árfolyamok

2021.07.29 11:00 UTC

BGN183,029Ft
CHF332,026Ft
CNY46,717Ft
CZK14,046Ft
DKK48,182Ft
EUR358,355Ft
GBP421,233Ft
HRK47,757Ft
JPY2,747Ft
NOK34,462Ft
PLN78,154Ft
RON72,806Ft
RSD3,030Ft
RUB4,125Ft
SEK35,193Ft
TRY35,500Ft
UAH11,272Ft
USD301,765Ft

hírek

Single chip synchronisation for 5G radio access equipment

2021-07-29 15:30 ElectronicsWeekly Steve Bush

Microchip has announced its first single chip frequency synchroniser for 5G packet switching, which needs ten times more accurately than 4G networks, it said. “Our ZL3073x/63x/64x network synchronisation platform implements measure, calibrate and tune capabilities, significantly reducing network equipment time error to meet the most stringent 5G requirements,” said the company’s v-p of timing components Rami ...

This story continues at Single chip synchronisation for 5G radio access equipment

Or just read more coverage at Electronics Weekly

>>>


Fable: The One-Trick Pony

2021-07-29 13:24 ElectronicsWeekly David Manners

A chip company founded 52 years ago had success with a programmable logic technology which used a matrix of circuit lines in which there was a fuse at every interconnection. When a fuse was blown by an electric charge it became an insulator. In this way the chip could be programmed to fit an application. ...

This story continues at Fable: The One-Trick Pony

Or just read more coverage at Electronics Weekly

>>>


Eval board for FTDI USB Power Delivery ICs

2021-07-29 11:51 ElectronicsWeekly Steve Bush

FTDI Chip has now announced a development board for USB Power Delivery, centred around its FT4233HP multi-channel interface ICs. The 138 x 77mm board has a pair of Type-C power delivery ports – one that can both sink (receive) and source (provide) power, and another that can only sink power. “The FT4233HP high-speed USB serial and ...

This story continues at Eval board for FTDI USB Power Delivery ICs

Or just read more coverage at Electronics Weekly

>>>


Power bipolars in DPAK

2021-07-29 11:02 ElectronicsWeekly Steve Bush

It is not often you see a new bipolar transistor introduced, and Nexperia has just revealed not just one, but nine of them. All in surface-mount DPAK (TO-252, SOT428C) packaging, they are intended for automotive (in AEC-Q101 qualified form) and industrial applications, and fill out the company’s MJD to span 2 – 8A and 45 – ...

This story continues at Power bipolars in DPAK

Or just read more coverage at Electronics Weekly

>>>


25-45W Intel Tiger Lake-H Xeon, Core, and Celeron embedded processors coming soon

2021-07-29 10:29 CNXSoft Jean-Luc Aufranc (CNXSoft)

While doing some research, I noticed an Intel Core i7-11850HE “Tiger Lake-H” processor on the...

>>>


28V-capable MCU includes USB Power Delivery 3.1

2021-07-29 10:21 ElectronicsWeekly Steve Bush

Infineon has introduced a high-voltage microcontroller with USB Power Delivery 3.1 support, claiming it to be an industry first. Branded EZ-PD PMG1-S3, it can be fed at up to 28V and is aimed at embedded systems consuming up to 140W. “The device supports higher power capabilities defined in the USB PD 3.1 specification and leverages ...

This story continues at 28V-capable MCU includes USB Power Delivery 3.1

Or just read more coverage at Electronics Weekly

>>>


ST expects 2021 revenues of $12.5bn

2021-07-29 06:25 ElectronicsWeekly David Manners

ST had Q2 net revenues of $2.99 billion for a gross margin of 40.5%, an operating margin of 16.3% and net income of $412 million. H1 net revenues were $6.01 billion with a gross margin of 39.7%, an operating margin of 15.5% and net income of $776 million. The Q3 forecast is for net revenues ...

This story continues at ST expects 2021 revenues of $12.5bn

Or just read more coverage at Electronics Weekly

>>>


Samsung Q2 profit up 73%

2021-07-29 06:11 ElectronicsWeekly David Manners

Samsung’s Q2 operating profit was up 54% y-o-y at $11 billion on revenues which were up 22% y-o-y at $55 billion. Net profit was up 73% at $8.4 billion. Q2 semiconductor revenues were $20 billion up 24.7% y-o-y for an operating profit up 27.6% y-o-y to $7 billion. “We expect that memory chip demand for ...

This story continues at Samsung Q2 profit up 73%

Or just read more coverage at Electronics Weekly

>>>


21% leap in IC unit shipments this year, predicts IC Insights

2021-07-29 05:30 ElectronicsWeekly David Manners

After a 6% drop in IC unit shipments in 2019, and an 8% increase in 2020, IC Insights forecasts a huge 21% jump in IC unit shipments this year. Unit shipments in 2021 are forecast to reach 391.2 billion, more than 11x the 34.1 billion units shipped over 30 years ago in 1990. The 2020-2025 ...

This story continues at 21% leap in IC unit shipments this year, predicts IC Insights

Or just read more coverage at Electronics Weekly

>>>


Quantum Computing: the video

2021-07-29 05:10 ElectronicsWeekly David Manners

Oxford Instruments has created a video about quantum computing with contributions from PsiQuantum, Intel, IBM, SeeQC and other leaders in the field. Called ‘Quantum Technology: Our Sustainable Future’ the video can be seen on YouTube here. Speakers are: · Alan Ho: Head of Product, Google Quantum AI · Alexandre Blais: Advisory Board, Q4Climate · Carl Williams: Deputy Director, Physical Measurement Laboratory at ...

This story continues at Quantum Computing: the video

Or just read more coverage at Electronics Weekly

>>>


binder takes M12 and M12 Power from the cabinet to the field

2021-07-29 05:10 ElectronicsWeekly Alun Williams

binder, the circular connector specialist, has announced M12 and M12 Power control cabinet feed-throughs, to help ensure a secure connection between modules in the cabinet and automation components in the field. Specifically, new control cabinet feed-through connectors will be available for the M12 product series 713, 813, 814, 823 and 824, which are available with ...

This story continues at binder takes M12 and M12 Power from the cabinet to the field

Or just read more coverage at Electronics Weekly

>>>


China and Europe dominate base station market

2021-07-29 05:09 ElectronicsWeekly David Manners

Chinese and European suppliers of base station equipment are expected to maintain a global market share of more than 70% in 2021, says TrendForce. The top three suppliers (along with their respective market shares) are: Huawei (30%), Ericsson (23%), and Nokia (20%). Remarkably, although Huawei remains banned by the US government, the company still manages ...

