Originally published on Substack: https://charslab.substack.com/p/i-turned-a-dead-sony-nasne-into-a

Sony's nasne (CECH-ZNR2J) is a small black box that was never quite like anything sold outside Japan. It is a network TV recorder: plug in an antenna, insert a B-CAS card, connect it to your LAN, and it records and streams Japanese television — terrestrial, BS and CS satellite — to PlayStations, phones and PCs around the house.

It is also a product with an expiry date. Sony's official services for nasne are scheduled to end, and with them the official apps that make the box convenient. A perfectly good tuner then becomes much harder to use — not because the hardware failed, but because the software path to it disappears.

There is another path. A community project called SonyNasne-debian replaces nasne's internal software with Debian Linux, and serves live broadcasts through a small receiving server you control yourself. No official app, no cloud account between you and your own tuner — just a URL on your own network that you open in VLC.

I wanted to know whether that actually works, starting from the worst realistic position: a nasne whose hard drive had just died.

It does work. This article is the story, the verified results, and the complete first stage of the procedure. The full step-by-step guide — every command for the rebuild, the NV reset, the Debian build and the streaming setup, with every failure fix and a final checklist — will be published as a PDF on Gumroad in Week 2.

It started with a hard drive that stopped spinning

I had done the sensible part first. I set up the work environment, prepared the tools, and confirmed I could boot the official firmware after the preliminary flash work. An SPI backup — a copy of the unit's flash memory, the thing that lets you go back if everything later goes wrong — was part of that plan.

Then, immediately after that confirmation, the internal hard drive stopped spinning.

Not a software error. Not a corrupted file system. The disk simply no longer rotated. The backup I had just made, and every recording on the drive, were trapped inside a dead piece of hardware. The nasne itself was unharmed. Its tuner worked. Its network port worked. It just had nowhere to run from.

That is the situation a lot of owners will eventually face, official service or not. These units are old enough that the hard drive is the most likely thing to fail first. So the useful question is not "how do I modify a perfect nasne?" It is "can a nasne be rebuilt from a blank replacement disk, and then converted?"

The answer, verified on my own unit, is yes — with conditions, and with two traps that cost me real time.

My replacement parts were not expensive. A refurbished 2.5-inch 500 GB SATA hard drive cost ¥1,500 on Amazon Japan — and, in a very second-hand-market touch, the drive that arrived was a different model number from the one I ordered. It turned out to be a newer model, and it passed a SMART health check, so I used it. A USB-SATA adapter cable cost about ¥900. Plus a precision screwdriver, and my existing M5 MacBook Air. That was the whole shopping list.

What the finished system does

Before the how, here is the what — because this is the part you cannot get from the project documentation alone. The docs tell you what each command is supposed to do. They do not tell you what the finished, working system feels like, or what its limits are in daily use.

Live terrestrial TV, on your own terms. After the conversion, a receiving server called nasne-recpt1 runs automatically on the Debian system, listening on port 8301. A channel scan returns every receivable terrestrial channel in my area of Osaka, with station names. Open a channel URL in VLC on a PC or a phone, and you are watching live TV. No account, no official app, no pairing.

It is measurable, and I measured it. In an 8-second capture test on NHK General (physical channel 24 in my region), the stream delivered about 7.2 MB in 8 seconds — roughly 7.2 Mbps, which is a healthy, stable terrestrial broadcast rate, not a struggling trickle. The file played cleanly. This was not "the tuner responded, so it probably works." Data actually flowed, at broadcast speed.

BS and CS satellite work too. A satellite scan found a large number of BS and CS services through my building's shared receiving equipment. NHK BS streamed stably at about 13 Mbps in testing. One honest limit: the scan listing a pay channel and you being able to watch it are two different things. WOWOW or a sports channel will only produce video if your own B-CAS card actually has a contract for it. The system respects exactly the same entitlement rules as the original — it changes who controls the delivery, not what you are entitled to receive.

