Andrew Zonenberg

Security and open source at the hardware/software interface. Embedded sec @ IOActive. Lead dev of ngscopeclient/libscopehal. GHz probe designer. Open source networking hardware. "So others may live"

Toots searchable on tootfinder.

ngscopeclient: https://www.ngscopeclient.org/
Blog: https://serd.es
Location: Seattle area
GitHub: https://github.com/azonenberg

Public Key

npub1wpgh8qdt8suzxy8f2luspksj4wpdfw53weq4v8dr7llqp3t72tds3rf386

NIP-05 Address

azonenberg_at_ioc.exchange@momostr.pink

Profile Code

nprofile1qqs8q5tnsx4ncwprzr5407gqmgf2hqk5h2ghvs2krk3l0lsqc4l99kcpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0ds306xct

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1wpgh8qdt8suzxy8f2luspksj4wpdfw53weq4v8dr7llqp3t72tds3rf386

Show more details

Last Notes

2026-06-18 15:15:36 UTC

- reply

Yeah they don't charge a registration fee so this would be all me.
My gut feeling is that for the amount I would potentially charge and this being most likely a one-time thing or at best once a year, it's probably not worth the hassle.

2026-06-18 14:59:59 UTC

EU folks: What are the tax etc. implications if I, as a non EU resident/citizen with no existing business connection to the EU, want to offer a paid training at a conference in Europe? Do I have to collect VAT or sales tax or something along those lines? Or if I'm not a European citizen/business or collecting less than X million EUR in the tax year am I tax-exempt?
Is this such a paperwork nightmare I shouldn't even attempt it unless an existing EU entity wants to handle all of this and just contract with me for my time?

2026-06-16 02:16:42 UTC

- reply

instructions unclear have root on my guitar now

2026-06-06 18:01:49 UTC

- reply

curious: is anyone in the US and/or EU pushing legislation to ban commercial residential proxies?

2026-06-05 16:22:34 UTC

Soo according to Anthropic (https://www.anthropic.com/institute/recursive-self-improvement) they are thinking of slowing or halting frontier model development, but only if their competition does so as well and they have a nuclear-disarmament-style inspection process to verify this is the case.
There's two possible reads on this, of course.
The cover story is that they are worried about the models self-improving and becoming God and running away etc. Because they're vibe-coding 80% of their new stuff and model collapse is totally not a real thing.
My preferred interpretation is that training costs are getting out of hand, they are terrified of shareholders finding out how much they are spending on training when they go public, so they want to cut back but they're afraid of falling behind and losing their edge so the only way they're willing to do so is if they can be assured the competition isn't out-spending them.
There has to be an antitrust case here if somebody is willing to bring one (good luck in today's political climate but hey I can dream)

2026-06-04 05:29:40 UTC

- reply

gluten free filament

2026-05-28 00:18:44 UTC

- reply

The big cisco is emptying out as i move stuff over to 40g or 4x 10g breakouts.
Pulling sooo many 10g optics. I might have to start unloading 10g sfps and only keep 10/25g dual rate capable ones
https://files.ioc.exchange/iocexchange/media_attachments/files/116/649/325/263/080/713/original/29a9eeaea003078a.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/649/325/686/710/653/original/063c14c71c4a6e0e.jpg

2026-05-27 23:55:32 UTC

- reply

Now to dig up all of those MPO to LC breakout cables I've had lying around unused...

2026-05-27 23:41:28 UTC

New switch setup time
https://files.ioc.exchange/iocexchange/media_attachments/files/116/649/180/260/162/528/original/ab25ea985247a0f2.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/649/180/659/266/031/original/9d469af465bfa29f.jpg

2026-05-25 04:53:03 UTC

- reply

These also look like poly resistors but are a different color, maybe they have two poly layers with different silicide doping or one of them is a resistive metal layer or something
https://files.ioc.exchange/iocexchange/media_attachments/files/116/633/420/644/746/592/original/6c07eec26092654e.jpg

2026-05-25 04:52:14 UTC

- reply

Poly resistors
https://files.ioc.exchange/iocexchange/media_attachments/files/116/633/417/700/118/874/original/7380b3feab6578f3.jpg

2026-05-25 04:51:38 UTC

- reply

Random logic showing 3 metal layers and what looks like maybe poly resistors
https://files.ioc.exchange/iocexchange/media_attachments/files/116/633/415/490/202/210/original/b338bbb8680fdf71.jpg

2026-05-25 04:50:51 UTC

- reply

I've figured out what foundry made a lot of chips based on various hints from fiducials and metal fill etc.
But this is the first time I've seen someone actually write the foundry and process on the mask set. Makes the ID simple lol
https://files.ioc.exchange/iocexchange/media_attachments/files/116/633/410/905/714/785/original/48c1029f071eccf2.jpg

