FromBeyond
Member
I think the salient line may be “If I can wait”.
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Frangipani
Top 20
And then there are also those who have recently joined BrainChip, such as Winston Tang and Aras Pirbadian, bringing along their individual talents, skills and interests:
Winston Tang - Accenture | LinkedIn
Experience: Accenture · Education: University of California, Berkeley · Location: San Francisco Bay Area · 172 connections on LinkedIn. View Winston Tang’s profile on LinkedIn, a professional community of 1 billion members.www.linkedin.com
View attachment 83933
View attachment 83936
View attachment 83937 View attachment 83938
Blog: "How to Solve the Size, Weight, Power and Cooling Challenge in Radar & Radio Frequency Modulation Classification" | Semi IP Hub posted on the topic | LinkedIn
Blog: "How to Solve the Size, Weight, Power and Cooling Challenge in Radar & Radio Frequency Modulation Classification" By Aras Pirbadian and Amir Naderi, BrainChip https://lnkd.in/ewpnjDrw In this post, we’ll show how BrainChip’s Temporal Event-Based Neural Network (TENN), a state space...
How to Solve the Size, Weight, Power and Cooling Challenge in Radar & Radio Frequency Modulation Classification
In this post, we’ll show how BrainChip’s Temporal Event-Based Neural Network (TENN), a state space model, overcomes this challenge. You’ll learn why...
How to Solve the Size, Weight, Power and Cooling Challenge in Radar & Radio Frequency Modulation Classification
By Aras Pirbadian and Amir NaderiJune 27, 2025
Modern radar and Radio Frequency (RF) signal processing systems—especially those deployed on platforms like drones, CubeSats, and portable systems—are increasingly limited by strict Size, Weight, Power and Cooling (SWaP-Cool) constraints. These environments demand real-time performance and efficient computation, yet many conventional algorithms are too resource-intensive to operate within such tight margins. As the need for intelligent signal interpretation at the edge grows, it becomes essential to identify processing methods that balance accuracy within these constraints.
One such essential task in radar and RF applications is Automatic Modulation Classification (AMC). AMC enables systems to autonomously recognize the modulation type of incoming signals without prior coordination, a function crucial for dynamic spectrum access, electronic warfare, and cognitive radar systems. However, many existing AI-based AMC models, such as deep CNNs or hybrid ensembles, are computationally heavy and ill-suited for low- SWaP-Cool deployment, creating a pressing gap between performance needs and implementation feasibility.
In this post, we’ll show how BrainChip’s Temporal Event-Based Neural Network (TENN), a state space model, overcomes this challenge. You’ll learn why conventional models fall short in AMC tasks—and how TENN enables efficient, accurate, low-latency classification, even in noisy RF environments.
Why Traditional AMC Models Fall Short at the Edge
AMC is essential for identifying unknown or hostile signals, enabling cognitive electronic warfare, and managing spectrum access. But systems like UAVs, edge sensors, and small satellites can’t afford large models that eat power and memory.Unfortunately, traditional deep learning architectures used for AMC come with real drawbacks:
- Hundreds of millions of Multiply Accumulate (MAC) operations resulting in high power consumption and large parameter counts demanding large memory
- Heavy preprocessing requirements (e.g., Fast Fourier Transform (FFTs), spectrograms)
- Still fail to maintain accuracy under 0 dB Signal-to-Noise Ratio (SNR), where signal and noise have similar power.
BrainChip’s TENN Model: A Low-SWaP-Cool Breakthrough for Real-Time RF Signal Processing
BrainChip’s TENN model provides a game-changing alternative. It replaces traditional CNNs with structured state-space layers and is specifically optimized for low SWaP-Cool high-performance RF signal processing. State‑Space Models (SSMs) propagate a compact hidden state forward in time, so they need only constant‑size memory at every step. Modern SSM layers often recast this recurrent update as a convolution of the input with a small set of basis kernels produced by recurrence. Inference‑time efficiency therefore matches that of classic RNNs, but SSMs enjoy a major edge during training: like Transformers, they expose parallelizable convolutional structure, eliminating the strict step‑by‑step back‑propagation bottleneck that slows RNN training. The result is a sequence model that is memory‑frugal in deployment yet markedly faster to train than traditional RNNs, while still capturing long‑range dependencies without the quadratic cost of attention of Transformers.TENN introduces the following innovations:
- A compact state-space modeling that simplifies modulation classification by reducing memory usage and computation—offering a leaner alternative to transformer-based models.
