HNHacker News
TopNewBestAskShowJobs

bunnie

2,002 karma · joined March 9, 2013

submissionscomments
bunnie··on The New Essential Guide to Electronics in Shenzhen
Ah yes. The Precursor is real now, you can read more at https://precursor.dev. As a baseline it can function as a password/authentication token manager, but as another user mentioned we're adding messaging features now.

Another thing I'm working on is IRIS, a non-destructive method for inspecting silicon chips after mounting on a circuit board: https://arxiv.org/abs/2303.07406 It's part of a long yak-shaving arc to make the Precursor more accessible by migrating from an FPGA-based system to an ASIC.

bunnie··on SanDisk Extreme Pro failures result from design flaw, says researcher
Hard to tell from appearance only but my initial impression is that's an inductor, not a capacitor. The circuit looks like a switching power regulator. The capacitors would be beige with silver ends, this one looks like an over molded inductor, similar to [1], and is used as the main power inductor in a buck regulator.

If this is an inductor, my gut reaction is it has an insufficient current rating for the application and it is overheating. Inductors have a bunch of loss mechanisms that contribute to heating. Depending on the type of metal used to build the core, it can 'hard saturate' and effectively walk itself off a cliff once the current draw gets too high. At some point, it gets hot enough to desolder itself from the circuit board. It's possible they did not see this in validation because the power draw of SSDs depend heavily on the work load and process variations in the chips; erase current can have a fairly wide variation.

fwiw, voiding of solder joints is a problem. The solder is applied as a paste - fine particles of metal solder suspended in solder flux. During reflow the flux evaporates and leaves the metal behind, but if the process isn't tuned right bubbles of gas can be trapped in the joint. This can lead to reliability problems. It can also increase the effective thermal resistance to the circuit board, which for tiny components like this can often be the primary path for heat removal during normal operation.

[1] https://www.digikey.com/en/products/detail/pulse-electronics...

bunnie··on Regarding Proposed US Restrictions on RISC-V
The concern is that the restrictions would follow the pattern of existing rules around semiconductor equipment and fabricated chips [1].

They are fairly broad and stipulate that US persons cannot transfer technology to any company on an entity list. iirc there was a clarification to the rule that also put a burden on US persons to confirm that they aren't working with a trading front or shell company for the listed entities, which makes any interaction potentially hazardous, because how are you even supposed to check for that?

Note that violations of these rules can "include up to 20 years of imprisonment and up to $1 million in fines per violation, or both" [2]. And the use of "US persons" language means that it's not just companies, it's individual developers too.

If you read the text of the rules, they restrict any kind of technology transfer to these companies. In the case of selling fabrication equipment, that's...fairly clear cut. Don't sell stuff to people on the list, and your deal was probably a few million dollars anyways so you can afford lawyers to research if you're dealing with a shell company.

But the concern is for RISC-V "technology transfer", what happens if you push to a git repository, and...someone from one of these entities pulls your commits? Are you now liable? What if you give a lecture on RISC-V, and someone from one of these entities was in attendance?

Or perhaps you're in charge of crafting an IP strategy for your SoC company. If you pick RISC-V, the ambiguous and changing rules around who you can license your products to is problematic; a client you have today could end up being banned tomorrow. Whereas if you license an ARM core, you don't have to worry about going to jail. It then becomes a no-brainer to just license the ARM core.

Basically, the concern is they will copy/paste the existing rules that are somewhat more tightly confined because they apply to atoms, but instead insert language referring to bits and ideas.

[1] https://www.federalregister.gov/documents/2022/10/13/2022-21...

[2] https://www.bis.doc.gov/index.php/enforcement/oee/penalties

bunnie··on Tell HN: Submit comments to IRS re tax treatment of software dev expenses
I see now that Section 6.04 explicitly states that if a provider retains any right to use the developed software, they are not allowed to deduct their expenses:

> "However, even if the research provider does not bear financial risk under the terms of the contract with the research recipient, if the research provider has a right to use any resulting SRE product ... costs paid or incurred by the research provider that are incident to the SRE activities performed by the research provider under the contract are SRE expenditures of the research provider for which no deduction is allowed ..."