2026-05-25 04:49:52 UTC

- reply

I wonder if any of these people still work at Qualcomm?
https://files.ioc.exchange/iocexchange/media_attachments/files/116/633/407/095/150/397/original/d6c86627ab5c9cd6.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/633/407/672/255/366/original/77e7c88893a12827.jpg

2026-05-25 04:48:50 UTC

- reply

Took a quick break but the PM7500 is done, cleaned, and imaged.
Sure looks like a PMIC. Lots of big metal polygons and multi-bonded nets that are probably inputs and outputs for LDOs and buck converters.
Made on an older, 3-metal planarized process.
https://siliconprawn.org/map/qualcomm/pm7500/azonenberg_mz_mit20x/
https://files.ioc.exchange/iocexchange/media_attachments/files/116/633/400/067/747/528/original/d1e703eab9dea71d.jpg

2026-05-25 02:38:21 UTC

- reply

A more normal amount of NO2 generation (now that I feel safe taking off my gloves and letting it sit for a bit, i can take pics)
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/890/556/635/643/original/7080b31c7684861e.jpg

2026-05-25 02:22:22 UTC

- reply

For reference usually I see very thin wisps of NO2 at the lip of the beaker by the watch glass that become invisible as soon as they mix with the air, this was throwing out enough that I saw visible brown puffs moving up from the beaker and remaining visible all the way up to the top of the hood. Faint, but visible.

2026-05-25 02:20:32 UTC

- reply

Like, I knew it was going to react - my plan was to let it sit there at room temp until most of the copper was gone, and only begin heating to attack the organics after I had taken off the copper.
But with this much copper it seemed to be entering thermal runaway, self-heating and increasing the rate of reaction to the point that I was worried about whether the filtration in the fume hood would be able to cope with the amount of NO2 it was throwing off.
Maybe I should get like a cold water bath or something to put the beaker in when I resume? I wasn't expecting it to get this spicy at this scale (~5 ml of acid iirc).

2026-05-25 02:17:41 UTC

- reply

Hmm, I perhaps could have seen this coming.
As soon as I added the acid to the VSC8512, with the large copper plane that I had sanded the soldermask off of to image, it started reacting aggressively even at room temperature and throwing off more NO2 than I was comfortable with.
I ended up pouring the acid off to a separate beaker to let things calm down a bit. Will give it a few minutes then see how much copper is left... I might want to do a pre-etch with a dilute nitric solution or something so i can remove the remainder of it without it getting too energetic.

2026-05-25 02:08:32 UTC

- reply

Everything but the acid is prepped. Time to glove up and get spicy, won't be posting for a while
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/771/690/746/579/original/093ab2fc759b98db.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/772/015/353/619/original/30cbdb7278f5d7c9.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/772/404/448/135/original/639565018b347119.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/772/740/790/601/original/0154e1645117300e.jpg

2026-05-25 02:03:45 UTC

- reply

And with that I think I have enough components to keep this wet lab session busy.
Time to fire up the fume hood...

2026-05-25 01:19:38 UTC

- reply

Next, probably also from the same device, a Qualcomm PM7500 PMIC.
112 pin dual row QFN (gross)
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/541/322/716/762/original/f4e0b1b2488bcb1c.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/579/117/846/179/original/b19a2f48847724e8.jpg

2026-05-25 01:02:22 UTC

- reply

Next is another Qualcomm part with a very similar part number, the RFR6500. Seems to be an early generation CDMA receiver from circa 2005. This must have come out of one of my old phones, although I don't think I ever used a CDMA based carrier in that time frame (T-Mobile was GSM based).
I had a Samsung SGH-T439 as my first phone, perhaps this was the chipset from it?
Nothing noteworthy package wise, just a 56 pin punch singulated QFN.
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/507/667/292/448/original/d2471a98afa98881.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/507/750/494/315/original/4b247f7e9408bddd.jpg

2026-05-25 00:45:34 UTC

- reply

Next is a Qualcomm MSM7500, seems to be part of a very old phone chipset with an ARM1136JF-S.
Maybe from an embedded cellular module or something? It was sitting as a loose chip in my microscope-food bin and I have no record of its origin.
It's very sticky on one side, some kind of adhesive or thermal compound residue that didn't come off with acetone. I have no special interest in the part so didn't bother trying to clean it further (plus it wasn't obscuring any markings)
The underside is... weird. It's 28 balls square with three concentric rings, each consisting of two rows of balls then a void space, then a bunch of power/ground balls in the middle. This much is normal.
But there's some strange gaps in the soldermask between the rings, and almost all of the balls seem to have connections going to these void areas. I wonder if there is some kind of test point or plating ground or something in this area that is removed late in the packaging process? Never seen anything like this before.
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/403/499/735/449/original/19abf8bfd92a7858.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/439/257/803/271/original/19b6b1928b3f40e5.jpg

