Darpa Plans a Major Remake of U.S. Electronics | Hacker News Reader

Darpa Plans a Major Remake of U.S. Electronics

spectrum.ieee.org

165 pointsrch67 comments

From the article: "It’s pretty rare to go to GitHub, find high-quality hardware blocks that are available, and have the verification tools and everything you would need to trust that, even though that block has been altered by many different designers, it is in a state that is useable for your design."

And so they are giving money to the likes of Intel, Nvidia, Synopsis, Xilinx? The reason why there is very little open source hardware is because it isn't in those company's interests.

I won't hold my breath, I think this is just government appropriations and a lot of self-congratulating kick-off conferences that won't amount to anything useful. We're more likely to get open hardware the same way it always has started -- from the community, not from corporations. They will only adopt it when they have no other choice.

Once IBM constructed a personal computer. But, instead of patenting everything about it, they decided to open the architecture, so that extension cards and entire clones could be made.

In a few years, IBM PC and its clones ate the market, and the resulting architecture kept eating it for like 20 years, pushing away both dedicated game machines, microcontrollers, and ultra-powerful servers. Only the advent of the smartphone slowed it down.

I'd like to notice that this was not a grassroots community effort; this was a shrewd business strategy ("it's better to own 10% of a $1B market than 90% of a $20M market"), and for-profit companies, working for the the benefit of themselves and their customers.

Not that community efforts are futile; they gave us [a long list of great things follows]. But the power of market forces, if harnessed right, should not be under-estimated.

…so that extension cards and entire clones could be made.

IBM had no intention of allowing clones. They relied on their proprietary BIOS to give them control of the product even though they were licensing an OS from another party. This failed when the clean room reimplementations of IBM’s BIOS withstood IBM’s legal challenges and clones were born. IBM attempted to regain control of the platform by creating their own OS, OS/2, but ultimately failed and was driven from the market.

"They will only adopt it when they have no other choice"

That's not why companies use and fund open source software. Open source electronics would probably be pretty much the same, added value being delivered from complete products (which are much more than just the chip) rather than just an agglomeration of components.

I wouldn't write off government investment just yet. It did, after all, give us the Internet. No community could have built that. Computer networks and hardware are both capital investment-heavy, long-term public benefits.

The OSS analogue would be ham radio. Which is great and all, but...

So much of our aerospace industry is based off of government initial R&D. If the government is going to do anything to help the tech space, this is it.

I think he's pointing out that the investment should go to a community-owned initiative, not companies who may take the funds and not take the initiative to heart.

This is really exciting. I've thought about building a datapath on an FPGA that uses partial reconfiguration to synthesize new hardware blocks based on the math being doing at a particular instant, and it sounds like DARPA wants to do that on steroids. Server CPUs coupled with FPGAs are already being used for a lot of interesting things and I think it makes a lot of sense to integrate together even tighter. I'm also really happy to hear that they're funding new EDA software that is actually smart. The best we have right now is some whitelist based rule checking for schematics and layout. I want full simulation, a perfect, proven parts library with symbols, footprints, 3D models, non-ideal SPICE models, and a thermal model for every part in existence, smart rule checking that looks not only for things like unconnected nets, but logic level mismatches and clock edges that are too fast.

Little pieces of this exist already, but in my opinion, software for creating hardware is way behind software for creating software.

Quite interesting is the bit about hiring the "former lead of GNU Radio" who had experienced pain in getting closed/proprietary hardware things to be useful. I wonder if this will lead to more open platforms. If for no other reason, that bit makes the article worth reading.

> Quite interesting is the bit about hiring the "former lead of GNU Radio" who had experienced pain in getting closed/proprietary hardware things to be useful. I wonder if this will lead to more open platforms. If for no other reason, that bit makes the article worth reading.

This makes me more worried, not less.

The reason why you couldn't repurpose those things targeted at "specific technologies" is that they were engineered for purpose.

You don't want a general "radio decoder". You want a 4G modem. You don't wan't a wideband ADC--you want low power for battery usage so you want to narrow the bandwidth as much as possible. You are willing to give up generality in order to get a couple more dB in order to increase your range by another couple miles, etc.