Thus, the rule as written would make it such that contractors who write Windows drivers could deduct their expenses (as they would have no rights to Windows), but contractors who write Linux drivers may not (as they would have some rights to Linux).

In effect, the tax rules would explicitly double-tax open source software development, due to its over-broad test of whether you could have any right to the code base you've written, irrespective of whether the code you wrote actually has any practical residual value to the author.

bunnie··on Tell HN: Submit comments to IRS re tax treatment of software dev expenses
Seems like somewhere in all of this there is an assumption that one writes software to profit from its sales.

However, in some cases people contribute software to open source projects that are not only Libre but also "Free as in Beer". Sometimes they get donations or grants to work on such projects. Heck, they might even have a contract from a company that uses this free-as-in-beer software, but, once the contributions are made, the code is free for everyone to take.

In this model, there is explicitly no forward looking profit to be had. All parties contractually agree that the value of the work, after it is done, is $0.

In this case I don't see how it would even make sense then to amortize donations/grants/contract money on a 5-year schedule when the asset value of the deliverable is $0?

I would think in this case that the person doing the development is providing a bona-fide "service industry" activity, and not doing an R&D activity. Thus the wages of that person should be as deductible as any other service-industry worker: There is as much expectation for future profits from writing this code as the person painting your house has future expectations of profits from the higher resale value of your place because of the improved paint job.

So at least for individuals being paid to work on free-to-download code, the entity paying you to do the coding may have to write off your contract as an R&D expense, but the individual contractor should still be able to depreciate that payment as a wage within the tax year because they are providing "coding services", and not doing R&D.

This logic has precedent if you consider how capital assets work. Let's say a shop purchases a donut machine. The purchaser of the donut machine has to depreciate that purchase over several years. However, the individuals who were paid to assemble the donut machine are paid wages as factory labor, and their expenses are deducted by the donut machine marker that tax year.

In this case the donut shop is whoever is paying you to write the code, and the coding contractor is the donut machine assembly technician. So long as the donut assembly technician doesn't have a stake in the donut shop's future-looking profits, I think it's quite clear who is providing a service and who owns a capital asset.

I think this logic would also mean that individual contractors who work on proprietary, for-profit code bases but assign all value of their work product immediately to the contracting body would also be a service provider and thus can depreciate their own wages within the tax year. But, the example is not as clear-cut as a work product that explicitly has a $0 expected market value.

bunnie··on The risk of RISC-V: What's going on at SiFive?
variable length instructions == CISC is not quite a correct equivalence.

The compressed 'C' extention is designed to re-use a lot of the existing decode infrastructure. On RV32 the C instructions are a strict subset of the full length instructions, so at least on RV32 it is very light weight, and it adds barely any logic to a core. It's almost always worth it to turn on C extensions versus making the cache bigger or trying to speed up main memory.

In my experience I-cache pressure is real, especially on lightweight implementations that don't have multiple levels of cache hierarchy and huge amounts of associativity to reduce the impact of an instruction cache miss.

I have played with both C and non-C variants, and also played with compiler tuning that saves code size versus 'performance' (which includes loop unrolling and thus more I cache misses). Generally smaller code size is better for power and system complexity, while keeping performance at par. Of course if you aren't as restricted on power or complexity (as is the case on a high end CPU), the calculus is different.

This kind of simplicity to me embodies the heart of RISC. If your CPU is hitting cache lines more often, you don't have to speculate as deep, don't have to re-order as much, thus less logic, less complexity, less power, higher clock rates and fewer side channels.

On the other hand, I suppose if you are already committed to deep speculation and out of order, compressed instructions might extract a disproportionate cost, maybe less so in decode and more so in precise exception handling and in tricks like register renaming.

bunnie··on The next Raptor OpenPOWER systems are coming, but they won't be Power10
I can't seem to find any information on Solid Silicon, the company making the CPU.

Anyone have a link or more details on this company? The website I can find for Solid Silicon [1] is delightfully devoid of details.

[1] https://www.solidsilicon.com

bunnie··on Arm Announces Public Filing for Proposed Initial Public Offering
I'm not an accountant, but I think I'm referring to gross profit margin [1], which is revenue minus cost of goods sold (which is small for an IP licensing company, any company that actually sells hardware will be a fraction of that).