2026-05-25 00:27:40 UTC

Prepping some chips for a decap run as I do the lab cleanout.
First on deck (other than the VSC8512 I prepped a while back) is a DisplayLink DL-5900. Seems to be some kind of USB displayport adapter? No idea where it came from.
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/374/136/542/262/original/cbe49ccbffb25c9c.jpg
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/374/104/050/712/original/e48530d0873bef38.jpg

2026-05-25 00:06:49 UTC

Brochure from a vendor found while cleaning out the lab.
Presented without comment
https://files.ioc.exchange/iocexchange/media_attachments/files/116/632/280/472/838/549/original/91abce208549af63.jpg

2026-05-21 17:37:28 UTC

STMicro finally released a security advisory related to a silicon errata I found last year so I can talk more freely about it https://www.st.com/resource/en/security_bulletin/sb0070-risaf-limitation-about-pcie-region-protection-on-stm32mp25xx-stmicroelectronics.pdf
Tl;dr the STM32MP2 AXI firewall (RISAF) does not fully block transactions to the PCIe region that are prohibited by register settings.
Writes have the data replaced with 0x00, but still execute (so an attacker who should not be able to touch the device can write zeroes to any location in the BAR of their choice).
Reads have the response data replaced with 0x00 as seen on the bus, but any side effects of the read like clearing flags or popping a fifo will still execute.
In short, it seems to behave as though the AXI data channel is blocked but not the control channel.
The security violation interrupt still fires so the trustzone monitor will be aware of the issue, but depending on the application it may be too late since the register write already happened.
I found this by accident while trying to bring up the STM32MP2 bare metal and couldn't understand why all of my PCIe accesses were acting funny. So if anyone is wondering why I went dark all of a sudden on that side project, now you know why.

2026-05-15 03:04:08 UTC

Fun engineering/privacy problem: design a device that plugs into an ESP32 based sensor node that has a microphone and reports ambient noise levels in dBa via MQTT or similar.
That's the easy part.
The hard part is to do so in a way that you can prove it's incapable of being used as a bug/listening device even in the worst case scenario (attacker running arbitrary code on any programmable component).
My initial thought is to record audio from a MEMS mic into a tiny MCU with limited RAM capacity, then report results to the main system processor through some kind of R-C low pass filter that only lets like one bit per second pass through.
Or maybe a bank of analog filters and peak hold blocks? I want proper weighted dBa.

2026-05-14 02:37:52 UTC

- reply

have you left a wrench on top of the mill drawbar yet?

2026-05-13 02:35:41 UTC

- reply

this is what happens when a person who doesn't eat any animal products rubs a balloon through their hair and builds up an electrostatic charge

2026-05-10 13:45:08 UTC

Can't link to full 800mpix res because siliconprawn is being derpy, but this is clear enough to see all of the traces so good enough for now.
Substrate of a VSC8512 seen from the underside, after sanding off solder mask. The silver marks are solder residue showing where the BGA balls had been.
This image is mirrored to match the datasheet orientation, as if you were looking at the substrate from the top of the package. Ball A1 is at the northwest corner.
https://files.ioc.exchange/iocexchange/media_attachments/files/116/550/576/368/043/303/original/61f4e52f140953e0.jpg

2026-05-02 16:41:49 UTC

- reply

The optic is NOS straight from the tray
https://files.ioc.exchange/iocexchange/media_attachments/files/116/505/974/506/288/189/original/6b1a487385b7da63.jpg

2026-05-02 16:41:03 UTC

Did i put this in backwards or something? The fiber isn't clicking
https://files.ioc.exchange/iocexchange/media_attachments/files/116/505/962/353/332/164/original/ec525353f8d27397.jpg

2026-04-27 20:23:45 UTC

- reply

and "Penelope" with "cantaloupe".

2026-04-27 20:22:44 UTC

- reply

don't forget "gorilla" rhymes with "quesadilla" too

2026-04-21 22:32:58 UTC

The existence of the New York Post implies the existence of the New York GET and HEAD

2026-04-19 07:52:28 UTC

- reply

Pulled all but one dimm out of each. No change in behavior

2026-04-19 07:42:23 UTC

- reply

oh. And the machine that is bootlooping originally came up after reboot enough that i could ping and start an ssh connection but not make it all the way to a shell. No idea what was going on there, had no visibility into the failed state

2026-04-19 07:35:23 UTC

- reply

Already pulled the GPUs from both, both because debugging supermicro stuff is easier using the VGA port, and to reduce PSU load in case that was part of the problem.