The problem is that it is so expensive to do your own chip that everybody tries to shoehorn their solution into something that exists in volume but really isn't quite right simply to avoid that expense.

If you want to fix electronics, find a way to make VLSI fabrication cost $500 and have a 5 day turnaround.

You don't want a general "radio decoder". You want a 4G modem.

That's consumer products. DoD wants generality even if it costs more.

The electronics industry ignores what DoD wants because DoD won't buy a million parts a month. That's been going on for a long time. It really upset some USAF generals in the 1980s; they were used to driving the industry, not following it. It's gotten worse since, because the center of the electronics industry has moved to Asia.

This project is a niche thing for military short-run production. That's fine. There may be commercial spinoffs. But it's aimed at DoD's needs.

> That's consumer products. DoD wants generality even if it costs more.

Ah, I missed that this was the DoD whingeing about the fact that they can't get people to design complicated systems for them for free.

However, even the DoD doesn't want only generality. The DoD generally has hard constraints on power and size and that fights against (and generally beats) generality.

Why does DoD want generality?

It's cheaper for contractors to slap together a bunch of COTS stuff and get good functionality for a "low price" versus building the perfect system from all-in-one designs. The manufacturers aren't going to make military spec all-in-one components because the military isn't buying that much compared to the commercial market so the manufacturers aren't going to invest the time and money into something that isn't going to sell volume. Even if they would make these parts, they are going to cost a lot more than what it would cost to design a slightly more complex, less integrated system. Everyone is better off with more general parts. More components in a system also allow the designer to swap portions out for newer designs as time goes on where as an all-in-one design is going to require a lot of NRE to update.

Military radios are unlike civilian radios in that they have to transmit and receive on a really wide range of frequencies. The Wikipedia article on the AN/PRC-152 handheld radio claims it can receive from 30 to 870 MHz. (As of 2014, they cost $13,000 each, in a tiny production run of 1500 units https://www.militaryaerospace.com/articles/2014/08/air-force...)

The military also wants the ability to change waveforms and protocols on the fly to avoid detection, another thing civilian radios never need.

They're trying to replace all their current radios with the Joint Tactical Radio System, which is the usual defense procurement boondoggle. Billions spent, not much produced: https://arstechnica.com/information-technology/2012/06/how-t... https://arstechnica.com/information-technology/2018/03/the-a...

>The final contract for Rifleman Radios, awarded to Thales and Harris, was for up to $3.9 billion to deliver 193,276 more radios. That amounts to around $20,000 per radio, including accessories and support; in 2014, the military expected the Rifleman Radio to cost about $5,600 per unit. The Manpack radios were expected in 2014 to cost about $72,000; the actual line item for Manpacks under the 2017 budget was $114.9 million for 1,459 radios; that's nearly $78,000 and does not include the maintenance, accessories, and services. That's not exactly the kind of cost savings one would expect from a commodity, standards-based system.

Personally I'd rather have a 4G modem that can I can reprogram for something useful after my phone is obsolete, but I'll take the special purpose chip if it's cheaper. What I don't want is a general purpose SDR that's been artificially nerfed to provide only 4G while 90% of the cost is in licensing.

There are things (anti-things?) you want and there are the things the market provides.

https://www.digikey.com/product-detail/en/analog-devices-inc...

The AD9361 is exactly that all purpose programmable radio.

http://www.analog.com/en/products/rf-microwave/integrated-tr...

https://zeptobars.com/en/read/AD9361-SDR-Analog-Devices-DAC-...

With a receive and transmit current in the range of 1/4 to 1/2 an ampere.

Not going to make many portable radios out of that, methinks.

That is what RF amplifiers are for. The output current for an IC that already does SDR is huge. I think you are at the limits of what you can reliably critique.

> You don't want a general "radio decoder". You want a 4G modem

Might we not want both, for different purposes? Software-defined radio won't match the performance or efficiency of specialized RF hardware, but that doesn't mean it doesn't have any utility.