- Tensor contraction optimization, applying efficient strategies to minimize memory footprint, computation and maximize throughput.
- Hybrid SSM architecture that replaces CNN layers and avoids attention mechanisms, maintaining feature richness with lower computational cost.
- Real-time, low-latency inference by eliminating the need for FFTs or buffering at inference time
Matching Accuracy with a Fraction of the Compute
The Convolutional Long Short-Term Deep Neural Network (CLDNN), introduced by O’Shea et al. (2018), was selected as the benchmark model for comparison with BrainChip’s TENN. Although the original RadioML paper did not use the CLDNN acronym, it proposed a hybrid architecture combining convolutional layers with LSTM and fully connected layers—an architecture that has since become widely referred to as CLDNN in the AMC literature.This model was chosen as a reference because it comes from the foundational paper that introduced the RadioML dataset—making it a widely accepted standard for evaluation. As a hybrid of convolutional and LSTM layers, CLDNN offers a meaningful performance baseline by capturing both spectral and temporal features of the input signals in the In-phase (I) and Quadrature (Q) (I/Q) components, which are used to represent complex signals in communication systems.
While more recent models like the Mixture-of-Experts AMC (MoE-AMC) have achieved state-of-the-art accuracy on the RadioML 2018.01A dataset, they rely on complex ensemble strategies involving multiple specialized networks, making them unsuitable for low-SWaP-Cool deployments due to their high computational and memory demands. In contrast, TENN matches or exceeds the accuracy of CLDNN, while operating at a fraction of the resource cost—delivering real-time, low-latency AMC performance with under 4 million MACs and no reliance on using multi-model ensembles or hand-crafted features like spectral pre-processing.
With just ~3.7 million MACs and 276K parameters, TENN is over 100x more efficient than CLDNN, while matching or exceeding its accuracy—even in low-SNR regimes. Moreover, the latency in the table refers to the simulated latency on a A30 GPU for both models.
On the RadioML 2018.01A dataset (24 modulations, –20 to +30 dB), TENN consistently outperforms CLDNN especially in mid to higher SNR scenarios. Here is the performance of TENN compared to CLDNN's over the SNR range of -20 to +30 dB:
Ready to bring low SWaP-Cool AI to your RF platforms?
Today’s RF systems need fast, accurate signal classification that fits into small power and compute envelopes. CLDNN and similar models are simply too resource intensive. With TENN, BrainChip offers a smarter, more scalable approach—one that’s purpose-built for edge intelligence.By leveraging efficient state-space modeling, TENN delivers:
- Dramatically reduces latency, power consumption, and cooling requirements
- Robust accuracy across noisy environments
- Seamless deployment on real-time, mobile RF platforms
Schedule a demo with our team to benchmark your modulation use cases on BrainChip’s event-driven AI platform and explore how TENN can be tailored to your RF edge deployment.
Book Demo
Tools and Resources Used
- Dataset: RadioML 2018.01A – A widely used AMC benchmark with 2 million samples:
- DeepSig Inc., "Datasets," [Online]. Available: https://www.deepsig.io/datasets
- BrainChip Paper: Pei , Let SSMs be ConvNets: State-Space Modeling with Optimal Tensor Contractions arXiv, 2024. Available: https://arxiv.org/pdf/2501.13230
- Reference Paper: O’Shea, T. J., Roy, T., & Clancy, T. C. (2018). Over-the-air deep learning based radio signal classification. IEEE Journal of Selected Topics in Signal Processing, 12(1), 168–179. Available: https://doi.org/10.1109/JSTSP.2018.2797022
- Framework: PyTorch was used to implement and train the TENN-based SSM classifier
- Thop is a library designed to profile PyTorch models; it calculates the number of MACs and parameters.