2026-04-19 07:34:58 UTC

- reply

The two media machines, also running Trixie with the same filesystem mounted, are unaffected.
The two failed machines both have Supermicro X12SPI-TF mainboards. One has a recently acquired secondhand Xeon 8362, the other a 5320 that had previously lived in my VM server for the last few years and had no problems.
RAM on both is recently installed, but hand-me-downs from other machines with no errors during POST or operation. Everything is ECC.

2026-04-19 07:31:16 UTC

- reply

The crash hit all 3 of my physical workstations running Debian Trixie, but not the VM running Trixie, or the xcp-ng VM server, both of which also had the same Ceph filesystem mounted.
And one of the three booted up fine afterwards with no apparent ill effects. Thankfully it's my office machine, the one I spend the most time sitting at, but it's also not in the lab so losing *both* of my lab workstations is going to be a big hit to productivity until I figure out what happened and fix it.

2026-04-19 07:25:55 UTC

- reply

One of them is hanging in DXE doing USB initialization, and the other is able to get into the UEFI configuration menu, the boot device selection menu, but when I select either the internal disk or a removeable Debian live disk it thinks for a few seconds then reboots.

2026-04-19 07:24:59 UTC

OK so... on my surviving workstation, the crash dump shows it was a null deref in the Ceph kernel module in ceph_mds_check_access.
Annoying, but this should have been trivially fixed with a reboot. On this machine, it was.
So why are my other two workstations not coming up?
https://files.ioc.exchange/iocexchange/media_attachments/files/116/430/173/171/170/352/original/444e7bb828def51a.png

2026-04-16 18:23:41 UTC

Randomly reminded of the time that me and my brothers decided that a hypothetical mixture of Coca-Cola and toilet paper (which as far as I know was never actually created) needed a name, and dubbed it "garblax".

2026-03-28 00:07:13 UTC

Zombie: "Braaaaains"
Fedi user: "Traaaaaains"

2026-03-26 13:59:33 UTC

The existence of toblerone implies the existence of related triangular hydrochocolate chains such as the base alkane (toblerane), as well as the closed ring form (cyclotoblerone)

2026-03-21 17:11:18 UTC

- reply

it took my wife a long time to train me not to wear shoes around the house. My childhood home had enough legos and random junk on the floor that if you went barefoot around the house it would rapidly prove a painful mistake.
I used to keep my shoes next to the bed when sleeping and pretty much only take them off to sleep or shower

2026-03-19 15:52:59 UTC

Every time I post a bunch about crazy inner layer rework I get a couple of reactions from people who think that being this good at bodging boards makes me some kind of EE god.
Nothing could be further from the truth.
You know how you get this good at fixing board design errors? Yeah, exactly how you'd think. Making a lot of them :P

2026-03-19 15:43:35 UTC

- reply

What's the return loss of the kitty?

2026-03-09 18:16:27 UTC

- reply

every time I hear "hugging face" I think of xenomorphs and no slop company propaganda will ever change that

2026-03-06 22:58:15 UTC

Wrote up the blog-format version of the TCXO failure analysis including a bit more backstory and commentary all in one place
https://serd.es/2026/03/06/TCXO-failure-analysis.html

2026-02-26 04:50:26 UTC

- reply

Chat can i retire early?
https://files.ioc.exchange/iocexchange/media_attachments/files/116/135/123/879/205/359/original/82d77dc60068e2a6.jpg

2026-02-26 04:37:51 UTC

VM server going down for the third attempt at updating to xcp-ng 8.3 after two previous failures resulted in rollbacks and changes to my SR configuration to hopefully be compatible with 8.3.
Also doing the first of two rounds of hardware upgrades after I scored a bunch of nice stuff from some servers that were headed to the scrapyard.

2026-02-21 00:02:39 UTC

Anybody have a non emergency service contact for Puget Sound Energy? They have a contact to report outages and downed wires but not lower priority issues.
We've pretty definitively figured out what's going on at my friends' place: intermittent brownouts on one of the two 120V phases affecting the entire house but not neighbors. Waveforms show glitches in the sine right above and below the zero crossing consistent with series arcing. And it happens more often in windy weather.
All of the evidence points to one crimp on the secondary side of the transformer going bad. Just a matter of figuring out how to get a lineman out on the pole to replace it.

