A recent post on zhichai.net analyzes the 25-channel ultra-low-noise single-photon-level detector module released by Beijing Zhongke Guoguang Quantum (Guoguang Quantum), targeted at the readout chain for kilo-qubit-scale superconducting quantum computing. The post cross-checks the vendor announcement against IT Home and Phoenix Tech coverage and offers both an endorsement of the underlying physics and several pointed corrections.
The thermal argument at the fridge door
Superconducting qubits operate near 20 mK inside a dilution refrigerator whose cooling budget is only about 10–50 µW at the mixing chamber. Each NbTi coaxial cable leaks 0.1–0.3 µW of heat from room temperature, and a kilo-qubit system needs roughly 2,000–3,000 readout lines—static heat load alone would exceed the budget. Optical fiber's thermal conductivity is about 1/10,000 that of copper coax, so converting readout signals to the 1550 nm telecom C-band at the cold stage and carrying them out over fiber is a physically sound strategy. The module's role is to be the room-temperature receiving terminal that catches these weak single-photon-level optical signals after electro-optic conversion, with 25 channels per module and a stacked modular architecture scaling to hundreds-to-thousands of qubits, integrated with the company's fiber array and quantum computing OS.
Technical corrections the post insists on
- It's an optical receiver, not a qubit readout device. Qubit dispersive readout uses 4–8 GHz microwave photons (~25 µeV each); 1550 nm photons carry ~0.8 eV—50,000–100,000× higher energy. SPAD/SNSPD devices detect the converted optical photons downstream, not the qubit states themselves. Qubit readout still happens via microwave resonators and parametric amplifier chains inside the fridge.
- A physical tension in the specs. Truly ultra-low-dark-count SNSPDs require 2–4 K cryogenics; InGaAs/InP SPADs that work at room temperature or TEC-cooled have kHz-level dark counts. The vendor claims both "room-temperature terminal" and "ultra-low dark counts" without giving efficiency, dark-count rate, crosstalk (dB), or channel-uniformity figures, and with no third-party evaluation. Treat these as qualitative claims until tested.
- "Most peers have only 4 channels" is anecdotal. It comes from an anonymous quote in the vendor release. Brokerage reports list Single Quantum's R12 and R48 modules at 12 and 48 channels (plus kilopixel arrays), mainly for QKD and photonic quantum computing—different segments, but enough to weaken the blanket industry claim.
- "Domestic first" is unverified, and "industry consensus / inevitable path" is vendor framing: cryo-CMOS, frequency-multiplexed readout, larger and modular multi-fridge architectures are all parallel routes, and microwave-optical transduction itself still faces pump-induced heating, duty-cycle, and pump-filtering problems. Fiber also doesn't mean zero heat load—intense optical pumps still deposit heat at the mK stage.
- "One coax per qubit" is a simplification: real systems use 2–4 lines per qubit, and readout lines can be frequency-multiplexed.
The Feynman-style question
Is the bottleneck on the chip, or at the fridge door? It depends on where heat enters. This module fills the step *after* transduction—catching the light at room temperature. Reading "fills a link" as "connects the whole line" costs an extra measure of imagination. The post also notes the company background: founded in late 2021 by a USTC PhD team, modest headcount, recent tens-of-millions-RMB strategic funding, with a stated "million-qubit" vision that should be kept separate from the current product.
Sources cited in the post: Phoenix Tech coverage, the company's official WeChat announcement, and Qichacha corporate registry data.