I think your math refers to operating margin, that's after they subtract all the operating overheads which are fixed, and not proportional to sales volume. In other words a 96% gross margin means they have virtually no friction to increasing sales.

Compare to Intel statement [2], they have a 35.8% gross margin and a negative operating margin this year, which would be the apples to apples comparison against 96% gross margin and 19.6% operating margin in the F1.

My read on the income line is they do a good job of 'spending money', but they are signaling that can be turned into things like dividends to shareholders once they IPO, perhaps by doing short sighted things like cutting R&D expenses and/or accounting tricks.

[1] https://en.m.wikipedia.org/wiki/Gross_margin

[2] https://www.intc.com/news-events/press-releases/detail/1637/...

bunnie··on Arm Announces Public Filing for Proposed Initial Public Offering
Total revenue: 2.679bb Gross profit: 2.573bb

(page 18, FY2023)

96% gross profit margin? Ahh to be an IP licensing company....why would anyone invest in building physical things when there are opportunities like this out there! </s>

edit: clarify gross profit margin

bunnie··on LK-99 is an online sensation but replication efforts fall short
Perhaps a charitable interpretation of that statement is "there are years of work to go from a proof of concept to a usable product". Even if LK-99 is some kind of a superconductor, there are a whole host of daunting technical issues that will need to be solved. For example: how do we get the yield up? How to turn it into a wire? If the material is not ductile, how to form the wire into a useful shape? How to reliably couple to the wire? How does the material age over time and and humidity? How to deposit and pattern it as a thin film?

We have a lot of arrows in our quiver to shoot at these problems -- I would be optimistic that the promising applications of a room temperature superconductor would attract plenty of investment and talent. On the other hand, I imagine it may take years or even decades before we have answers to these hard questions. As the quote says, "there's no guarantee" that we can solve them, because right now we know so little about the properties of the material at this time. Of course, that is no reason to throw in the towel right now!

bunnie··on Byron Bay data breach victim told to pay Adidas, NBA $1.2M by US courts
Looking at the court documents, it seems shady organizations registered a bunch of domain names that are used in the trade of brand-name athletic gear. The victim's identity was used to register one of those domain names, but just looking at the other registrations it's pretty clear the organizations are based out of China. In this context, the case brought by the brand owners is a little more reasonable -- if they did not try to defend their brands against this level of counterfeiting, then, they can be found to have lacked diligence in defending their trademarks.

However, in the case of the victim, her owing $1.2mm in penalties hinges on her identity being used as the owner of some domain names that contain these brands' names.

I suppose if the victim had "infinite resources" the next step as the victim would be to file a lawsuit against the domain name registrars for claiming she owned the sites. If the registrar would remove her stolen identity from the site then the suit would have no basis to link her to the domain names and she would be cleared.

But then again, what incentive would a domain name registrar have to remove your stolen ID as the owner? If they simply agreed to do it when asked, then anyone could send a fake letter to the registrar claiming you don't own your DNS records and remove you as the owner of them.

Genuine question to HN readers -- if you woke up tomorrow and found a whois entry that had your name and details listed as owner for a site that traded in illicit goods and/or morally objectionable content, how would one go about correcting that?

bunnie··on ASUS to manufacture and sell Intel’s NUC products
I have used the NUCs to integrate controllers into small robotic platforms and custom lab equipment. They are a good 'it just works' platform that you can scale performance v cost over time as needs vary (need image processing? drop in a core i7. just motion control? save money and go with an i3 or atom). Also they are more reliable and performant than RPi, and relative to the cost of the motion control systems, the NUC is a rounding error.

Notably the NUC has a standardized base plate mount that requires only two screws, and it has nice rubberized stand offs that isolate vibration. So, mechanical integration is super easy, and swapping out a NUC in the field is also very fast. In some cases it is easier to do a firmware upgrade by switching out the whole NUC than trying to walk a lab tech through punching a hole in the corporate firewall, or running an ethernet cord through the production floor, etc etc.

bunnie··on Ask HN: Any hardware startups here?
We're building a device to secure your most precious secrets. https://betrusted.io (project overview) and https://precursor.dev (details). It has the form factor of a screen-and-keyboard retro-phone.