2026-02-16 18:36:43 UTC

- reply

the whole circuit is switched by a single 120V SPST switch controlling all four bulbs wired in parallel. There is to my knowledge no other switch.
And the capacitors are inside the LED bulbs as part of the SMPS circuit.

2026-02-16 18:20:51 UTC

- reply

all four bulbs are in parallel and the other 3 turn off fine. It's definitely discharging stored energy in a capacitor

2026-02-16 17:39:55 UTC

Someone please explain this to me
https://files.ioc.exchange/iocexchange/media_attachments/files/116/081/524/037/393/749/original/17aaa77ee52cb062.mp4

2026-02-11 20:55:35 UTC

- reply

E2EE for open groups is pointless, an adversary can just join your group and see everything. No cryptanalysis and no hacking required.
For small closed groups, use signal.
The niche we need to fill is large public group chats with a better UX than IRC.

2026-02-11 20:54:43 UTC

- reply

What I want to see is basically IRC + formatting fixes, reply chains, and image embedding for group chats, plus strong binding of user accounts to signal accounts and a nice UX for clicking on a user in a group chat and jumping into a signal DM with them

2026-01-27 01:58:34 UTC

Steps to greatly advance the state of the art in any relatively obscure field:
1) Find a bunch of autistic nerds whose special interests broadly overlap with said field
2) Put them in contact and let them share ideas or, better yet, lab space
3) Buy them whatever they ask for
4) Wait and see what they come up with. Excitement guaranteed, you just have no idea exactly what kind.

2026-01-01 20:28:42 UTC

- reply

Has any state made serious efforts towards outlawing the data broker industry outright?
As in, rather than forcing individual residents to opt out of data collection, mandate that as soon as you become aware someone is a California resident they implicitly treat that as a deletion request.
And banning the operation of data brokers within state lines (even if not handling CA resident data) including employing CA residents, selling data to CA companies, etc.

2025-12-27 05:52:30 UTC

- reply

If you want Good BNCs, get Rosenberger.
They'll cost a fortune, but they have this incredibly smooth feel to them - snug, but no burrs or binding or sticking. Some of them are rated to 10 GHz which is just ludicrous for a BNC. Insertion and return loss performance are excellent.

2025-12-24 23:06:02 UTC

- reply

the cockroaches will be IRCing over HF radio after whatever apocalypse wipes out humanity.
It's the messenger that will outlive all others. Like cockroaches it's not pretty but also nearly impossible to kill.

2025-12-10 22:51:29 UTC

Random thought: the centralization of authentication to a few big OAuth providers like MS and Google, combined with services that time out your cookies and force relogins every so often, makes phishing people so much easier.
Want someone's account creds? Just pop up something that looks like a ms or google login form, odds are they're so conditioned by login fatigue that they'll automatically type their creds and TOTP token into it.

2025-12-05 16:33:45 UTC

Seeing all of the Flock posts lately got me wondering: is there a privacy-preserving way to accomplish the public benefit of traffic/speed cameras (reducing speeding and red light running) without also turning them into blanket ALPR spies tracking where everyone goes?
The best idea I came up with is a split system architecture that treats the camera as adversarial.

2025-11-24 08:33:21 UTC

- reply

After a while it didn't even make sense or refer to any of the stories at all.
(long list of European kings)
B: "King Henry the 8th"
M: "King Henry the 9th"
(we ran out of Europeans and switched to the Zulu)
B: Shaka
M: Dingane
(list went on until we couldn't remember any more)
Eventually the ritual would end and we'd actually have a real conversation about which to play.
Is this a normal kid thing or just a neurospicy version? Lol.

2025-11-24 08:30:58 UTC

- reply

M: "Well, I don't see why not. What type?"
B: "Well, I don't see why not. What type?"
M: (chanting) "Pooper-scooper kitty 1945" (this had started out as "super kitty 1945")
B: "No-no-no-no-no-no-no-no-no" (he was tired of it because it was my favorite for a long time)
My memory gets fuzzy but we went on for several minutes like this suggesting various ideas and being shot down.

2025-11-24 08:29:00 UTC

- reply

Towards the end it went something like this:
Brother: "Your Majesty, wilt thou express thy intentions to do a P. if there are any. Or, if there are not any, wilt thou express thy intentions to turn into a hippopotamus and fly away?"
Me: (chanting) "Could we turn into a hippopotamus and fly away?"
(pause)
M: "Could we do a Puh?"
B: "Duh"
M: "Sku"
B: "Sting"
M: (chanting) "Could we do a Puh-Duh-Sku-Sting?"
B: (chanting) "Could we do a Puh-Duh-Sku-Sting?"