This story continues at China and Europe dominate base station market

Or just read more coverage at Electronics Weekly

>>>


Achieving full MCU partition isolation: Heaps

2021-07-29 05:01 Embedded.com Ralph Moore
Achieving full microcontroller (MCU) partition isolation is essential for achieving high security for MCU-based systems. Continuing a series on this complex topic, this article discusses the need for multiple heaps and the heap features that are useful in partitioned embedded systems.

This is the third article in a series on achieving high security for MCU-based systems. The first article covered security basics and partitions; the second article covered MPU management. In this article, we cover the need for multiple heaps and the heap features that are useful in partitioned embedded systems. The right heap is important for achieving full MCU partition isolation. This is not intended as a tutorial. Ref. 2 may be helpful for that.

There was a day when embedded systems were so simple that heaps were seldom used. However, since then, complexity has grown so much that most embedded systems, even RTOSs, now use heaps. If only one heap is in use, heap efficiency is not of much concern. However, in typical partitioned systems, there may be several heaps. Multi-heap support becomes necessary because using a common heap between partitions breaks partition isolation. A hacker can wreck a common heap and bring all partitions using it down or possibly invade them through the common heap. In a multi-heap situation, heap efficiency and features become important. Below, various heap features are discussed that are useful in fully partitioned embedded systems.

Doubly Linked Chunks

Heaps consist of linked chunks. In each chunk is metadata used by the heap and the data block returned to the user. In dlmalloc (see Ref. 1) and many of its derivatives, the block size is at the start of each chunk and also at the end of each free chunk. This achieves a very high memory efficiency – only 8 bytes per inuse chunk. However, it is only possible to trace forward (by adding sizes to the heap start), not backward through the heap.

A better design for embedded systems is a forward link and a backward link in every chunk. This permits continual forward heap tracing during idle periods to find broken links, and when a broken link is found, backward tracing to fix it. (Backward tracing is also necessary to check each link when tracing forward.) Heaps contain vital information such as task stacks, memory protection arrays, and other system and application control structures. Hence, heap self-testing and self-healing is important for embedded systems exposed to background radiation, electrical disturbances, excess heat, and other environmental phenomena encountered by many embedded systems — not to mention hackers.

Heap Bins

dlmalloc was one of the first heaps to use heap bins, and most other heaps developed since also use them. A heap bin consists of doubly linked free chunks, in a certain size range, linked to the bin header. Normally the smallest-size bins have only one size and are organized by successive chunk sizes (e.g. 24, 32, …, 56) up to a maximum size. This is referred to as a Small Bin Array (SBA). Allocating or freeing an SBA chunk is very fast – the size is used as an index into the SBA, and the first chunk is taken or returned. Above the SBA are the upper bins [1], which consist of a mixture of small bins and large bins. Small bins have one size; large bins have a range of sizes (e.g. 128 to 248, 256 to 504, etc.). The last upper bin is called the top bin, and it has all of the remaining sizes.

Accessing the correct upper bin requires size comparisons, then searching the bin for the best match, and if a big-enough chunk is not found, taking the first chunk in the next larger bin. This can be time-consuming, but not as bad as searching a serial heap with no bins. Each upper bin is, in effect, a sub heap.

The problem with dlmalloc and its variants is that the number of bins and trees is fixed. This may be ok for the main heap, but it is not a good match for partition heaps. A given partition may use only a few block sizes, hence it needs only a small SBA, if any, and only a few large block sizes. Thus a configurable bin structure is highly desirable for partition heaps, both for memory efficiency and for performance. For example, the following bin structure might work well for a network partition:

u32 const binsz2[] =
/*bin  0    1     2     3     4    end */
      {24, 512, 1024, 1526, 1534,  -1};

In this case, there is no SBA. The first bin covers all chunk sizes from 24 to 504 [2]. The next two bins cover all chunk sizes up to 1518. Bin 3 has a single chunk size of 1526, which can hold the maximum Ethernet frame size of 1518 bytes. Like an SBA bin, bin 3 requires no searching – the first chunk is taken or placed. Bin 4 is the top bin and it contains all chunk sizes from 1534 on up.

An even simpler partition might contain only one bin:

u32 const binsz1[] =
/*bin  0   end */ 
      {24, -1};

This could be a partition that allocates mostly permanent blocks. In this case, the chunks would come primarily from the top chunk [3] and seldom from bin0.

Large Bin Sorting

Generally, embedded systems have significant idle time in order to be able to handle peak loads. Large bin sorting can be done during idle times. If a large bin is sorted, the first big-enough chunk is also the best-fit chunk in the bin. Assuming the rule is to always take the first big-enough chunk, sorting reduces unnecessary chunk splitting and saves bigger chunks for larger requests. (Note: Finding the best-fit chunk in an unsorted bin usually takes too much time.)

Merge Control

When a chunk is freed it may be merged with adjacent free chunks. dlmalloc has no merge control – freed chunks are always merged with adjacent free chunks. This tends to defeat the purpose of bins – it results in removing a free chunk from a lower bin, merging it with the freed chunk, and placing the resulting chunk into a higher bin. The next time this chunk size is needed, its bin may be empty and it may be necessary to get a larger chunk and split it. Hence two unnecessary operations have occurred: merge and split. This hurts performance.

A better policy for embedded systems is not to merge until a certain upper threshold is reached, such as HEAP_USE_MAX, and then merge all freed blocks until a lower threshold is reached, such as HEAP_USE_MIN. This operates like a thermostat. Other criteria could be used, such as number of free blocks, total free block bytes, etc. In defense of dlmalloc, research has shown (See Ref. 2) that there is no universal solution to avoiding allocation failures due to fragmentation. Hence, merge control may be dangerous and should be turned off. However, embedded systems, particularly individual partitions, tend to have very regular behaviors and merge control probably will not cause allocation failures for them.

Heap Recovery

In the event of an allocation failure, a heap recovery function is automatically called. It traces through the heap to find enough adjacent free blocks to satisfy the request, merges them, and returns control to the allocation function. Thus the heap hiccups but does not fail.

