Andrew Zonenberg

@azonenberg@ioc.exchange
2.8K Followers
391 Following
18.5K Posts

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.

ngscopeclienthttps://www.ngscopeclient.org/
Bloghttps://serd.es
LocationSeattle area
GitHubhttps://github.com/azonenberg
Andrew Zonenberg2h ago

The time has come!

My term as Conebearer is ending and I have the honor of passing the cone on to the next worthy recipient (a friend's gf is getting bottom surgery a week from tomorrow). I wonder where it will go next?

Show threadAndrew Zonenberg1d ago

Well, they're not final.

The front-side launches all look solid (better than +/- 10% impedance match in TDR, better than -17 dB return loss out from 10 MHz to 30 GHz).

But the back side BGA launch is about 20% low on impedance still, and while the ARF6 launch isn't done re-running it's only slightly better than the BGA. I probably need to increase the clearance from the vias to the internal ground planes more.

On the plus side after going back and forth with Sonnet support to work around some bugs in the newly released custom palette code, I have Viridis working in the current density viewer.

Andrew Zonenberg1d ago
First time I've seen a "no vomiting" sign in a taxi.
Andrew Zonenberg1d ago
Bus display is missing a bootloader...
Andrew Zonenberg5d ago

REcon talk done! Seemed to be well received.

(Photos courtesy of @david_rysk )

Andrew Zonenberg5d ago
Apparently if you know the right folks to call you can just order up pubs by the TEU.
Andrew Zonenberg6d ago

Taken just a few minutes ago. Where?

CW guesses or answers. Be as precise as possible.

#osint #geolocation #CatchMeIfYouCan

Andrew Zonenberg6d ago
Found the YVR shörk repository
Andrew ZonenbergJun 25

The full frequency sweep won't finish before I leave for REcon but I think I've finally got the ARF6 launch dialed in: better than -20 dB return loss from DC to 11.9 GHz, and better than -15.8 dB from DC to 30 GHz.

* Shrink vias from 0.45 pad / 0.25 drill to 0.35 / 0.15
* Use round vias for the signal lines as coupling seemed to be significant vs the square ones I tried to use to simplify meshing
* Keep non-functional via pads
* Single large ground plane cutout on L2 to compensate for pad capacitance
* No plane cutouts on lower layers
* About 0.175 mm via to plane clearance
* Add dummy feedlines to the pads so we can use box-wall ports in the simulation

When I get back I need to back-annotate the changes from the Sonnet model into KiCAD, re export new gerbers, and see how the final sim looks.

Also, I figured out how to get proper Viridis working in the current density view. After talking to support there's a custom/user defined palette option. It's a bit flaky since it's a new feature and I have a ticket open with several bug reports, but once you get it set up right it looks much better than the classic rainbow.

Andrew ZonenbergJun 25

In between other stuff I've been running sims to dial in the ARC6 connector launch to back-side diff pairs.

After sweeping lots of variations I've concluded it's impossible to get it matched using 0.45 / 0.25mm ViP. You can't get above about 90 ohm Zdiff no matter how far you pull back the ground planes, just because of coupling between the vias themselves.

Orange trace is a first test using 0.35 / 0.15mm ViP which seems much happier. Slightly overshot so I need a smaller plane void but i've got it bracketed.

Also, the coupling between vias at this pitch is strong enough that the square-via approximation is *not* valid, round vias give substantially different results from square ones. That threw me off for a bit.