Frangipani
Top 20
And then there are also those who have recently joined BrainChip, such as Winston Tang and Aras Pirbadian, bringing along their individual talents, skills and interests:
Winston Tang - Accenture | LinkedIn
Experience: Accenture · Education: University of California, Berkeley · Location: San Francisco Bay Area · 172 connections on LinkedIn. View Winston Tang’s profile on LinkedIn, a professional community of 1 billion members.
View attachment 83933
View attachment 83936
View attachment 83937 View attachment 83938
Meanwhile, one of our former interns left us after two months and is now interning with Accenture:
Winston Tang - Accenture | LinkedIn
Experience: Accenture · Education: University of California, Berkeley · Location: San Francisco Bay Area · 172 connections on LinkedIn. View Winston Tang’s profile on LinkedIn, a professional community of 1 billion members.
We should have some feedback from this exhibition soon:
https://brainchip.com/sensors-converge-2025/
Join BrainChip at Sensors Converge 2025. Discover the future of sensing, processing, and connectivity at Sensors Converge 2025, North America’s leading event for design engineers. BrainChip is proud to exhibit and present how Akida™ technology is driving smarter, energy-efficient solutions for next-gen systems and IoT devices.
https://brainchip.com/sensors-converge-2025/
June 24-26 | Santa Clara Convention Center, CA
Join BrainChip at Sensors Converge 2025. Discover the future of sensing, processing, and connectivity at Sensors Converge 2025, North America’s leading event for design engineers. BrainChip is proud to exhibit and present how Akida™ technology is driving smarter, energy-efficient solutions for next-gen systems and IoT devices.
smoothsailing18
Regular
The momentum is definitely picking up
Bravo
If ARM was an arm, BRN would be its biceps💪!
Qualcomm targets Taiwan drone, space startups as edge AI shift accelerates
Qualcomm has backed nearly 70 Taiwan startups since 2019 and is now focusing on drone, space technology, and cybersecurity companies as artificial intelligence processing moves from cloud to edge devices.
www.digitimes.com
They may as well target us, since these are all the areas we excel at.
EXTRACT
Qualcomm has backed nearly 70 Taiwan startups since 2019 and is now focusing on drone, space technology, and cybersecurity companies as artificial intelligence processing moves from cloud to edge devices.
BrainShit
Regular
Hi TECH,
are you privy to know more about the depth of engagement that our competitors have with their partners?
Have a look at the ecosystems of two of those companies that, just like us, already have products on the market. Pretty unknown names, huh? Oh, wait…
Innovation -
innatera.com
View attachment 87728
View attachment 87733
About Us | SynSense
SynSense focuses on neuromorphic intelligence and takes the lead in commercialization of neuromorphic chips; SynSense provides application solutions and services.www.synsense.ai
View attachment 87729
View attachment 87730 View attachment 87732
Who ever underestimate INNATERA is a fool....
Frangipani
Top 20
Edge AI Milan 2025
Edge AI Milan 2025 Join BrainChip for Edge AI Milan 2024 July 2-4, an inspiring and forward-thinking event exploring how edge AI is bridging the gap between digital intelligence and real-world impact. Attendees can engage with industry leaders and experience the latest innovations including...
brainchip.com
Edge AI Milan 2025
Edge AI Milan 2025 Join BrainChip for Edge AI Milan 2024 July 2-4, an inspiring and forward-thinking event exploring how edge AI is bridging the gap between digital intelligence and real-world impact. Attendees can engage with industry leaders and experience the latest innovations including...
brainchip.com
Edge AI Milan 2025
Join BrainChip for Edge AI Milan 2024 July 2-4, an inspiring and forward-thinking event exploring how edge AI is bridging the gap between digital intelligence and real-world impact. Attendees can engage with industry leaders and experience the latest innovations including Brainchip’s Akida neuromorphic, event-driven AI at the edge.
Register
BrainChip will be exhibiting at Edge AI Milan 2025 next week.