Aligned Blocks

Dynamically allocated protected blocks and protected messages must be aligned on power-of-two boundaries in order to be used as MPU regions. An efficient process for doing this is as follows:

  1. Find the first large-enough free chunk for the desired block size, sz.
  2. Find the first alignment boundary, 2n, inside the chunk’s data block.
  3. Test if the remainder of the chunk is >= sz.
  4. If not, go on to the next large-enough chunk.
  5. When an acceptable chunk is found, put its Inuse Chunk Control Block (ICCB) below the boundary – i.e. just below the aligned data block.
  6. The resulting space below the ICCB is called free space, and it is handled as follows:
    1. If the preceding chunk is free, combine the free space with it.
    2. Else, if the free space is large enough, make it into a free chunk.
    3. Else, combine the free space with free space at the end of the preceding inuse chunk, and if the result is big enough, make it into a new free chunk.
  7. Split off space after the block, if large enough for a free chunk, else make it free space.

This process is illustrated in Figure 1. In this figure, ICCB = Inuse CCB, FCCB = Free CCB. In this case, option 6b has been taken and a small free chunk has been formed below the new data block. Some spare space is left at the top of this chunk because the space is not large enough to form a free chunk. Over a period of time, the heap will start to be organized into chunks having aligned data blocks, and aligned allocations will become faster.


Figure 1: Aligned Allocation. (Source: Author)

Region Blocks

The foregoing is adequate for v8M MPU regions when specifying alignment and size as multiples of 32, but v7M MPUs require additional steps:

  1. Determine the region size as the next larger power of two. For example, if sz = 630, then region size = 1024.
  2. Determine the subregion size and the number of contiguous subregions needed. In the example, subregion size = 128, and 5*128 = 640 > 630, so N = 5.
  3. Do aligned search steps 1 – 3 with alignment = 128 and size = 640.
  4. After step 3: Verify that all N subregions are in the same region – i.e. find the next region boundary (e.g. multiple of 1024) and verify that the last subregion ends before it.
  5. Do steps 5 -7.

This results in a subregion-aligned v7M region block that is contained in contiguous subregions within a region, as shown in Figure 2:


Figure 2: Region Allocation. (Source: Author)

In this example, subregions 1-5 will be enabled, and subregions 0, 6, and 7 will be disabled. Note how the disables protect the surrounding heap CCBs and spare space. Because the region block is only subregion-aligned, it is much easier to find and causes less heap disruption.

Chunk Types.

Figure 3 shows three types of chunks, each with a different Chunk Control Block (CCB) (Orange). All three CCBs have a forward link to the next CCB and a backward link to the previous CCB. In addition, the free CCB has chunk size, forward and backward bin links, and bin number. This requires 24 bytes, so that is the minimum chunk size. The inuse CCB requires 8 bytes for links, so the smallest data block is 24 – 8 = 16 bytes.

The third type of chunk is a debug chunk. The debug CCB is an inuse CCB with chunk size, time of allocation, owner, and a fence added. In addition the data block has N fences above and below it. These things are useful during debugging to find memory leaks, block overflows, and other problems. Whether an inuse chunk or a debug chunk is generated by an allocation depends upon whether the heap’s debug mode is off or on, respectively. This permits limiting debug chunks to code of interest, which is useful since they can be much larger than inuse chunks.


Figure 3: Chunk Types. (Source: Author)

Integrated Block Pools

16 bytes is rather large for many C++ objects. To alleviate this problem, smaller block pools can be integrated with the heap. Then allocations of less than 16 bytes are taken from block pools. This is very fast and commensurate with the needs of object-oriented code. A useful by-product of integrating block pools into a heap is that if a block pool runs out of blocks, the heap is used instead. This may result in slower performance, but the system does not break. When freed, blocks go back to their source.

It is likely in modern embedded systems that some partitions (especially third party software) will be written exclusively in C++. A heap with integrated block pools may result in much better performance for the partition and may use less memory.

Need for Mutexes

All RTOSs use some mechanism to protect critical sections of system service routines. Whatever method is used, the net result is that no other task can run during these critical sections. This approach is not workable for multiple heaps that share the same heap code. Instead, we must use a mutex per heap in order to limit one task to access a heap, while still allowing higher priority tasks to preempt and access other heaps.

Conclusion

As can be seen from the foregoing, the heap choice is an important part of partitioning a system. The right heap can improve performance and reduce code rewriting. Providing dynamic regions is also an important heap requirement. In the next article in this series, we will look into partition portals to see how they work and how they are used.

References

  1. Doug Lea, A Memory Allocator, December 1996.
  2. Paul R. Wilson, et al. Dynamic Storage Allocation: A Survey and Critical Review, September 1995.

Footnotes

[1] This discussion is not applicable to dlmalloc, which uses trees; trees are much more complicated than large bins.

[2] Keep in mind that block sizes are 8 bytes less than chunk sizes, and that chunk sizes are spaced 8 bytes apart.

[3] A heap starts out as a donor chunk for the SBA and a top chunk for upper bin chunk sizes. If there is no SBA, there is no donor chunk and the whole heap starts out as a single top chunk.

Ralph Moore is a graduate of Caltech. He and a partner started Micro Digital Inc. in 1975 as one of the first microprocessor design services. In 1989 Ralph decided to get into the RTOS business and he architected the smx RTOS kernel. After 20 years of selling Micro Digital products and managing the business, he went back into product development. Currently he does the whole job from product definition, architecture, design, coding, debug, documentation, patenting, to promotion. Recent products include eheap, SecureSMX, and FRPort. Ralph has three children and six grandchildren and lives in Southern California..

Related Contents:

For more Embedded, subscribe to Embedded’s weekly email newsletter.

The post Achieving full MCU partition isolation: Heaps appeared first on Embedded.com.

>>>


NuCurrent showcases NFC charging at 3W, data rates up to 848 kb/s

2021-07-29 04:57 CNXSoft Jean-Luc Aufranc (CNXSoft)

The NFC Forum approved the Global Wireless Charging (WLC) specification last year with the new...

>>>


Upcoming events about RISC-V, RT-Thread IoT OS, and Embedded Linux

2021-07-29 03:50 CNXSoft Jean-Luc Aufranc (CNXSoft)

Three events about open/open-source technologies have been recently announced with namely the RT-Thread IoT OS...

>>>


Epoxy’s diverse role in IoT device PCB manufacturing

2021-07-29 00:59 Embedded.com Emily Newton
IoT device makers can apply epoxies at numerous design and manufacturing phases to fulfill specific requirements or needs.

The Internet of Things (IoT) market is booming. That success pushes engineers to explore practical solutions to improve the printed circuit boards (PCBs) that become integral parts of today’s IoT gadgets.

Epoxy is a material that serves various functions during the PCB manufacturing process for IoT products. Here’s more about the vital role it plays in IoT manufacturing.

