signal from R1. This combo forces U1c to pump F * -0.1 V/amp * 500 pF = -2 kHz * Vload * 50 pC * Iload into the input of the A1b inverting integrator.

 The melodious result is:

Vw = R1 * Iload * 2000 * Vload * R6 * C6

or, 

Vw = Iload * Vload * 0.1 * 2000 * 1 MΩ * 500 pF = 100 mV/W.

The R6C5 = 100-ms integrator time constant provides >60-dB of ripple attenuation for Vload > 1-V and a low noise 0- to 5-V output suitable for consumption by a typical 8- to 10-bit resolution ADC input. Diode D1 provides fire insurance for U1 in case Vload gets shorted to ground.

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

• Voltage inverter design idea transmogrifies into a 1MHz VFC
• A simulated 100-MHz VFC
• A simple, accurate, and efficient charge pump voltage inverter for $1 (in singles)
• 100-MHz VFC with TBH current pump

The post Amps x Volts = Watts appeared first on EDN.

Walmart onn. coverage

Walmart’s onn. (or is it now just “onn”?) line of streaming media boxes and sticks are regularly represented here at Brian’s Brain, for several good reasons. They’re robustly featured, notably more economical than Google’s own Android TV-now-Google TV offerings, and frequently price-undershoot competitive devices from companies like Apple and Roku, too. Most recently, from a “box” standpoint, I took apart the company’s high-end onn. 4K Pro for publication at EDN in July, following up on the entry-level onn. 4K, which had appeared in April. And, within a subsequent August-published teardown of Google’s new TV Streamer 4K, I also alluded to an upcoming analysis of Walmart’s mid-tier onn. 4K Plus.

An intro to the onn.

That time is now. And “mid-tier” is subjective. Hold that thought until later in the write-up. For now, I’ll start with some stock shots:

Here’s how Walmart slots the “Plus” within its current portfolio of devices:

Note that, versus the Pro variant, at least in its final configuration, the remote control is not backlit this time. I was about to say that I guess we now know where the non-backlit remotes for the initial production run(s) of the Pro came from, although this one’s got the Free TV button, so it’s presumably a different variant from the other two, too (see what I did there?). Stand by.

And hey, how about a promo video too, while we’re at it?

Now for some real-life photos. Box shots first:

Is it wrong…

that I miss the prior packaging, even though there’s no longer a relevant loop on top of the box?

I digress. Onward:

Time to dive inside:

Inside is a two-level tray, with our patient (and its companion wall wart) on top, along with a sliver of literature:

Flip the top half over:

and the rest of the kit comes into view: a pair of AA batteries, an HDMI cable, and the aforementioned remote control:

Since I just “teased” the remote control, let’s focus on that first, as usual, accompanied by a 0.75″ (19.1 mm) diameter U.S. penny for size comparison purposes:

All looks the same as before so far, eh? Well then, last but not least, let’s look at the back:

Specifically, what does the product-code sticker say this time?

Yep, v2.32, different than the predecessors. Here’s the one in the baseline onn. 4K (v2.15, if you can’t read the tiny print):

And the two generations that ship(ped) with the 4K Pro, Initial (v2.26):

And subsequently, whose fuller feature set matched the from-the-beginning advertising (v2.30):

Skipping past the HDMI cable and AA battery set (you’re welcome), here’s the wall wart:

Complete with a “specs” close-up,”

whose connector, believe it or not, marks the third iteration within the same product generation: micro-USB for the baseline 4K model:

“Barrel” for the 4K Pro variant:

And this time, USB-C:

I would not want to be the person in charge of managing onn. product contents inventory…

Finally, our patient, first still adorned with its protective translucent, instructions-augmented plastic attire:

And now, stark nekkid. Top:

Front:

Bare left side:

Back: left-to-right are the reset switch, HDMI output, and USB-C input. Conceptually, you could seemingly tether the latter to an OTG (on-the-go) splitter, thereby enabling you to (for example) feed the device with both power and data coming from an external storage device, but in practice, it’s apparently hit-and-miss at best:

And equally bare right side:

There’s one usual externally visible adornment that we haven’t yet seen. Can you guess what it is before reading the following sentence?