Away-from-home viewing, without exposing the server. The receiving server has no authentication. Putting its port directly on the internet would be irresponsible, so I did not. Instead I used Tailscale: my Mac at home advertises my home LAN as a subnet route, I approved that route in the Tailscale admin console, and my phone joins the same encrypted tailnet. I tested it the strict way — iPhone Wi-Fi switched off, on mobile data only, Tailscale on — and opened the same VLC URL I use at home. The picture played. The constraint is real and worth stating: the relay computer at home has to be there, and awake. If it sleeps, remote viewing stops.

Channel selection is usable. Typing a raw tuner URL every time you want to change stations is not a system, it is a party trick. The server can generate a playlist with station names, which VLC presents as a selectable list. One catch I found: the generated playlist reflects the last type of scan you ran, so if you want terrestrial and satellite in one list, you need a combined scan. I will cover the exact commands in the full guide.

And the honest limits. The server handles one channel, one service, at a time. Ask for a second stream while one is playing and, after about ten seconds, you get a 503 — tuner busy. The same can happen for a few seconds right after switching channels. For one viewer, or one household taking turns, it is fine. It is not a whole-house multi-room system.

The proof that the conversion really took

The most important verification in the whole project is also the least visible.

nasne's hardware has a watchdog — a supervisory timer that, under the original design, expects the original software to keep it happy. If a converted system leaves that watchdog active, the unit force-reboots on a cycle of about 333 seconds, forever. It would look like it boots. It would even stream — for five minutes at a time.

So after installation — which itself is about 5 minutes of fully automatic processing once you power the unit on with the prepared disk: root image creation, a fresh SPI backup, writing and verifying the new kernel, and a reboot, all hands-off — I checked three things over SSH. The stage-1 log showed the kernel hash verification OK and the kernel write as "flash + verify OK." The watchdog log contained the line that matters: "MCU watchdog disabled." And uptime kept counting past 333 seconds, past ten minutes, past an hour.

That is the difference between "I ran the installer" and "the machine is now a Debian machine." The hardware is unchanged. What runs on it, and who it answers to, is completely different.

The two traps that cost me the most time

A full trap-by-trap account is in the paid guide, but two deserve a public warning, because both produce symptoms that point at the wrong suspect.

Trap 1: the USB adapter that freezes mid-write. My USB-SATA adapter uses a JMicron chip (USB ID 152d:0583). Inside my virtual machine, in its default UAS transfer mode, it would repeatedly reset during long writes and the rebuild would hang. The disk's SMART data was clean, which sent me looking in the wrong place. The kernel log told the truth: repeating timeouts. The fix is a one-line setting that pins the adapter to the older, slower, stable usb-storage driver — after which writes completed reliably. The catch: the setting is runtime-only, so you must re-apply it after every VM reboot. If your rebuild hangs during writes, suspect the bridge before you suspect the disk.

Trap 2: the unit that lights up but never appears. After rebuilding on the new disk, my nasne would power on — lamp lit — but never become reachable on the network. The cause was not the disk and not the rebuild. The unit's non-volatile memory (NV) still held information tied to the old disk, and the mismatch with the new one stopped a clean official boot. The project documents an NV reset procedure for exactly this case: save the current NV to disk first, then erase, verify, and let the unit reboot. After the reset, the official firmware booted normally. One side effect to know in advance: the reset also erases the official app's pairing, so if live TV does not appear in the phone app afterwards, re-registering the pairing — not another reset — is the fix.

There were two more traps in the Debian build itself — one involving DNS inside the build container, and one involving a CPU-detection script from 2008 meeting a modern architecture it had never heard of. Both are documented with their exact fixes in the full guide.

Stage 1, in full: preparing the work environment

So you can judge the level of the full guide — and start correctly if you decide to go ahead — here is the first stage exactly as it should be done.