Our hardware is simple and Open so it is easier to verify. There is no "hard CPU" (it is all FPGA based) so you don't have to trust anyone about what's executing your code. The entire system runs on a pure-Rust OS (Xous) which is an async message-passing microkernel, details at https://betrusted.io/xous-book/.

Out of the box it comes with U2F/FIDO2, a TOTP authenticator, and a password vault. It's one of the few U2F/FIDO2 tokens that you can make a backup of. As the system is fully open source, you can extend it with your own favorite apps.

bunnie··on U.S. universities are building a new semiconductor workforce
It's also been a while since I stepped foot in a fab, but my impression is that for modern processes there are no people in the fab proper. It is fully automated, people sit outside the fab or they service machines in drop-down bays underneath the fab floor, accessible only through airlocks.

What the technicians are there for isn't to operate the robots...they're supposed to analyze the performance of the robots constantly and summarize anomalies and flag them to the next shift. So a work shift might involve metrology, statistics, and communication skills; I wouldn't expect that skill set from any but the most exceptional US high school graduates, and those folks are bound for the university, not a job with 12-hour days wearing clean room suits and limited bathroom breaks.

Some of the more fussy processes like EUV require regular cleaning of excess tin in the machine as well to keep yield up. afaik it's a trade secret as to how this process works, but I know one of the big delays in rolling out EUV is that the cleaning process was hard, and that lead to a lot of downtime on the line. People on both side of the Pacific were able to make chips with EUV for a while, but they just couldn't make money while making them. TSMC cracked that nut, but I haven't read any articles describing exactly what their magic was to get the uptime so high. However, I suspect it might involve a platoon of PhD-level experts that also work like an F1 pit crew.

Anyways, the days of carrying boxes of wafers from machine to machine and pressing buttons are long gone. Everything that can be automated has been, what's left is debugging and servicing robots, and searching for early signs of degradation in an otherwise fully automated machine.

So, the real value of a highly trained technician in this case is they can look at bumps in the data and go "huh..." and flag it to management so that you don't have 80% of the fab full of trash wafers. In this situation, you want someone who actually understand the physics of the process going on inside the machine, and not just someone who is following an SOP. Any analysis that could be easily automated probably already has been.

Of course, you need even more highly skilled people to set up the robots in the first place, and that process also consumes hundreds if not thousands of engineers, depending on your schedule, scale and the complexity of the machines. A lot of that can be pushed back onto vendors, but you still need an on-site team who actually understands the physics to get your process yields up to several nines and hold it there with almost no downtime.

bunnie··on U.S. focuses on invigorating ‘chiplets’ to stay cutting-edge in tech
Kinda ironic how the article is about chiplets, but all the factory photos are of an old school wire bonded PGA packaging line... But maybe that's the point, you can't find a chiplet factory within the borders of the USA to photograph for a story?
bunnie··on The AT protocol is the most obtuse crock of shit
I had the same reaction. (-_-)
bunnie··on Ask HN: Why is TSMC so competitive in semiconductor fabrication?
Except in manufacturing, getting your yield up and tuning a process is basically 'just work'. You define a space of parameters you want to explore and then you brute force it by building samples at each parameter step and then analyzing it.

This can be pipelined and parallelized to some extent, but then you have to convince enough PhD level employees to do night shifts, because each process step is basically a miniature physics or chemistry experiment that has to be monitored and tuned constantly (at this stage).

It's one thing to make one perfect transistor, it's a totally different ballgame to make 10 billion perfect devices with better than 90% tool uptime, and an essential component of closing that gap is brute force experimentation.

bunnie··on New NASA Director Swears Oath on Carl Sagan’s ‘Pale Blue Dot’ Instead of Bible
But if you don't happen to believe in God then swearing on a book of (from the oath takers perspective) fairy tails is equivalently silly / virtue signaling.

I think it's better if someone calls out that they don't believe in the thing they are swearing on and substitutes it with something they do believe in? otherwise one is making a promise on top of a lie that they also believe in the contents of that book, so the whole stack is forfeit.

