Extend all tails of the bell curve. Daily nerd snipe(s). Robotics. Photonics. RF/microwave. Nanofab. KO6FFN. Twitter as an open journal.
SF Bay Area
- Tweets 22,467
- Following 1,691
- Followers 19,046
- Likes 70,169
Pinned Tweet
[Silicon photonics and co-packaged optics: what is inside the box?](outside5sigma.com/tutorial/s…)
Great photo from @sokol_cc for the greatest intel 100G CWDM4 SiPho transceivers, and cover image is also intel's optical engine.
I should put in more quiz next time to make it a proper tutorial.
11
17
190
20,642
heterogeneous integrated TFLN on 8 in SiPho from CUMEC. CUMEC has appeared enough for me maybe I should figure out how many wafers do they make per month.
EO modulator has 3 dB bw around 100 GHz, PD around 50 GHz. The LN waveguide is so fat thus such long modulators. Vertical coupler IL is 0.1 dB. Edge coupler is pathetic 1.6 dB IL.
1
1
14
684
TiN etch stop used for defining the trench for bonding.
- Wu2026: [Heterogeneous Back-End-of-Line Integration of Thin-Film Lithium Niobate on Active Silicon Photonics for Single-Chip Optical Transceivers](doi.org/10.1002/lpor.71344)
3
250
another high bw TFLN EM modulator demo, but the more interesting design is this dual band grating coupler that sends C band one direction and O band the other direction, and a reflection based mux.
the sim looks like they have higher order modes along the thickness direction tho, 300 nm SiN + 300 nm LN is too thick.
1
4
45
1,781
- Wang2026: [Dual-band silicon nitride-loaded lithium niobate electro-optic modulator with a record-breaking speed of 600 Gbps](doi.org/10.1364/OPTICA.60983…)
microwave loss of the CL-TWE is 12 dB/cm at 100 GHz, they need to switch to thick BOx or quartz substrate.
TE1 slab mode will start appearing at ~470 nm of total thickness, 300+300 definitely too thick.
the GC's dual band directionality is better than id expect.
1
6
511
only in photonic computing will you see people brag about lower speed, oh only the activation data needs to be fast and the programmable weights can be slow? Can you fit all the weights onto your chip? Will you ever be able to? Still want to load your weights slow? Have you thought about why you have this much controls and IOs for a mere 16x16 array?
cool setup tho, although those crap coax is not for the high bw modulators you claimed, wait a sec, you only ran them at only 1 MS/s?
3
5
63
3,015
- Ren2026: [A native signed incoherent photonic matrix core on thin-film lithium niobatesupporting in situ backpropagation](arxiv.org/abs/2609.27709)
1
6
548
we are dreaming too small. The total number of agents will double every 9 months, and number of agents per user will double every 2 years, millions of agents in 2024, trillions of agents in 10 years. We'll need EDA tools for harness architecture.
the problem is most people just consume. Do you need hundreds of agent swarm to consume? I guess if you consume things generated on the fly will create that much demand.
11
9
122
7,325
"to show you the power of programmable quantum photonic processor in orbit,"
*LOUD ANGLE GRINDER NOISES*
"I sawed this lens in half!"
3
3
48
2,087
impressive they mounted all these optics and fiber couplers onto a carbon fiber board.
when in doubt, use more epoxy.
the PIC involved is old school fs laser direct write in borosilicate glass
- Steiner2026: [In-orbit operation of a programmable quantum photonic processor](arxiv.org/abs/2609.25248)
1
11
728
it's a paper about the Center for Ubiquitous Connectivity (CUbiC) at Columbia
- Bergman2026: [Beyond the Bit Pipe: CUbiC’s Ubiquitous Edge-to-Cloud Connectivity for AI Infrastructure](doi.org/10.1109/MM.2026.3732…)
2
4
826
the more I chat with packaging people the more respect I have for cerebras, their IO in and out of the wafer sucks, but their power delivery absolutely mogs CoWoS or EMIB-T, so much cleaner than these clown show with amateur clowns. Good luck to those powerpoint warriors that dream up as many dies and as much power without any basic thermal and PI checks, and then scrambling to try fixing their broken supply chain with broken capabilities.
3
5
102
7,306
real nice and clean photonic chips before dicing, but how are they dicing this, doing it twice?
these are very low power phase modulators with heterogeneous InGaAsP on silicon.
7
13
130
4,007
maybe the stealth dicing just works if you shoot along the lanes and pull them apart, never done it like this myself, feels risky..
1
6
670
the guy leading this AI4Materials team was a stanford chemistry PhD, had his battery startup in the US, and decided it's better to head back
From a research question to candidate materials.
Xiaomi MiMo-V2.6-Pro worked with Xiaomi’s materials researchers to explore new MOF designs for capturing PFAS — reviewing literature and patents, proposing and checking design hypotheses, and running computational “dry-lab” experiments to identify promising candidates for further validation.
The team estimates a 10× productivity gain, shortening the R&D cycle from one month to 2–3 days.
As Prof. Jinhu Dou of Peking University put it: “In my view, its work on this project—from literature review to materials design and computational evaluation—was on par with that of a well-trained doctoral researcher.”
Read the full case study:
mimo.xiaomi.com/blog/mimo-v2…
#XiaomiMiMo #XiaomiMiMoV26 #AIforScience #XiaomiAI
6
52
969
79,046