Tuned to Meet Specific Requirements

Manufacturers can either choose specialty epoxies or alter specific epoxy properties to meet particular performance or manufacturing needs. For example, additives can make an epoxy harder or thicker, making it maximally suitable as a conformal coating. Here are some other ways to tune particular epoxy properties.

Electrical and Thermal Conductivity

Using silver as a filler for one- or two-part epoxy can create an electrically conductive adhesive to replace soldering. Electrically conductive adhesives are either isotropic or anisotropic. Those in the first category are electrically conductive in all directions. However, anisotropic adhesives conduct electricity in only one direction. They’re sometimes used to bond antenna structures in radio frequency identification (RFID) products.

Epoxies also aid with thermal conductivity. One option is to use such adhesives to join two surfaces and transfer heat to the cooler one. However, since most epoxies lack adequate intrinsic thermal management capabilities, fillers make up the deficit. Powders such as copper, boron nitride and aluminum significantly increase heat transfer properties.

Extreme Temperature Tolerance

Additives and hardeners also get blended into epoxies during and before curing to make the adhesives resistant to cryogenic temperatures. Conversely, epoxies exist that withstand temperatures warmer than the approximately 300 degrees Fahrenheit that non-extreme-heat-tolerant types can.

Low Outgassing

Epoxies used in the aerospace industry must be low-outgassing types. Outgassing causes the release of volatile compounds around a spacecraft due to space’s vacuum.

NASA uses two testing parameters to ensure epoxies meet outgassing requirements: total mass loss (TML) and collected volatile condensable materials (CVCM). More specifically, NASA’s standards dictate that an epoxy adhesive or potting compound has a TML of less than 1% and a CVCM of less than 0.1%.

Companies that offer low-outgassing, high-purity epoxies first test those products under strict conditions in specialized chambers. They then publicize the results, catering to customers requiring low-outgassing adhesives.

Coefficients of Thermal Expansion (CTEs)

Most materials experience thermal expansion due to the increase in the energy of molecular interactions due to temperature changes. CTE expresses how much change occurs with each one-degree temperature rise.

CTE mismatches can occur between two substrates or between an adhesive and a substrate. Thus, a common approach is to select adhesives with as low a CTE as possible. Another option is to insert specialty negative CTE fillers or ceramics into unfilled adhesives. However, doing that causes a significant increase in the tensile modulus, making the epoxy stiffer.

Glass Transition (Tg) Temperature

Epoxy’s glass transition (Tg) temperature is a range over which it goes from a rigid, glass-like consistency to a softer, more rubbery one. It can span from approximately 50-250 degrees Celsius. However, the choice of epoxy, the fillers used and the cure time can all affect Tg.

Epoxies with a Tg of more than 150 degrees Celsius typically have superior high-temperature resistance. However, types with a Tg in the 120-130 Celsius range provide excellent chemical resistance properties.

Proper Adhesion to Various Substrates

Epoxy adhesives bond to and seal a wide variety of substrates, ranging from metals and most plastics to wood and concrete. However, there are a few unsuitable materials, such as low-surface energy plastics, including polyolefins, silicones and fluorocarbons. Moving ahead with decisions to use epoxies on those materials requires pretreating them to change the substrate’s surface.

Cure Time and Storage Requirements

Epoxy adhesives are available as one- and — more commonly — two-component formulations. One-component options typically come as pastes and require people to apply them with trowels to fill gaps. These epoxies require heat to cure, as well as cold storage to maintain their shelf lives.

Two-component types necessitate mixing and using the products within a specific timeframe that could range from a few minutes to several hours. These epoxies cure at slightly warmer than room temperature (approximately 75-85 degrees Fahrenheit), although more heat expedites the process.

Two-component epoxies also have less stringent storage requirements compared to one-component types. Manufacturers may keep these specifics in mind while choosing epoxies that align with their production requirements.

Viscosity

Centipose (CPS) is a viscosity value applied to epoxies to indicate how fast it flows. A low-CPS epoxy flows quickly, while the flow rate slows as the CPS rises. An epoxy’s viscosity dictates its potential use cases and the methods of applying the products.

Reduced viscosity also helps reduce voids. Many manufacturers sell epoxies in a wide range of viscosities, such as from 100-1,500,000 CPS. However, heat also affects viscosity, and exposure to it thins an epoxy’s consistency.

Low-viscosity epoxies may take 12-24 hours to cure — longer than their high-viscosity counterparts. High-viscosity epoxies suit surface-coating applications. However, processing them requires not exceeding the maximum thickness specified by the manufacturer, which is often 1-2 centimeters.

Used as a Material Throughout the PCB

Engineers frequently work with epoxies while developing PCBs. Specific epoxies behave in different ways, and engineering professionals must typically use whatever a manufacturer provides.

However, knowing about a particular epoxy’s functions helps the design project go smoothly. While some have adhesive properties, others offer thermal conductivity. A mismatch between an adhesive compound’s characteristics and the product’s materials could lead to issues affecting manufacturing or usability once a product reaches the market.

For example, a circuit board’s prepregs are often made from a semi-cured glass-epoxy material. Prepregs are dielectric materials with binding and insulating properties. A PCB’s internal core usually features fully cured glass-epoxy material with copper laminated to both sides.

Moreover, companies have started combining epoxy with other substances during PCB manufacturing in an ongoing effort to cut costs associated with dielectric materials. One common practice is to use it with polyphenylene oxide (PPO) or polyphenyl ether (PPE), which are thermoplastics.

Using PPO without epoxy typically increases the overall manufacturing costs. However, relying on it reduces expenses while still meeting performance requirements.

You can get an idea of the multiple uses of epoxy on PCB components for an IoT device with the example of a newly developed implanted blood-oxygen sensor. This advanced product bonded a piezoelectric crystal with conductive silver epoxy, then attached it to a PCB. The developers also used ultraviolet-curable epoxy to surround the wire-bonded areas within the PCB.

Chosen to Improve Heat Transfer

As mentioned earlier, specific epoxies have different characteristics. Thermal management is a significant concern for most companies that design and manufacture IoT devices. Excessive temperatures can damage delicate electronics and cause gadgets to malfunction. Some engineers have developed ways to make IoT devices benefit from warmth, such as body heat. However, the goal is typically to avoid hot spots and overall overheating.

The need to control heat becomes even more crucial as IoT devices get smaller. Traditional methods include using fans and heat sinks. Another option is to apply thermal greases between the components that give off heat or have cooling capabilities. People can also get the desired results by using specific kinds of epoxies.