>If you want to fix electronics, find a way to make VLSI fabrication cost $500 and have a 5 day turnaround.

Agreed. The root problem is that producing chips has become so expensive and proprietary that it has become extremely centralized. I would bet that no more than a few hundred people decide the big design questions of the chips in nearly all computers produced today.

It's worse than that. As a new player I can't even get my calls answered.

How much did you pay for the VLSI software tooling?

It's not relevant. The reason why all the semiconductor companies are buying each other is that it is more cost effective than trying to pry a design win out of somebody's hands that might or might not result in volume.

Or, more directly, someone who already has the design win can probably undercut your price and still make a profit (don't have to redesign anything so all the NRE is already sunk). So, if you want the design win, you have to take an initial loss and then hope you can squeeze the cost of the chip to gain some profit. Not a good bet.

The pot of money for semiconductors is effectively fixed for right now so companies are going to just keep eating one another until that changes or governments step in.

In our case we're an FPGA shop looking to transition. There really needs to be a support group for people like us. We have a hard time figuring out what's even feasible.

There used to be companies like that. Unfortunately, the market for that is so small that it simply isn't worth it and they all died.

The difference these days is that with the death of Dennard scaling and Moore's law there's a place now for customized hardware. You really can beat Intel and even NVidia now for specialized workloads with FPGAs. Of course you'd be even better with an ASIC which is why we're leaning that way.

Intel sees this threat which is why they are working to glue an altera asic onto the xeon core. I don't think it's likely they will succeed though.

Why don't you want a software defined radio?

This is the same project that didn't give any funding to Clifford Wolf (of YOSYS fame) so they could give it to ADI, TI, Xilinx, Cadence and Synopsys, etc.

I expect nothing to come of this. With the fox watching the hen house, there will be no change.

https://twitter.com/ico_TC/status/1012766234784366593?s=01

For reference.

Originally known as ARPA, Advanced Research Projects Agency, the agency helped create the internet: In 1973, the U.S. Defense Advanced Research Projects Agency (DARPA) initiated a research program to investigate techniques and technologies for interlinking packet networks of various kinds. The objective was to develop communication protocols which would allow networked computers to communicate transparently across multiple, linked packet networks. This was called the Internetting project and the system of networks which emerged from the research was known as the “Internet.” The system of protocols which was developed over the course of this research effort became known as the TCP/IP Protocol Suite, after the two initial protocols developed: Transmission Control Protocol (TCP) and Internet Protocol (IP). Source: www.internetsociety.org

The motivations for this new initiative could be numerous, including a step back toward the ARPA days, a way to outdo the Chinese in defense applications, and/or a way to make a political claim for the next election.

I'd be curious to hear what more informed people think the motivation might be. Regardless, it seems to be a positive step towards electronics innovation.

One would expect the competition between the US and China to get ferocious over the next 20 years. The US hasn't had a serious direct competitor across the board (military + economic + political) in 70 years. My personal opinion as an American, is that the US has gotten lackadaisical about its formerly overwhelming superpower position, its focus, and the need to compete at a high level. That isn't surprising of course, one would almost expect that to always occur, and China has come on extremely fast. Hopefully China's rapid rise forces a reawakening of an older can-do spirit, across all spheres of US life (perhaps most especially in regards to the invaluable contributions a better functioning government can bring, as in the past). If not, then China will de facto rule Asia and half of Europe.

A great book on the subject of the decline of Western superiority is “Has the West Lost it?” by Mahbubani [1]. He gives a fantastic overview of the stalling of the West following the confirmation of the allegeded superiority of the West’s “rational governance” following the breakup of the USSR.

He discusses the concurrent rise of Asia’s economic prowess and that the west must internalize that they have “lost it” before they can beat strategize for the coming decades.

My favorite portion of the book states that the most important event that happened in 2001 was not the terrorist attacks on 9/11, but China joining the WTO.

[1] https://www.amazon.com/Has-West-Lost-Kishore-Mahbubani-ebook...