Or am I missing something?

bunnie··on Upgrading my Chumby 8 kernel part 4: reboot/poweroff
Wow! someone is still using a chumby 8! I designed the hardware and some of the firmware... if you get stuck I might have some old notes laying around somewhere still, happy to share them. But from the looks of your post, you've figured out more than I can easily recall about the design.
bunnie··on Infra-Red, in Situ (Iris) Inspection of Silicon
Just adding on to taniwha's correct comments -- it is possible to indirectly image the charge using thermal stimulation of the laser to induce the Seeback effect. The local heating will cause the current consumption of the chip to shift ever so slightly in a fashion that's dependent on the state of the transistor's bias. [1] is where I first read about the technique.

[1] https://eprint.iacr.org/2018/717.pdf

bunnie··on Infra-Red, in Situ (Iris) Inspection of Silicon
His images are great! Discovering his photofeed was one of the puzzle pieces that sent me down this path in the first place. I give him a shout-out in the post :)
bunnie··on Infra-Red, in Situ (Iris) Inspection of Silicon
Hi! thanks, I'm the post's author.

I had the same concern you have about finding "just the right lens", but was pleasantly surprised to find the $180 camera/lens combo linked in the post worked just fine. In fact basically every piece of optics I had in my lab could pass light at 1050nm, and I have a bit of everything from cheap stuff bought over the counter in Shenzhen to brand-name microscopes. I tried lenses from 6 different vendors and they all worked about the same in terms of passing light (but of course the fidelity of the image scales with cost, better lens tolerances are just more expensive).

My guess is that uncoated glass should be transparent in the 1050nm band, but some lenses have AR coatings that limit the bandwidth. A lot of my lenses do have AR coating (based on the greenish tint I see in the reflections), but I guess the bandwidth of the coating used on most visible-light lenses is just wide enough for 1050nm to slide through.

bunnie··on “How America took out the Nord Stream pipeline”
What I don't get is the risk of an accidental signal triggering it. The system as described sounds similar to some existing tech used to communicate with submarines when they are under deep cover... presumably those comms come with both strong encryption and strong authentication off the rack. seems unlikely that random noise would be a concern for a trigger...

A more likely rationale for a fixed time line would simply be a concern on the life of the battery in the receiver, and the diplomatic fallout that would happen if the battery died, and an intact detonator was recovered on a later routine inspection that could be traced back to the USA.

bunnie··on Missing radioactive capsule found in WA outback after frantic search
I'm guessing that even though tho capsule is tiny, its radiation signature could be detected from quite a distance using a Geiger counter.
bunnie··on The ThinkPad TrackPoint tried to build a better mouse (2020)
I find trackpoints are more precise for CAD work. You can end a gesture with zero force precisely; but letting go of a mouse button inevitably has a tiny bit of positioning jitter. I use the track pad to scroll, and the trackpoint to draw.
bunnie··on Analog Chess
I can relate to this!
bunnie··on Towards a More Open Secure Element Chip
> But his mention of "if we release an updatable chip" sounds... concerning? It sounds like he is focusing on people being able to audit the code they're shipping rather than ship their own applications. Perhaps he could clarify what "non-updatable" means for the normal SKU?

This is driven by the paying customer (not the end user: but the entity actually writing checks with lots of zeroes on it and taking in truckloads of chips).

It is a fact that the biggest, highest volume paying customers to date for SEs would assume that such chips could only be updated through means that the customer controls. Thus, a business manager doing volume-driven research to craft a chip specification for such a market would reasonably conclude this is a mandatory checkbox for generating volume orders.

The flip side is the assumption that nobody in their right mind would pay for a chip that could be fully wiped and/or updated by end users. And to a first order, the conclusion is correct: I can't think of anyone who would write a check big enough to fill distribution channels with such a SKU. Keep in mind that chips are run by the wafer, and things don't get interesting until you run about 25 wafers in a single lot, so you're looking for someone to underwrite the cost for upwards of 100k devices until a sustainable market for such fully updateable chips materializes.

This is where the "obviousness" of an open source approach comes crashing into the reality of market-driven hardware economics. It seems pretty obvious to people like you and me that we should have fully updateable SKUs that can be wiped and provisioned by end users. But there's a chicken-and-egg problem of building the hardware, so that OSS developers can write the software stack, then hopefully getting adoption by volume businesses; but volume business wouldn't adopt until the hardware was cheap enough and the software stable enough.