For example, one- and two-component epoxies enhance the heat transfer across interfaces. People may also choose them to complement other heat dissipation methods, such as using epoxy to bond a heat sink to a PCB.

When people discuss how quickly heat dissipation happens with certain epoxies, they refer to the substances’ conductivity. If an epoxy has a heat conductivity value of 0.3-0.4 watts per milli-Kelvin, that means the warmth dissipates comparatively slowly. However, values of 1.7-2 watts per milli-Kelvin indicate quicker heat conductivity.

However, the Tg is another aspect to consider when using epoxies for thermal management during PCB manufacturing. Any epoxies used must have compatibility with the Tg of the accompanying substrates.

Selected as a Conformal Coating

When companies engage in IoT manufacturing, representatives must consider the likely environmental characteristics the gadget will get exposed to during normal use. For example, some IoT devices get placed outdoors in dusty or moist environments. In other cases, the IoT products perform constant monitoring in remote areas and are not frequently checked by humans.

Thus, it’s vital to build the PCBs for IoT devices to withstand potentially harsh elements. One common way to do that is to apply conformal coatings. Epoxy that is used this way is both hard and opaque, providing good protection against chemicals, abrasion and moisture. Epoxy conformal coatings are also wise choices for IoT devices exposed to high humidity.

Conformal coatings are extremely thin yet protective. They add a safeguarding layer directly on top of PCB components without thickness that would add undesirable bulk. Since conformal coatings also extend a PCB’s lifespan, they’re an easy way for an IoT device manufacturer to provide the prolonged performance a client expects.

Similarly, conformal coatings can reduce expensive repair costs that could cut into a manufacturer’s profits. PCBs breaking prematurely inside of IoT products could also damage the maker’s reputation. Choosing to apply conformal coatings during PCB manufacturing is a relatively straightforward way to prolong functionality, thereby keeping customers happy.

Applied to Discourage Reverse Engineering

Reverse engineering occurs when someone — often a competitor — attempts to determine how a manufacturer produced an item. It’s a risk in numerous industries and applies to chemical and biological processes, as well as physical products.

Numerous preventive measures exist to protect against reverse engineering. For example, some manufacturers place sensors within the PCB to detect and prevent such attempts. However, a less involved but still effective technique is to practice potting.

It involves using a shell or similar layer to completely encase a PCB or another electronic component. People pour a compound into that case area, which hardens and becomes part of the PCB. Epoxy resin is a commonly chosen potting compound. Its opacity stops people from learning visual details that help them understand more about the design.

Some potting compounds are also non-removable. That’s a good thing when it comes to protecting against design copying. However, it also could make it difficult or impossible for an authorized person to repair a PCB.

Depending on the project at hand, engineers may also use silicones for encapsulation rather than epoxies. In addition to maintaining their mechanical properties across a wide temperature range, silicones are soft and flexible, making them appropriate for covering sensitive electronics.

Potting is usually selected along with several other measures that stop people from reverse engineering a PCB design. Thus, manufacturers must determine which options provide optimal protection and consider whether they may need to remove the potting compound later.

Epoxy Helps IoT Manufacturing Progress

These examples show that IoT device makers can apply epoxies at numerous design and manufacturing phases to fulfill certain requirements or needs. As IoT devices continue rising in popularity and becoming even more widespread, epoxy will keep being a critical part of PCB manufacturing.

Emily Newton is a technology and industrial journalist who enjoys discovering how the IoT is impacting different industries. Emily is editor in chief of Revolutionized – an online magazine exploring trends in science, technology and industry. Subscribe to her newsletter to keep up with the latest.

Related Contents:

For more Embedded, subscribe to Embedded’s weekly email newsletter.

The post Epoxy’s diverse role in IoT device PCB manufacturing appeared first on Embedded.com.

>>>


Arm M&A Nonsense

2021-07-29 00:00 ElectronicsWeekly David Manners

A lot of nonsense is being talked about the Nvidia bid for Arm. It revolves around whether an IPO would be better for Arm than being bought by Nvidia. Everyone involved seems to agree that Arm’s open licensing model, with all customers being treated equally, should be preserved. Which of the two options of either ...

This story continues at Arm M&A Nonsense

Or just read more coverage at Electronics Weekly

>>>


OKdo program will recycle your old Pi and give you £10 voucher

2021-07-28 19:35 LinuxGizmos Eric Brown

OKdo, Sony, and the Raspberry Pi Foundation have launched an “OKdo Renew” recycling program that will exchange your old RPi 3 and 4 boards and give you a £10 OKdo voucher. Meanwhile, RPi’s Eben Upton says the next Pi will likely be a RPi 4A due in 2022. The tech industry likes to talk a […]

>>>


Top Ten (less 5) Base Station Vendors

2021-07-28 18:20 ElectronicsWeekly David Manners

Thanks to TrendForce for this one – the top five base station vendors:

This story continues at Top Ten (less 5) Base Station Vendors

Or just read more coverage at Electronics Weekly

>>>


RPi 4 based mobile bot features optional 6-DOF arm

2021-07-28 15:57 LinuxGizmos Eric Brown

Elephant Robotics’ $700 to $1,200 “MyAGV” mobile robot runs on a Raspberry Pi 4B with a 360° lidar, 5MP cam, a 220 rpm/min motor, 4x omnidirectional wheels, and an optional 6-axis MyCobot arm. In May, Elephant Robotics launched a $699, six-axis manipulation bot called the MyCobot Pi, which is built around a Raspberry Pi 4 […]

>>>


Take it away… Gadget and Music intersectionality

2021-07-28 15:29 ElectronicsWeekly Alun Williams

This is just a bit of fun - a way to take a different angle on some of the Gadget Master posts we've featured over recent years.

This story continues at Take it away… Gadget and Music intersectionality

Or just read more coverage at Electronics Weekly

>>>


Re-use and recycle Raspberry Pis

2021-07-28 15:09 ElectronicsWeekly Steve Bush

Distributor OKdo has partnered with Raspberry Pi and the Sony Technology Centre in a scheme to re-use certain Raspberry Pi boards in the UK, with plans for it to become global later this year. OKdo Renew, as it will be know, is open only for later Rapberry Pis: 3b, 3b+ and 4. “In return, they ...