It’s a pity, though, that no one from our company will be giving a presentation at that conference, especially since Innatera will be spruiking their technology.
In addition, Innatera’s Petruț Antoniu Bogdan will give a workshop on neuromorphic computing “on the current state of neuromorphic computing technologies and their potential to transform Edge AI” in his capacity as co-chair of the Edge AI Foundation Working Group on Neuromorphic Computing, which BrainChip also belongs to.
Our CMO Steve Brightfield is co-chair of another Edge AI Working Group, namely Marketing.
EDGE AI Working Groups
<p>EDGE AI FOUNDATION Working Groups are populated by industry partners and academia to drive cross industry initiatives and best practices in a neutral and collaborative environment. Working Groups are open to Strategic Partners and academia.</p>
How Neuromorphic Chips Could Redefine Edge AI Devices | Pete Bernard
We have a Neuromorphic working group at EDGE AI FOUNDATION that is comprised of the leading technology companies in this space as well as leading academic researchers. This is an area that should be focused on as it gets broader commercial adoption! Petruț Antoniu Bogdan Charlotte Frenkel...
Frangipani
Top 20
And that’s a wrap on Sensors Converge 2025. | HaiLa
And that’s a wrap on Sensors Converge 2025. Huge thanks to all who came by and spoke to us about really low power communications over Wi-Fi and Bluetooth and to those that participated in one or more of the conference sessions Charlotte Savage and Patricia Bower participated in. Special shout...
Frangipani
Top 20
The hypersonic threat is real. | Air Force Office of Scientific Research (AFOSR) posted on the topic | LinkedIn
The hypersonic threat is real. Cyber attacks are constant. How do we outpace adversaries in testing and securing critical technologies? The answer lies in foundational breakthroughs. Our Agile Science for T&E program fuels the #BasicResearch that makes rapid innovation possible. This AFOSR...
Frangipani
Top 20
It was fun to be the part of the EDGE AI FOUNDATION panel at #SensorsConverge discussing the future of #EdgeAI. | Scott Smyser
It was fun to be the part of the EDGE AI FOUNDATION panel at #SensorsConverge discussing the future of #EdgeAI. It was an insightful discussion with fellow providers in the industry: Syntiant Corp., BrainChip , Ambient Scientific, Innatera. Each company has a unique offering, with EMASS’s...
Frangipani
Top 20
Listen to your body. | Gary Spittle
Listen to your body. The ear is a fantastic place to gather biometric information. At Sensors Converge we got an update in the latest in heart rate, motion, brain activity monitoring and many others. Was great to see IDUN Technologies, Neuranics, BrainChip, InnoPhase IoT, Syntiant Corp...
Gary Spittle - Sonical Inc | LinkedIn
Experience: Sonical Inc · Education: University of York · Location: San Francisco · 500+ connections on LinkedIn. View Gary Spittle’s profile on LinkedIn, a professional community of 1 billion members.
Sonical
Headphone 3.0
Products
Last edited:
BrainChip and our new partners HaiLa were both on-hand at the 40th annual Sensors Converge conference in Santa Clara this week demonstrating how combining BrainChip’s Akida neuromorphic technology… | BrainChip
BrainChip and our new partners HaiLa were both on-hand at the 40th annual Sensors Converge conference in Santa Clara this week demonstrating how combining BrainChip’s Akida neuromorphic technology with HaiLa’s ultra-low power wireless connectivity unlocks unprecedented power efficiency at the...
Hi TECH,
are you privy to know more about the depth of engagement that our competitors have with their partners?
Have a look at the ecosystems of two of those companies that, just like us, already have products on the market. Pretty unknown names, huh? Oh, wait…
Innovation -
View attachment 87728
View attachment 87733
About Us | SynSense
SynSense focuses on neuromorphic intelligence and takes the lead in commercialization of neuromorphic chips; SynSense provides application solutions and services.
View attachment 87729
View attachment 87730 View attachment 87732
Hi Frangi,
For the last decade every poster on HC and this forum know that I am totally one-eyed.....my posts generally contain a combination
of fact and personal opinion.