The angle I'm hoping to exploit is that because the chips are done with no OTP and only ReRAM, we can release a fully-open and updateable SKU as a late-binding wafer sort option. So instead of having to buy a whole lot of wafers, we can split out a portion of a larger run as the "open source developer's" SKU, at a volume that could be underwritten with say, a crowdfunding campaign.

Thus, you'd have the same design, fabricated using the same masks and literally on the same wafer, but based on how some ReRAM bits are configured in the factory, one will accept updates, and the other will not. It's a compromise, but that's the angle I'm trying to work, at the very least. And until someone is willing to step in with a few million bucks to make the demands of other paying customers irrelevant, it's maybe the only practical approach I can see at the moment to break the chicken and egg cycle.

I think the team is on-board with the approach, but there's a lot of technical details to work out, which is why there is an "if" qualification. For example, if you totally wipe all the ReRAM, but you need some extant code to write ReRAM in the first place, you again have a chicken and egg problem. This problem is solvable at wafer sort using probe pads that can put the initial code in place to allow bootstrapping, but how to extend this to a convenient user-accessible solution is an open problem. This is particularly troublesome for SKUs that are destined to be packaged in minimal packages, e.g. an 8-pin SO package or a smart card format, where you simply lack the pins to expose a provisioning port. This is also, unfortunately, where all the volume is.

It could be that the open source SKU simply comes in a package with more pins and we burn precious I/O to expose something like a permanent JTAG interface. Muxing over the JTAG pins to other I/O can also help mitigate the challenge, but it adds complexity and verification/shipment risk if the mux circuit is defective, and again, a hardware bug costs millions of dollars in this scenario so no decision can be considered "trivial". Nobody wins if I try to force some clever provisioning solution that turns out to be broken (also keep in mind that smart card I/O is weird and fussy, it's not a generic CMOS I/O cell).

Also keep in mind that creating a version with a larger package could be a non-starter if we don't also have demand from a "paying" customer to fund the R&D and supply chain to push out that package format. Chip packages aren't free, and also come with substantial up-front costs and high minimum volume requirements.

bunnie··on Towards a More Open Secure Element Chip
> The mention of using ReRAM is interesting, I wonder what's motivating that.

Your previous paragraph succinctly addresses the motivation: high performance processes can't use flash. However, ReRAM is already qualified for use in these processes. Since ReRAM is non-volatile (with a storage lifetime measured in 100+ years) and dense, you can get the benefits of flash and of OTP, and also in a high performance process. By implementing "OTP" as a bank of ReRAM (with a write-protect bit, but still an erase-all mode), the option to do a full device wipe-down is theoretically available with an OTP-like behavior.

bunnie··on Towards a More Open Secure Element Chip
Stealth mode simply refers to a stage in a company's evolution, not an on-going state of operation.

For example, because the company is not yet fully formed, it doesn't have a website -- nobody has made one, because it's not the most pressing thing to do right now. I don't even think the company has an official logo yet.

They are basically walking and chewing gum by trying to put together their advisory board, while they also put together the details of the corporate entity. In essence most startups go through a "stealth mode" phase because many disclosable details simply do not exist. It's perfectly consistent to say "we want to do open source", but also not have prepared a logo, a website, and an overall corporate narrative.

bunnie··on Towards a More Open Secure Element Chip
One goal of an open-SDK SE is for all the base primitives to be open, precisely because stuff like GG-20 gets broken. I don't think it's realistic to think that we're at a point where we could burn such a novel and complicated class of algorithms into a chip, and expect it to have no revisions.

The specific citation was provided because MPC algorithms span a pretty big range of computational difficulty, depending on the features you require and the primitives you assume. Narrowing it down to an exemplar algorithm (even if broken) helps pin the analysis to a given complexity range.

It's also interesting to note that some MPC algorithms assume some amount of secure and real-time communication with an oracle to assist with key distribution, which means you're possibly having to stuff a TLS stack into the SE.

← PreviousPage 4 of 7Next →