2025-11-24 08:26:23 UTC

- reply

But before we could do any of these, we had to agree on which one to play that night.
This became a very long, involved ritual that was constantly having new verses added.

2025-11-24 08:25:13 UTC

- reply

Another involved "super kitty", a cat on a Pacific island in late WW2 trying to fend off a Japanese invasion, I think with the aid of other cats?
And then there was the KWSTG ("killer whale and sea turtle game") involving me as the sea turtle and my brother as his orca friend adventuring in the ocean.

2025-11-24 08:24:08 UTC

- reply

There were a bunch of different storylines that continued for years in parallel. Nearly all involved intelligent animals, but closer to Redwall / Watership Down style (i.e. human level intelligence but physically unchanged) rather than anthro type.
The original, which became the namesake for them all, involved prairie dogs or "P.D." (pronounced "puh duh" or just "puh" for short). I forget most of the details of that one.

2025-11-24 08:21:28 UTC

When I was probably 12ish and the older of my younger brothers was around 10, we all shared a bedroom with me on the top bunk and both of them on the bottom (it was a twin + double stack meant to sleep 3).
If we weren't ready to sleep yet, we'd pass the time with D&D-esque verbal roleplaying games (no dice or anything, story flow purely driven by him as the DM and remaining characters).

2025-11-10 18:18:01 UTC

- reply

You don't even need tape, big hard drives are cheap and you can hook them up to the cheapest i3 motherboard on the market if you just need to use it for backup storage over the internet.
My offsite backup is six of the largest cost-effective hard drives (4TB at the time I set it up) at a friend's house; we have a "poor man's colo" arrangement hosting offsite backups for each other and peering over OpenVPN.
Large spinning rust is not *quite* as dense as tape, but more than good enough for 99% of users.

2025-11-04 02:34:23 UTC

- reply

you can also buy individual sleeves (although sold in pairs they're not physically connected)
https://www.northernsafety.com/Product/16751/Honeywell-North-Silver-Shield-20-Chemical-Resistant-Sleeves-Pair

2025-10-22 18:35:18 UTC

Does anyone have (freely available non-NDA encumbered, i.e. reverse engineered) area figures for ARM Cortex-A processors, ideally on TSMC 28nm?
Even rough single data points would be helpful.
I have some Zynq-7s (28HPL dual A9) in my microscope food bin, but it seems like nobody's published an xc7z substrate die shot yet and it might be a while before I get to mine.

2025-09-30 16:16:01 UTC

- reply

I'm amazed more people don't understand that having your entire economy dependent on an expensive commodity with highly volatile pricing, that's primarily produced by countries hostile to your interests, which are always fighting each other, is a significant national security risk.

2025-09-28 15:42:51 UTC

- reply

Except that most AI GPUs have no video output.
Most of them can probably be repurposed for scientific computing purposes, although they may not perform great at that (I'm not sure about the ratio of float32 to float64 to neural network performance in e.g. one of the big Blackwells compared to a current generation workstation/HPC GPU).
For anything that needs video output, they're useless.
I am starting to look into multi-GPU for ngscopeclient to see how practical using a headless GPU (or just one other than the one your display is plugged into) for all the number crunching, then copying the final rasterized waveform textures to the display GPU for final video output.

2025-09-21 16:18:14 UTC

Internet hive mind: what are these?
Options I'm considering are cellular microcells, water/power meter collectors (my leading candidate), some kind of public wifi network.
Guesstimating 900 MHz band from the antenna size as best i can guess from ground level, but I don't have a SDR on me to be sure.
https://files.ioc.exchange/media_attachments/files/115/243/178/802/003/341/original/90e0b6e745ef85a2.jpg

2025-09-11 16:07:30 UTC

The ngscopeclient team is excited to announce our first official release, version 0.1, after many years of development.
This release includes contributions from 77 people across the globe and supports instruments from 23 vendors, running on all major operating systems.
https://www.ngscopeclient.org/news/2025-09-11-v0p1-release.html