This story continues at Re-use and recycle Raspberry Pis

Or just read more coverage at Electronics Weekly

>>>


PoE power injector works up to 5Gbit/s

2021-07-28 14:50 ElectronicsWeekly Steve Bush

Advantech has a DIN-rail power-over-Ethernet injector for industrial grade Wi-Fi 6 infrastructure. Called EKI-2701MPI-5G, it supports data rates of 10M, 100M, 1G, 2.5G and 5G, and IEEE802.3bt 90W output. It is”an excellent choice for obsolete infrastructure applications requiring heavy network connectivity,” said the company. It further serves as a PoE PSE – delivering up to 90W ...

This story continues at PoE power injector works up to 5Gbit/s

Or just read more coverage at Electronics Weekly

>>>


Disinfection LEDs from Anglia

2021-07-28 14:13 ElectronicsWeekly Steve Bush

Anglia Components has begun stocking UV-C ultra-violet LEDs from Bolb. “UV-C LEDs are gaining attention because they provide robust, toxic-free disinfection,” said Anglia marketing director John Bowman. “The addition of Bolb’s products to Anglia’s line up is an important step as we look to offer customers solutions for sterilisation that will play an important role ...

This story continues at Disinfection LEDs from Anglia

Or just read more coverage at Electronics Weekly

>>>


2MW Orkney tidal turbine hooks into the grid

2021-07-28 14:00 ElectronicsWeekly Steve Bush

Orbital Marine Power’s O2 has begun feeding power into the national grid from the European Marine Energy Centre in Orkney. The floating turbine, claimed by Orbital to be the worlds largest, has been fabricated using ship-building technology in Dundee, and is anchored in the Fall of Warness. 74m long O2 is Orbital’s first commercial turbine, ...

This story continues at 2MW Orkney tidal turbine hooks into the grid

Or just read more coverage at Electronics Weekly

>>>


Made-up BNC cables from Amphenol

2021-07-28 13:46 ElectronicsWeekly Steve Bush

Amphenol RF is aiming at test, measurement and industrial applications with a line of BNC cable assemblies using low-loss Times LMR‐UF cable. LMR cable is designed with improved shielding compared to standard RF coaxial cables and is intended to achieve low attenuation at high frequencies – bandwidth is up to 3GHz. The 50Ω ohm assemblies are ...

This story continues at Made-up BNC cables from Amphenol

Or just read more coverage at Electronics Weekly

>>>


The Contagion Games

2021-07-28 13:15 ElectronicsWeekly David Manners

Tokyo is facing a fifth coronavirus wave and will be in a state of emergency for the entirety of the Olympic Games. The seven-day average of new cases has more than doubled over the past two weeks, and with the Delta variant taking hold and just 22 percent of the population fully vaccinated, health experts ...

This story continues at The Contagion Games

Or just read more coverage at Electronics Weekly

>>>


The first good folding 3D printer I have seen

2021-07-28 13:12 ElectronicsWeekly Steve Bush

Lots of people have attempted to design and build fold-up 3d printers, many of which turn out quite compromised. Now I have seen the first that looks properly competent, and erects into something that prints well – I am guessing at least as well as ~£500 printers. It is a one-off by Kralyn3D, called the Positron, ...

This story continues at The first good folding 3D printer I have seen

Or just read more coverage at Electronics Weekly

>>>


£1.5m for UK silicon anode li-ion project

2021-07-28 11:06 ElectronicsWeekly Steve Bush

Innovate UK has approved funding for the £1.5m ‘Silicon anode battery for rapid electrification’ (Sabre) project, intended to improve energy density in electric vehicle grade lithium ion batteries. The money, via Faraday Battery Challenge at UK Research and Innovation, will go to Nexeon, Britishvolt and University College London (UCL) to deliver Li-ion test cells with a silicon ...

This story continues at £1.5m for UK silicon anode li-ion project

Or just read more coverage at Electronics Weekly

>>>


Voice-control of embedded vision AI for IoT and the edge

2021-07-28 10:37 ElectronicsWeekly Steve Bush

Renesas is developing voice-controlled AI for contactless image processing in IoT and edge systems, giving self-checkout machines, security cameras, video conference systems and smart appliances such as robotic cleaners as example applications. To do this, it has teamed up with Californian deep learning chip company Syntiant, and is combining Renesas’ RZ/V vision AI MCUs with ...

This story continues at Voice-control of embedded vision AI for IoT and the edge

Or just read more coverage at Electronics Weekly

>>>


MediaTek Kompanio 1300T processor is made for 5G tablets

2021-07-28 09:40 CNXSoft Jean-Luc Aufranc (CNXSoft)

MediaTek has launched several 5G mobile SoCs for smartphones in recent years, but with Kompanio...

>>>


Gadget Book: Understanding Infrastructure Edge Computing

2021-07-28 08:38 ElectronicsWeekly Alun Williams

This book may interest Gadget Masters who wanted to better understand edge computing and the infrastructures that can underlie it – it’s full title is: Understanding Infrastructure Edge Computing: Concepts, Technologies, and Considerations. Written by Alex Marcham – who apparently coined the term ‘infrastructure edge computing’ – it’s intended to provide an accessible view of ...

This story continues at Gadget Book: Understanding Infrastructure Edge Computing

Or just read more coverage at Electronics Weekly

>>>


Test paper for disaggregated O-RAN components

2021-07-28 08:20 ElectronicsWeekly David Manners

Small Cell Forum (SCF) has released a paper outlining the key test configurations for 3GPP TS38.141 compliance testing of split 6 products using the SCF nFAPI 2.0 interface. Led by Forum members Keysight Technologies and Picocom, the paper aims to inform operators and integrators on how split 6 products can be tested, with specifications derived ...

This story continues at Test paper for disaggregated O-RAN components

Or just read more coverage at Electronics Weekly

>>>


PsiQuantum raises $450m Series D to build million qubit quantum computer

2021-07-28 05:35 ElectronicsWeekly David Manners

PsiQuantum has raised $450 million in Series D funding to build a quantum computer with one million qubits – the level at which quantum computers become commercially viable. The funding round was led by BlackRock with participation from Baillie Gifford and M12 (Microsoft’s venture fund), Blackbird Ventures and Temasek. PsiQuantum has now raised a total of ...

This story continues at PsiQuantum raises $450m Series D to build million qubit quantum computer

Or just read more coverage at Electronics Weekly

>>>


Blaize raises $71m Series D

2021-07-28 05:27 ElectronicsWeekly David Manners

Blaize, the AI processor specialist, has raised a $71m Series D funding round. Franklin Templeton, a new investor, and Temasek, an existing investor, led the round, along with participation from DENSO and other new and existing investors. “With substantial power advantages making EVs more efficient and economical,Blaize SoCs offer best in class performance with lowerpower ...