That assumes the competition is inevitable. Hopefully, China can become a democracy, set free its people, and then the competition will be no more ferocious than the U.S.'s with Europe, Japan, India, Brazil, or any other democracy. The 'competition' is a massive waste of money and possibly of blood that benefits very few people at the expense of billions of others. Compare the competition to the relationships among democracies; what is gained by it?

China is 6000 years old and it had never been anything even remotely resembling a democracy. It’s going to remain a bureaucratic autocracy just as it always has been. And that’s fine, it’s worked very well for the Han.

Just a reminder that Taiwan, a country deeply steeped in traditional Chinese culture, is a functioning multi-party democracy.

Perhaps Mainland China is too big, or too poor, to become a stable democracy at this time, but Chinese culture should not by itself be a reason for China to remain a "bureaucratic autocracy".

And that's why the PRC (= Mainland) is trying everything it can to take over Taiwan, and subvert / destroy this democracy. After all the Chinese Communist Party's convenient narrative that justifies it's 'eternal' rule is just this, that the Chinese people neither want, nor need multi-party democracy.

(See also: Iran vs Saudi-Arabia. The former -- despite all the imperfections of its current mode of government -- is a living proof that Muslims don't need the iron fist of a hereditory ruling family.)

China has an amazing history, but has been rife with war and tumult. It has not been a continuous well run autocratic state for 6000 years. Take the Han dynasty:

After 92 AD, the palace eunuchs increasingly involved themselves in court politics, engaging in violent power struggles between the various consort clans of the empresses and empresses dowager, causing the Han's ultimate downfall. Imperial authority was also seriously challenged by large Daoist religious societies which instigated the Yellow Turban Rebellion and the Five Pecks of Rice Rebellion. Following the death of Emperor Ling (r. 168–189 AD), the palace eunuchs suffered wholesale massacre by military officers, allowing members of the aristocracy and military governors to become warlords and divide the empire. When Cao Pi, King of Wei, usurped the throne from Emperor Xian, the Han dynasty would eventually collapse and ceased to exist.

I can't help but wonder if it's worked out very well for the non-Han.

It definitely hasn't worked well for the Uyghurs and the Tibetans.

Perhaps China never dominated on the world stage because autocracy makes for a low-dynamism society/economy/tech sector. Perhaps China will come to appreciate and adopt enlightenment values, then roar past the U.S. on every mark.

China is currently busily roaring past the US on every mark mostly without adopting western values.

Made possible only by massive US assistance, both intentional and unintentional. China’s trajectory is probably self sustaining at this point, but who knows for how long.

As an American, I can assure you this is certainly not the case. We always go that extra mile for that special something that makes us shine. Don’t worry. You’re in good hands.

> The motivations for this new initiative could be numerous, including a step back toward the ARPA days, a way to outdo the Chinese in defense applications, and/or a way to make a political claim for the next election.

DARPA has been carrying out bleeding-edge R&D continuously for decades; as I understand it, that's their mission. How are these programs a step back? What is special about these programs, as opposed to other DARPA programs, that you think the motivations above would apply?

Why is DARPA investing in general purpose open source technology? I'd expect the U.S. military to seek competitive advantages, but this tech could be used by China, Russia, and Iran as easily as by the U.S. Perhaps this tech suits the U.S.'s needs better than the needs of its competitors?

DARPA has funded general purpose open source tech previously, but for the reasons above I haven't understood why. I'm glad to have the technology, I just don't grasp their motives.

Just because it's under the military doesn't mean it's only for military purposes. They want civilian technology to advance so that they can then leverage it, while also building businesses within the US to strength it economically.

Honest question: How is this not basically a subsidy for the electronics sector? The news I'm hearing from the US all the time is how China, for example, is subsidizing it's production sector and that that creates unfair market competition. From Harvard Business Review, for example: "[...] its overarching policy of aggressively subsidizing targeted industries in order to dominate global markets" and "Government subsidies to produce technologically advanced products and undercut foreign manufacturers have buttressed China’s trade prowess" [1]

The same can be said for the other technologies that were first financed by the government and then taken over by the private sector [2]

[1] https://hbr.org/2013/04/how-chinese-subsidies-changed

[2] https://www.youtube.com/watch?v=fUHrYJW-6MI

It is. If we can’t stop China from subsidizing their tech we should subsidize our own.