2025-09-11 00:44:18 UTC

Periodic reminder to new followers or anyone just swinging by: I have high speed digital design / SI experience and lots of nice lab toys. Most folks can't afford this kind of stuff so I try to share the love.
If your project is open source or generally noncommercial / hobby in nature, there's NO CHARGE for a quick design review or some basic lab measurements. If you send me a board for characterization and want it back, all I ask is that you send me a prepaid label or reimburse my return shipping costs.
If it's a major time commitment for me or you plan to make money from the project, I might still be interested but we'd have to discuss compensation. As a general rule I don't want to price people away; I use the day job and occasional industry-rate projects to subsidize the cost of offering steeply discounted services to folks who can't afford to pay industry rates at a "real" lab.
Capabilities include but are not limited to:
* Schematic/layout review
* 3D planar EM simulation (Sonnet)
* Oscilloscope measurements to 16 GHz
* BERT measurements to 25 Gbps NRZ
* VNA measurements to 8.5 GHz
* Metallographic cross section sawing, embedding, and polishing
* Wet chem for IC decapsulation, metallographic etching, etc
* Optical microscopy to ~300nm resolution including automated large-area scans
* UV/VIS/NIR spectroscopy
EDIT: almost forgot to add, advanced PCB rework capabilities including inner layer circuit edits and bodgewires down to 125 um and micro 100 ohm twinax for differential pair bodging

2025-09-05 03:59:17 UTC

Design review thread. I'm going to be looking at https://git.sr.ht/~mxsparks/ceres-hardware/ by @npub1lxl…pjmr from mostly a SI/PI viewpoint.
This board has already been fabbed (https://xgdr.net/2025/08/15/ceres-revA-bringup/) so I'm not going to be spending much time worrying about looking for pinout or basic "does it boot" issues.

2025-09-04 14:48:17 UTC

The existence of perjury implies the existence of juries in lower oxidation states

2025-09-04 10:03:35 UTC

Not enough developers understand the responsibilities of having a large userbase and how important preserving freedom of users is.
Every time you change the way something works, the cost of that change is borne by your entire userbase. If you have a million users and you move a button and everyone spends 30 seconds figuring out where it went, that's 8333 hours of wasted effort. At $150/hr for a somewhat average tech salary, you've cost your userbase $1.25M with that one change.
Sometimes breaking changes are unavoidable, but the more fundamental the change and the more extensive the retraining users need, the more cautious you need to be about doing it.
It's not just about your "creative expression" or some manager's vision of "the right" way to do things. If you can possibly add a preference to keep a workflow the way it does, add it.
Even today in ngscopeclient it hurts me every time I make a change that breaks a workflow. I rationalize it as "better now than later, while the userbase is small", and we're still very much in "experimental" stage in the v0.x series. But I do it fully understanding that I'm throwing a lot of people off every time I merge a dialog into something else, etc.
I don't know how many users we have since we have no analytics or telemetry, but it's enough that I'm sure every small change I make turns into thousands of dollars of wasted time at minimum.

2025-09-02 02:14:48 UTC

Kcmil has to be one of the most cursed units for area.
If you're not familiar with North American wire sizes, heavy gauge wire like power lines are often sized in kcmil (sometimes abbreviated MCM because M = roman numeral 1000, not SI "mega").
As in kilo circular mils, where a circular mil is the area of an 0.001 inch diameter circle.
Not even *square* mils. Circular mils. Because irrational units and conversion factors for area are the best

2025-08-27 09:06:07 UTC

Hmmm.
I should probably do something about this section in the documentation before we ship v0.1, right?
https://files.ioc.exchange/media_attachments/files/115/099/925/309/606/664/original/87e53f2f00beb7b0.png

2025-08-24 16:34:37 UTC

Anybody else feel like they're working the lunch counter at Subway whenever they put on sunscreen?
Especially the Coppertone that comes in the bottles that squeeze out the bottom like the Subway condiment bottles.
"Ok sir... what else on this? Black olives, Parmesan cheese..."
"And two squirts of SPF 50"

2025-08-20 05:38:24 UTC

IC reverse engineering software is absolute crack for neurospicy brains.
Every time I'm on a project for work where I get to use it, I have to force myself to put it down when I'm supposed to be done with work for the day. You can probably tell how that's working out given that it's 2230 local time.
It's the most addictive puzzle game out there by a massive margin. There's the "fix stitch errors" mini-game, the "fine tune inter layer alignment" mini game, the "optimize feature extraction" mini game, dozens to hundreds of "what is this standard cell" mini games, and then the main "wtf does this giant netlist do" storyline.
If only it wasn't dongle-locked, export-restricted software sold at EDA-tool seat prices...

2025-08-19 18:59:19 UTC

- reply

Name and shame. If not the actual company, the recruiter for not having the sense to throw the startup under the bus for being dangerously stupid.

2025-08-15 07:25:06 UTC

For all the marketing advice LinkedIn MBAs come up with, a surprising number of people forget one of the most basic steps: The Middle Schooler Test.
The process is simple:
1) Place a sample of your brochure, website, etc. in front of a recently calibrated 12-year-old boy.
2) If they start giggling uncontrollably, identify and correct the source of their amusement. Return to step 1.
3) If they aren't particularly amused, move on with the campaign.
https://files.ioc.exchange/media_attachments/files/115/031/572/687/523/103/original/12808be52ec59c9b.jpg

2025-08-11 11:22:00 UTC

Anybody else think "HBCU" sounds more like a bank than a place to get an education?