My work environment was a Mac running a Debian virtual machine under VMware Fusion. Whatever your host, you need a Debian environment and a terminal. Stage 1 itself is safe: nothing here writes to the nasne.

Step 1 — Clone the project and run its environment check.

git clone https://github.com/mar1mo-41414/SonyNasne-debian.git cd SonyNasne-debian sudo python3 scripts/nasne_setup.py check 

The sudo is not optional decoration. Run the check as a normal user and it can falsely report that debootstrap is not installed — a PATH artefact, not a broken environment. I nearly went debugging a healthy setup because of this. If the check says NG, re-run it with sudo before you change anything.

Step 2 — Generate an RSA SSH key.

ssh-keygen -t rsa -b 4096 

Do this now, in RSA format specifically, even if you already have a modern Ed25519 key you use everywhere else. SSH on the nasne side is built around older algorithms, and the public key you generate here (~/.ssh/id_rsa.pub) is the one you will embed in the Debian build later — it becomes your root login key on the converted unit. Generating the wrong type now means rebuilding later.

That is the whole of Stage 1. Everything after it follows the same pattern: one stage, one verification, and never proceeding until you have confirmed you can still go back.

The one rule that governs the whole project: do not touch the firmware until you have taken an SPI backup and verified it by hash. In my run, spi_mtd0.bin should be 13,631,488 bytes — about 13 MB — and its SHA-256 must match between the unit and your work machine before you go any further. My dead drive taught me why this rule exists. The backup is not bureaucracy. It is the entire difference between an experiment and a brick.

What the remaining stages cover

Without giving away the command sequences that are the guide itself, here is the shape of the rest, so you know exactly what you would be buying:

  • Stage 2 — Rebuild on a replacement HDD: identifying the correct disk safely by model, capacity and serial; building the disk foundation; letting the unit write its identity; verifying the result; and recreating the recording area when the unit's own auto-generated folders get in the way.
  • Stage 3 — NV reset (only if the official firmware will not boot): building the reset tool, preparing the reset image, preserving the original files, and confirming the saved NV backup exists before trusting the reboot.
  • Stage 4 — SPI backup and verification: what to capture, the expected size, and the hash check that gates everything after it.
  • Stage 5 — Building Debian: the build command and what every flag means, plus the two build traps — container DNS, and the 2008 config.guess that cannot recognise a modern CPU — with their exact fixes.
  • Stage 6 — Install, first boot, and streaming: the automatic 5-minute install, the three SSH log checks that prove it worked, terrestrial scanning and capture testing, the BS/CS scan and URL format, the named playlist, and the Tailscale subnet-router setup for viewing away from home — plus how to return the unit to its official state.

The full guide is coming in Week 2

This free article gives you the story, the measured results, the limits, the two worst traps, and Stage 1 in full. What it deliberately does not give you is the complete, ordered command sequence for the rebuild, the NV reset, and the Debian build — because that sequence, checked against a real dead-drive recovery, with a fix attached to every place I actually got stuck, is the product.

Coming in Week 2 on Gumroad: My Sony nasne Died. So I Rebuilt It as a Debian TV Server — Full Guide (PDF, $24) at gumroad.com/charslab — every command as typed, every trap with its cause and fix, the verification checkpoints at each stage, and a final checklist. Follow Char's Lab here so the release post reaches you; the Week 2 article on Substack, Medium and DEV.to will link straight to it.

If your nasne is sitting in a closet because its drive died: it is not e-waste yet. Mine is on my network now, running Debian, streaming the evening news to my phone — on a ¥1,500 disk.

Disclaimer: This conversion is not supported by Sony, it voids the manufacturer's warranty, and you do it entirely at your own risk. Use your own B-CAS card, and use the result for private viewing only. Do not redistribute broadcast content, and do not expose the streaming port directly to the internet — it has no authentication.

Project credit: SonyNasne-debian — https://github.com/mar1mo-41414/SonyNasne-debian. All results above were measured on my own CECH-ZNR2J in Osaka, Japan.