Yes, clever-among-you, that’s right: it’s the status LED. Flip the device over and…there it be:

Now for closeups of the underside marking and (in the second) the aforementioned LED, which is still visible from the front of the device when illuminated because it’s on a beveled edge:

Enough of the teasing. Let’s get inside. For its similar-form-factor mainstream 4K precursor, I’d gone straight to the exposed circumference gap between the two halves. But I couldn’t resist a preparatory peek underneath the rubber feet that taunted me this time:

Nope. No screw heads here:

Opening it up

Back to Plan B:

There we go, with only a bit of collateral clip-snipped damage:

The inside of the bottom half of the case is bland, unless you’re into translucent LED windows:

The other half of the previous photo is much more interesting (at least to me):

Three more screws to go…

And the PCB then lifts right out of the enclosure’s remaining top half:

Allowing us to first-time see the PCB topside:

Here are those two PCB sides again, now standalone. Bottom:

and top:

Much as (and because) I know you want me to get to ripping the tops off those Faraday cages, I’ll show you some side shots first. Right:

Front; check out those Bluetooth and Wi-Fi antennae, reminiscent of the ones in the original 4K:

Left:

And back:

Let’s pop the top off the PCB bottom-side cage first:

Pretty easy; I managed that with just my fingernail and a few deft yanks:

At the bottom is the aforementioned LED:

And within the cage boundaries,

are two ICs of particular note; an 8 Gbit (1 GByte) Micron DDR4 SDRAM labeled as follows:

41R77
D8BPK

And, below these ICs are the nonvolatile memory counterpart, a FORESEE FEMDNN016G 16 GByte eMMC.

Now for the other (top) side. As you likely already noticed from the side shots, the total cage height here is notably thicker than that of its bottom-side counterpart. That’s because, unsurprisingly, there’s a heat sink stuck on top of it. Heat rises, after all; I already suspected, even before not finding the application processor inside the bottom-side cage, that we’d find it here instead.

My initial attempts at popping off the cage-plus-heatsink sandwich using traditional methods—first my fingernail, followed by a Jimmy—were for naught, threatening only to break my nail and bend the blade, as well as to damage the PCB alongside the cage base. I then peeked under the sticker attached to the top of the heatsink to see if it was screwed down in place. Nope:

Eventually, by jamming the Jimmy in between the heatsink and cage top, I overcame the recalcitrant adhesive that to that point had succeeded in keeping them together:

Now, the cage came off much more easily. In retrospect, it was the combined weight of the two pieces (predominantly the heatsink, a hefty chunk of metal) that had seemingly made my prior efforts be for naught:

At the bottom, straddling the two aforementioned antennae, is the same Fn-Link Technology 6252B-SRB wireless communications module that we’d found in the earlier 4K Pro teardown:

And inside the cage? Glad you asked:

To the left is the other 8 Gbit (1 GByte) Micron DDR4 SDRAM. And how did I know they’re both DDR4 in technology, by the way? That’s because it’s the interface generation that mates up with the IC on the right, the application processor, which is perhaps the most interesting twist in this design. It’s the Amlogic S905X5M, an upgrade to the S905X4 found in the 4K Pro. It features a faster Arm Cortex A-55 CPU quad-core cluster (2.5 GHz vs 2 GHz), which justifies the beefy heatsink, and an enhanced GPU core (Arm Mali-G310 v2 vs Arm Mali-G21 MP2).

The processing enhancements bear fruit when you look at the benchmark comparisons. Geekbench improvements for the onn. 4k Plus scales linearly with the CPU clock speed boost:

While GFXBench comparative results also factor in the graphics subsystem enhancements:

I’d be remiss if I didn’t also point out the pricing disparity between the two systems: the 4K Plus sells for $29.88 while the 4K Pro is normally priced $20 more than that ($49.88), although as I type these words, it’s promotion-priced at 10% off, $44.73. Folks primarily interested in gaming on Google TV platforms, whether out-of-the-box or post-jailbreaking, are understandably gravitating toward the cheaper, more computationally capable 4K Plus option.

That said, the 4K Pro also has 50% more DRAM and twice the storage, along with an integrated wired Ethernet connectivity option and other enhancements, leaving it the (potentially, at least) better platform for general-purpose streaming box applications, if price isn’t a predominant factor.

That wraps up what I’ve got for you today. I’ll keep the system disassembled for now in case readers have any additional parts-list or other internal details questions once the write-up is published. And then, keeping in mind the cosmetic-or-worse damage I did getting the heatsink and topside cage off, I’ll put it back together to determine whether its functionality was preserved. One way or another, I’ll report back the results in the comments. And speaking of which, I look forward to reading your thoughts there, as well.

—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

• Perusing Walmart’s onn. 4K Pro Streaming Device with Google TV: Storage aplenty
• Walmart’s onn. full HD streaming device: Still not thick, just don’t call it a stick
• Walmart’s onn. UHD streaming device: Android TV at a compelling price
• Walmart’s onn. FHD streaming stick: Still Android TV, but less thick
• Walmart’s onn. 4K streaming box: A Google TV upgrade doesn’t clobber its cost

The post Inside Walmart’s onn. 4K Plus: A streaming device with a hidden bonus appeared first on EDN.