2025-07-30 19:10:00 UTC

- reply

Last time I checked old.reddit.com was free from such idiocy.

2025-07-30 19:09:23 UTC

- reply

i only go there in a browser *in a virtual machine that I don't use for anything but social media*.
Keep it locked up in a padded cell where it can't do anything I don't want it to.

2025-07-24 23:13:47 UTC

- reply

Yep.
Keep radio links for mobile deployments where you actually need the flexibility and can tolerate poor performance since the alternative is no connectivity whatsoever.
Using any kind of wireless link for a fixed site is just dumb.

2025-07-24 00:01:33 UTC

Is anyone aware of publications or research on what sort of bugs LLM-generated or LLM-assisted code tends to have?
Like, we have a huge body of knowledge in the security community about how to audit human-generated codebases for the types of bugs that human developers commonly write.
But we don't have that kind of data yet (AFAIK) for the vibe-coded monstrosities all of us are going to be pentesting soon.
Gut feelings:
* There are some common threads and patterns of errors, but they're very different from purely human-authored code
* There's a lot of subtle bugs where code looks good at a glance, but is missing some knowledge of interface behavior in another module or component that was outside the context window or something
Assumptions:
* We don't know exactly which lines in the subject codebase were written by humans and which by stochastic parrots
* The code at least appears to function correctly for a nontrivial fraction of inputs, i.e. it compiles and has been debugged sufficiently that a customer is considering shipping it

2025-07-15 20:36:28 UTC

- reply

How do you continue growing when the majority of the world's population already uses your products? Magically inventing more customers (at least, faster than the birth date) is not going to happen.

2025-07-15 04:53:15 UTC

Why does milk say "homogenized" and not "homoogenized"?

2025-07-10 05:50:08 UTC

- reply

The limited flash on the H750 should mostly be a non issue. My typical firmwares so far are barely over 128 kB all told, and I have a ton of executable RAM to work with.
So I can put a minimal bootloader and BSP (and maybe firmware update SFTP server) in the internal flash, bring up the FPGA, then copy my application image from spare space on the FPGA config flash to SRAM and jump to it.
The only other issue is that I normally run microkvs in the last few flash sectors. So in a single-sector flash part, I can't do that. Instead, I'd need to add a memory mapped interface to the FPGA-attached flash and store my config there. (I could also attach an extra QSPI flash to the MCU but that adds BOM cost, it's more pins to route, and most importantly involves dealing with the cursed OCTOSPI peripheral).
It'd take a bit of coding to get all of this done, but I think it's worth the effort. I may still go with the 753 for new designs long term but since I have 750s on the shelf, might as well use them up for one-off / low volume projects.
The larger BGA ball pitch alone is enough of a reason.

2025-07-10 05:44:11 UTC

Starting to look more seriously into the STM32H750/753.
The 753 is the part I really want to use, the 750 is the same die fused down to disable all but the first 128 kB of flash. But I picked up several of them during the chip shorage and I feel like I should put them to use.
They're mostly comparable to the H735 I use now. Key differences:
* CPU Fmax down from 550 to 480 MHz
* RAM nearly doubled from 564 to 1024 kB
* Flash increased from 1 MB to 2 MB in the 753, but reduced to 128 kB in the 750
* 40 more GPIOs (128 -> 168)
* Package enlarged from 201 ball 0.65mm (10x10 mm) to 265 ball 0.8mm (14x14 mm).
That last bit is what has me the most interested. The 0.65mm H735 package is usable on OSHPark 4L, barely, but you lose a lot of IO and the parallel memory controller etc. aren't practical to use.
While the H75x is available in the same footprint, you can also get it in the 0.8mm package which should be much more friendly to the large via sizes offered on OSHPark. So it could be a winner for prototypes where I care more about ease of access to IO and MCU-FPGA interfacing than about absolute max CPU performance.

2025-07-05 20:18:42 UTC

I just described bus timetables at a local transit center as "leading / lagging the ferry phase" in order to "provide sufficient setup/hold margin" for transfers between modes of transit.
Does that make me some kind of nerd?
It seriously is a synchronous CDC between, dare I say, buses.

2025-06-29 19:34:15 UTC

- reply

last time my blog was on the front page of HN my server was using 1% of 1 vCPU and like 128 MB of RAM.
And had a load average of like 0.01.