This story continues at Blaize raises $71m Series D

Or just read more coverage at Electronics Weekly

>>>


Ofcom consults on licensing changes for NGSO constellations such as OneWeb

2021-07-28 05:10 ElectronicsWeekly Alun Williams

Ofcom is proposing rule changes around its regulation of satellite constellations operating in non-geostationary orbits (NGSO), such as those of Amazon, SpaceX, Telesat and the UK government-owned OneWeb. It has announced a consultation period ahead of new proposals for its licensing of the NGSO systems. In order to ensure the quality of such satellite broadband ...

This story continues at Ofcom consults on licensing changes for NGSO constellations such as OneWeb

Or just read more coverage at Electronics Weekly

>>>


Boardcon EM3566 Linux SBC is powered by a Rockchip RK3566 System-on-Module

2021-07-28 05:01 CNXSoft Jean-Luc Aufranc (CNXSoft)

Another day, another platform based on Rockchip RK3566 quad-core Cortex-A55 AIoT processor with Boardcon EM3566...

>>>


Element 14 eBook covers IoT security trends

2021-07-28 05:01 ElectronicsWeekly David Manners

Element14 has released a new eBook for its members providing an overview of IoT Security trends, relevant algorithms and cryptography for security use, along with a rundown of the latest in communication, edge device, gateway server, and cloud level security. The IoT allows millions of servers, devices, and people across the world to communicate, share ...

This story continues at Element 14 eBook covers IoT security trends

Or just read more coverage at Electronics Weekly

>>>


Arrow signs ERNI

2021-07-28 05:00 ElectronicsWeekly David Manners

ERNI, the connector specialist, has signed a distribution agreement with Arrow. Arrow will offer ERNI’s full range of electronic connectors. With the agreement, ERNI aims at new customers who turn to Arrow for connectors. ERNI’s product portfolio ranges from SMCs (small multiple connectors) with more than 300 million units in use world-wide to the most ...

This story continues at Arrow signs ERNI

Or just read more coverage at Electronics Weekly

>>>


ESP32-H2 Bluetooth LE & 802.15.4 RISC-V SoC shows up in ESP-IDF source code

2021-07-28 03:09 CNXSoft Jean-Luc Aufranc (CNXSoft)

Espressif Systems is working on yet another RISC-V chip with ESP32-H2 SoC offering Bluetooth LE...

>>>


Top Ten Analogue Companies

2021-07-28 00:06 ElectronicsWeekly David Manners

Thanks to IC Insights for this – the top ten analogue companies in 2020:

This story continues at Top Ten Analogue Companies

Or just read more coverage at Electronics Weekly

>>>


16Mpixel image sensor for factory automation

2021-07-27 16:20 ElectronicsWeekly Steve Bush

XGS 16000 is a 16Mpixel global shutter image sensor for factory automation by On Semiconductor. Consuming 1W at 65frame/s, it fits into standard 29 x 29mm industrial cameras with C-Mount lenses – it has a 1:1 square aspect ratio. It also shares a common architecture and footprint with other XGS CMOS image sensors, enabling users to ...

This story continues at 16Mpixel image sensor for factory automation

Or just read more coverage at Electronics Weekly

>>>


Full-featured RK3566 SBC ships with custom Linux stack

2021-07-27 15:22 LinuxGizmos Eric Brown

Boardcon’s “EM3566” SBC runs Linux on Rockchip’s RK3566 via a “CM3566” module with up to 8GB RAM and 32GB eMMC. Features include WiFi/BT, GbE, HDMI, MIPI-DSI and -CSI, 6x USB, SATA, M.2 for NVMe, and mini-PCIe with SIM. Boardcon has announced its first single board computer based on the Rockchip RK3566, following earlier Rockchip-based models […]

>>>


ST’s first 200mm in-house SIC wafers

2021-07-27 14:20 ElectronicsWeekly Steve Bush

STMicroelectronics has manufactured its first 200mm silicon-carbide bulk wafers at its Norrköping Sweden facility. “The transition to 200mm SiC wafers will bring substantial advantages to our automotive and industrial customers as they accelerate the transition towards electrification,” said ST president of automotive Marco Monti. The company is claiming high quality with minimal crystal-dislocation defects through expertise ...

This story continues at ST’s first 200mm in-house SIC wafers

Or just read more coverage at Electronics Weekly

>>>


Microchip’s 1,700V SiC mosfets

2021-07-27 13:41 ElectronicsWeekly Steve Bush

Microchip has added 1,700V mosfets to its silicon carbide portfolio, as die, discretes and in modules. Features include gate oxide stability, according to the company, which “observed no shift in threshold voltage even after 100,000 pulses in repetitive un-clamped inductive switching [R-UIS] tests. R-UIS tests also indicated avalanche ruggedness and stability in other parameters. They also ...

This story continues at Microchip’s 1,700V SiC mosfets

Or just read more coverage at Electronics Weekly

>>>


Defining a standard federated model for multi-access edge computing

2021-07-27 13:09 Embedded.com Dario Sabella
Edge computing environments comprise a diverse ecosystem of equipment and providers, driving the need for collaboration to form the so-called MEC Federation needed to provide global edge computing services to customers.


(Source: ETSI)

Multi-access Edge Computing (MEC) offers application developers and content providers cloud-computing capabilities and an IT service environment at the edge of the network. This technology, standardized by ETSI ISG MEC, enables an open market and new business models, including the possibility to serve multiple use cases and applications. Edge computing environments are also characterized by a diverse ecosystem of market players, ranging from infrastructure owners, to service providers, system integrators and application developers. This scenario is characterized by a complex multi-vendor, multi-supplier, multi-set of equipment including both HW and SW devices. As an additional degree of complexity, typical real-world scenario include the presence of multiple Operators, each one with its own edge computing infrastructure, network capabilities and customer base.

In this heterogeneous environment, in the view of offering global edge computing services to customers, Operators are realizing the need to collaborate and form the so-called MEC Federation, which use cases and key issues are defined and analyzed in the recently published Report GR MEC 035 “to enable inter-MEC system deployment and MEC-Cloud system coordination”.

click for full size image

MEC Federation helps provide customers with seamless access across a diverse ecosystem of Mobile Network Operator (MNO environments. (Source: ETSI)

The study, led by the Japanese telecommunications operator KDDI, supported by other operators and in collaboration with many other MEC members and participants, was also taking into account the requirements from GSMA OPG (Operator Platform Group). The goal of that GSMA initiative is to make edge computing an Operator service, where customers using an edge application should have seamless access to applications with edge Quality of Experience, whether the application is running on their Operator’s edge cloud, or on the edge cloud of a different (but federated) Operator. The work in ETSI ISG MEC is indeed targeted at introducing a proper standard that could achieve this goal (also in collaboration with 3GPP and other bodies).