Yes, we have competition, which is very healthy, and no we will never command the whole pie, or 50%, or even 25% (guess) but we
will establish ourselves as a respected player (IP supplier) in the Edge AI market, to think otherwise would be extremely negative with
what we "believe" we have going on currently.
Thank you for your continued dedication (time) devoted to keeping our forum balanced.
Kind regards......Tech
The latest press release from BrainChip has shed light on their innovative approach to addressing the long-standing Size, Weight, Power, and Cooling (SWaP‑C) challenge, particularly in radar and RF… | Philip Dodge
The latest press release from BrainChip has shed light on their innovative approach to addressing the long-standing Size, Weight, Power, and Cooling (SWaP‑C) challenge, particularly in radar and RF modulation classification. For professionals engaged in real-time signal processing within...
DingoBorat
Slim
Sean has stated a couple of times at least now. how when New Technological frontiers are created, it is a bit of a "Wild West" for a while, until 2 or 3 dominant players, come to the fore.
He wants "us" to be one of those dominant players and so obviously, do we!
I would think this would look like, something of a 20 to 30% market share, of our targeted markets combined.
With our apparent foothold lead, in the Premier, Space, Military and Medical fields, this is well within reach, in my opinion.
But it's obviously the Big breakthrough in "bread and butter" consumer markets, that we eagerly seek.
Despite what our 20 cent share price and lack of any serious revenue shows, "we" have been laying the foundations and groundwork, to become one of those dominant players, for some time now.
Tony Lewis is hinting very strongly at the announcement of a new technological update, coming out imminently, which while strategically important and value adding from an IP perspective, is not the kind of announcement we really need.
I know Sean must have mixed up the Time Zones, as "his" Friday has only recently ended and the necessity to announce a new large IP deal on the ASX, missed even "our" late announcements deadline.
So Monday morning, is still on the cards, for the Big Announcement, I promised last week.
Last edited:
BULLIIIIIIIISH
The latest press release from BrainChip has shed light on their innovative approach to addressing the long-standing Size, Weight, Power, and Cooling (SWaP‑C) challenge, particularly in radar and RF… | Philip Dodge
The latest press release from BrainChip has shed light on their innovative approach to addressing the long-standing Size, Weight, Power, and Cooling (SWaP‑C) challenge, particularly in radar and RF modulation classification. For professionals engaged in real-time signal processing within...
smoothsailing18
Regular
FF. On crapper.........This part of the website was updated in the last 12 hours. So much now ready and being offered to customers
brainchip.com
Ready-to Use Akida Neural Network Models
BrainChipTM provides a set of pre-trained, documented and supported network architectures. Here is a variety of models that are available in our model library in the Developer Hub: Explore Akida’s Model Library Discover ready-to-use models built for low-power, real-time performance...
brainchip.com
🆘 𝗦𝗔𝗩𝗘 𝗣𝗘𝗢𝗣𝗟𝗘 𝟭𝟬𝗫 𝗙𝗔𝗦𝗧𝗘𝗥 𝗪𝗜𝗧𝗛 #𝗥𝗢𝗕𝗢𝗧 𝗪𝗔𝗧𝗘𝗥 #𝗟𝗜𝗙𝗘𝗚𝗨𝗔𝗥𝗗 🤖 | Jason De Silveira | 123 comments
🆘 𝗦𝗔𝗩𝗘 𝗣𝗘𝗢𝗣𝗟𝗘 𝟭𝟬𝗫 𝗙𝗔𝗦𝗧𝗘𝗥 𝗪𝗜𝗧𝗛 #𝗥𝗢𝗕𝗢𝗧 𝗪𝗔𝗧𝗘𝗥 #𝗟𝗜𝗙𝗘𝗚𝗨𝗔𝗥𝗗 🤖 ✅ remote controlled ✅ 10x faster than swimming ✅ keeps people from drowning I love these kinds of robotic applications #robots for help, not war 🙏👌 Follow Jason De Silveira for more #robotics insights | 123 comments on LinkedIn
See Philip Dodge's conment.
Similar threads