The United States does plenty of subsidizing, through the "Pentagon System".

https://www.youtube.com/watch?v=VSJjlaggbK0

http://www.thirdworldtraveler.com/Chomsky/PentagonSystem_Cho...

Ok, so I'm not just imagining stuff. Thanks for the reply.

That doesn't follow.

I feel you are conflating two things intentions, one is a Government encouraging/forcing companies to make something (emission control/eu privacy laws/lower taxes on electric cars/develop a stealth paint) for the good of the country, with a country helping local companies achieve global dominance by funding them thru years of losses (who collects the profits when it succeeds?)

Both have government money going to a company, both involve the good of the country. Idk one is to incentivise companies to make something you need, the other is to establish local capabilities that is self sustaining in the future. I guess some agricultural industries fall into that category in western countries

This is research. It's not directly subsidizing companies who want to build factories and products immediately.

Right, it's less directly subsidizing companies who then use the advantages gained through those subsidies to build factories and products at a comparatively reduced cost.

It's subsidy with extra steps.

Nonetheless, research can be just as pricey. This is a subsidy.

Not in the same sense, because none of it might pan out. It’s literally research grants, which might be called a subsidy because it’s for the common good, but it’s not about building expensive industries at a cheaper private cost. This is $1.5B. That’s nothing compared to what it costs to build a semi-conductor industry.

If you're interested in open source hardware, you might find the WOSET workshop interesting. It is a new workshop on open source EDA, related to this DARPA project. See http://scale.engin.brown.edu/woset/

It's interesting but how does that square with the desire to keep China and others from getting their hands on our designs when the bulk of our manufacturing is done overseas?

> The Chisel team showed that two to three student designers can do full system-on-chip designs by abstraction and automation.

> One of the things we’re trading off is the ultimate efficiency of the design. If 1 percent of area matters because you’re making millions of parts, then this level of abstraction may not be the right one for you. In our case, we’re not as cost-driven. So, if we give away 10 percent of the area but maximize the capabilities of the smaller design team, that’s the kind of tradeoff that would work for the Defense Department, but it also might work for startups and other smaller players like universities.

Presumably, one of their unspoken goals is to more easily enable advanced/secure chips to be designed and made in the USA for the military and kept confidential.

None of what you quoted mentions IC/chip fabrication. It is definitely a problem DARPA/DoD is concerned about, but this project is not addressing it.

Another reminder what a failure capitalism is. I'm so tired of hearing "markets are the best way to distribute resources" when we always see failures like this. Capitalism can't produce "big leaps" in innovation. Never could. We were so blind.

Just to check: Did you type this on a system whose entire hardware and software stack was built within some non-capitalist society, which you are claiming is superior to capitalism?

If “capitalism has failed”, as you claim, don’t you think it’s rather ironic that ALL the world’s most advanced technology comes from capitalist societies then? Shouldn’t there be at least ONE modern computer system that was built entirely without capitalism?

Open source counts as capitalist technology as well, since the vast majority of its contributors live within mostly capitalistic societies as a whole (and I’m not really aware of Soviet open source software).

Chinese technology, perhaps the greatest upcoming contender to the USA’s technological dominance, is also deeply capitalist throughout its inception (though elements of state control are deeply interwoven, the Chinese government wisely realized the benefits of capitalism and managed to use it to great success).

Capitalism is certainly deeply flawed, but regulated capitalism still seems to be the best system we have so far. Constantly saying “Capitalism has failed!” is utterly useless if you cannot propose a new system in its place that does better. (Hint: Most systems that people popularly propose have been tried time and time again.)

Certainly fair overall, but of course quite a lot of the technology we're using to have this conversation spun off from the Apollo program[1] originally...

1: https://spinoff.nasa.gov/Spinoff2018/index.html

Photonics will replace electronics by that time, of hope.

Reply on news.ycombinator.com