More in detail, the ETSI GR 035 report is highlighting the problems, gaps and solutions, and recommendations related to a MEC Federation, which is defined as “federated model of MEC systems enabling shared usage of MEC services and applications”. Various use cases are motivating the establishment of a MEC federation, and they are analyzed in this report, together with a description of possible solutions and related evaluations:

  1. MEC federation scenario of V2X services
  2. multi-operator agreements enabling MEC Federation for V2X services
  3. Application instance transfer between MEC and Cloud systems
  4. Inter-system communication involving a MEC system in an MNO’s network
  5. MEC federation scenario for connecting different services
  6. MEC federation scenario for immersive AR game
  7. MEC federation scenario for Edge Service availability on visited networks
  8. MEC federation scenario for edge node sharing

In particular, the use case #1 (“MEC federation scenario of V2X services”) is of interest for automotive stakeholders from 5GAA (5G Automotive Association), who recently joined ETSI MEC membership. This MEC federation scenario of V2X services (i.e., multi-MNO, multi-OEM, multi-MEC) is typical in Smart Cities, where cars with different SIM Cards and network subscriptions are connected to different operators, each one with different capabilities, in terms of availability of MEC platform and/or APIs. In this perspective, enabling inter-MEC system communications between MEC applications, but also application interoperability and portability, is a key need for the whole automotive ecosystem, to provide V2X services (starting from MNOs, OEMs, but also technology providers, city municipalities and their subsidiaries, regulators and policy makers). From a technology perspective, this scenario requires the MEC system to support:

  • MEC system discovery, including security (authentication/authorization, system topology hiding/encryption), charging, identity management and monitoring aspects as an essential prerequisite to form a MEC federation.
  • MEC platform discovery, by means of the MEC systems exchanging information about their MEC platforms, i.e. their identities, a list of their shared services, as well as authorization and access policies.
  • Information exchange at MEC platform level, for the needs of MEC service consumption, or for MEC app-to-app communication.

Another use case of interest for the automotive domain is the #8 (“MEC federation scenario for edge node sharing”), where the MEC federation may be also used to share edge capabilities from one operator to another, in situations where one of them has no edge resources in a certain region. This is particularly convenient for all operators that may want to join a federation and benefit from each other’s edge computing infrastructures, in the view of offering a global user experience to their customers, whichever country the customers are located. In addition to the requirements introduced by the first use case, this one is also requiring a connectivity between MEC platforms of Operator A and network gateway of Operator B (and vice versa) to optimize the service delivery form one operator to the other. Consequently, this implies the introduction in the MEC Standard of an information exchange mechanism to support application discovery/publishing as well as subscriber redirection.

The requirements defined by GR 035 (not only related to automotive domain, but in principle serving all MEC use cases) led naturally to the start of a normative work in ETSI MEC (work item GS 040), that will be focused of the definition of proper MEC Federation APIs as communication means between federation managers of different MEC Systems (i.e. the so-called EWBI, in OP parlance).

The ETSI MEC GR 035 report is the very first study in this domain, considering the GSMA requirements on MEC Federation, and the aim of ETSI ISG MEC is to collaborate with 3GPP and open source communities to deliver a consistent standard that could support an interoperable service offer to all stakeholders in the edge computing ecosystem.

Dario Sabella is Chairman of ETSI MEC (Multi-access Edge Computing), an Industry Specification Group (ISG) within ETSI. He works with Intel as Senior Manager Standards and Research, driving new technologies and edge cloud innovation for advanced systems, involved in ecosystem engagement and coordinating internal alignment on edge computing across standards and industry groups. Prior to becoming ETSI MEC chair, from 2019 he served as vice-chairman, previously Lead of Industry Groups, and from 2015 vice-chair of IEG WG. Since 2017 he is also a delegate of 5GAA (5G Automotive Association). Before 2017 he worked in TIM (Telecom Italia group), as responsible in various research, experimental and operational activities on OFDMA technologies (WiMAX, LTE, 5G). He is author of several publications (40+) and patents (30+) in the field of wireless communications, energy efficiency and edge computing, Dario is IEEE senior member and has also organized several international workshops and conferences.

Related Contents:

For more Embedded, subscribe to Embedded’s weekly email newsletter.

The post Defining a standard federated model for multi-access edge computing appeared first on Embedded.com.

>>>


Cree XP-P LED sample arrive

2021-07-27 12:54 ElectronicsWeekly Steve Bush

Thanks to Cree, there are now some XP-P led samples at (virtual) Electronics Weekly Tower. I was dying to experiment on some, and they were proving difficult to buy. They arrived beautifully packaged and even vacuum-wrapped. XP-P is Cree’s high-intensity led for long narrow beams – it has a 9Wmax 1 x 1mm die with ...

This story continues at Cree XP-P LED sample arrive

Or just read more coverage at Electronics Weekly

>>>


Gelsinger Nails Colours To The Mast

2021-07-27 11:05 ElectronicsWeekly David Manners

When the Captain nails the ship’s colours to the mast, the crew knows it’s crunch time – conquer or die. This is what Pat Gelsinger did yesterday. He gave several hostages to fortune and said ‘measure us against these goals’. It was a brave thing to do – particularly in the light of Intel’s many ...

This story continues at Gelsinger Nails Colours To The Mast

Or just read more coverage at Electronics Weekly

>>>


SpiNNaker2 tapes out

2021-07-27 10:35 ElectronicsWeekly David Manners

Technische Universität Dresden, Manchester University, Racyics GmbH, and GlobalFoundries have taped out the SpiNNaker2 AI processor. Based on the SpiNNaker infrastructure developed by Manchester University , the hybrid AI architecture created at TU Dresden, the adaptive body bias IP platform ABX by Racyics, and utilising GLoFo’s 22nm FD-SOI process, SpiNNaker2 is a realtime neuromorphic AI processor with ...

This story continues at SpiNNaker2 tapes out

Or just read more coverage at Electronics Weekly

>>>


$2 MXCHIP EMC3080 WiFi and Bluetooth LE IoT module integrates Cortex-M33 MCU

2021-07-27 09:56 CNXSoft Jean-Luc Aufranc (CNXSoft)

While nowadays most people from the maker community are working with ESP8266 or ESP32 modules...

>>>