You are browsing as a guest. Sign up (or log in) to start making projects!

Gerafftes

@Gerafftes

Joined June 2nd, 2026

  • 55Devlogs
  • 3Projects
  • 1Ships
  • 28Votes
hi.
Ship Changes requested

So welcome to this ship.
In short: tracking living things via wifi (csi or channel state information). In the future I can ad vital tracking, pose estimates and person identification over bone mass.

I made a demo video explaing everything important: https://youtu.be/k27eQqz_vok
In short:
I had the idea because of a video from Bilawal Sidhu. I found it quite interesting and wanted to try it. I completely underestimated the whole project. I remember the moment where I was a the gym and thought that I was near the finish line (At this moment I had like 60h). The whole gym session I tried to think of something that would increase the time so I can get more stardust. When I went home an added the CAD editor (it makes the whole setup easier and visual. First screenshot (Its a fallback setup because the project is to big and I need to move it to an external SSD. Not everything has moved jet, because of that the database uses the fallback setup). It is about 405gb, mostly because of long measurements.
I created 3 PCB versions, with the breadboard 4. I hand some dumb mistakes in the first 2. They were my first ever PCBs and I created them in Ki Cad. Sadly the timelapse of it got missing. (I wrote several emails, texted with multiple people over slack and open a GitHub issue. Nothing helped). The PCBs are for the mmWave sensor. Its purpose is to help me calibrate the system and give reference data to see how accurate the positing is.

But to be honest you can also see all this in the 2 min video.

Even though I spend over 350 hours on this project I am sorry to announce that the project isn’t completely finished jet. I tried my best :(

  • 51 devlogs
  • 353h
Try project → See source code →
Open comments for this post

9h 11m 56s logged

DEVLOG #51

I now have a full five-minute recording with peer data in both directions. RX1 received 263 suitable frames from RX4, and RX4 received 361 from RX1. Both still fall short of the required 600, and the time-window coverage failed too.

I fixed the host handling of out-of-order packets, which had previously stopped a recording. The completed pair run handled seven reordered packets without that earlier failure. I also installed the recovery improvements on all four receivers, though the deliberate Wi-Fi outage test is still pending.

A later firmware filters unsuitable peer layouts earlier, but captures were still expiring during pair preparation. The interesting catch was the timing model: the real scheduler uses 10 ms ticks, while earlier models assumed 1 ms. A send opportunity could look possible but be missed before the worker woke up.

The newest fix adds processing reserve and stricter readiness checks. It passes the offline expiry model, but still needs hardware validation.

Next: test the new candidate on hardware, then run a short peer pilot before repeating the full pair recording.

0
0
36
Open comments for this post

5h 31m 53s logged

DEVLOG #50

The shared firmware is now running on all four receivers, and I finally have a passed baseline recording. Across three minutes, every TX1-to-RX link reached all 173 valid analysis windows, above the required 156.

Getting there took another fix: the first pilot was missing the WLAN data filter and traffic generator used by the earlier firmware, so it barely collected useful CSI.

The peer links are still giving me trouble. I reproduced and fixed a timing race that ended their sessions too early. After installing that fix, both sessions stayed active during the short check, but peer captures were still expiring before being sent. The next pair recording was blocked, and I returned all receivers to TX-only operation.

I’ve now prepared another fix that separates waiting for a send slot from running out of peer quota. It passes the offline regression, but hasn’t been validated on hardware yet.

Next: validate the transport fix on hardware, then repeat the full baseline and two-way peer measurements.

0
0
45
Open comments for this post

9h 36m 32s logged

DEVLOG #49

I now have one shared firmware image for all four receivers. Each RX gets its identity, LED settings and measurement profile through configuration, which should make future changes much easier to maintain.

I also reworked Wi-Fi reconnection, peer scheduling and updates. The server and recorder now understand the new packet format, keeping device identity, timing and configuration changes attached to the raw CSI data. The shared image builds reproducibly and has passed its offline checks, but the hardware rollout and real measurements are still ahead.

Separately, RX4 finally has a confirmed boot on the earlier Gen02 firmware, with Wi-Fi and CSI running. One confusing detail: its old address was actually answering as RX3. That explains the later address mismatch, though it doesn’t prove what caused the original timeout.

The inconsistent file reads are still unexplained, so I’m keeping that issue open too.

Next: qualify the shared firmware on the receivers, then repeat the baseline and two-way peer measurements.

0
0
47
Open comments for this post

11h 3m 36s logged

DEVLOG #48

The baseline finally passed its timing checks: about 5.47 ms uncertainty against the unchanged 50 ms limit. The next blocker is the peer link, where most captured peer frames couldn’t be used.

Wrong CSI layouts had been consuming the same budget as useful peer frames. I fixed the layout filtering and added fairer scheduling for peer traffic. The newer firmware reached RX1, but there’s still no complete two-way result.

RX4 restarted during its update and was still running V1 afterward. Later USB attempts never reached an app write, and its final boot mode wasn’t confirmed. The update’s reset cause remains unknown.

I also hardened the installer and file readers. The write path now uses the exact firmware bytes it verified, but the underlying inconsistent-read problem is still unexplained.

Next: confirm RX4’s state, complete the reviewed update, and repeat the peer pilot before the full two-way measurement.

0
0
59
Open comments for this post

11h 8m 52s logged

DEVLOG #47

RX2, RX3 and RX4 now have the new firmware installed, with full readback and same-boot qualification completed. RX1 stays on its original firmware.

The short capture passed. A later baseline recording contains over 32,000 raw frames, but failed the clock checks: its required end observation arrived about 114 ms late. Having the data isn’t enough if I can’t prove its timing.

I reworked that end-timing path to keep local file work out of the critical sampling window. It passes offline checks, but still needs a successful real run. I also added verified source loading after inconsistent file reads; their underlying cause remains unresolved.

The latest attempt then missed its dispatch window, so it never started a capture. The peer comparison is still untouched.

Next: qualify the current boots, launch within the valid window, and verify the new timing path before continuing to the peer comparison.

0
0
66
Open comments for this post

10h 37m 45s logged

DEVLOG #46

RX1’s reconnect firmware passed its 15-minute hardware qualification, and I collected around 21 minutes of UART logs. The original failure didn’t reappear, so I still can’t call the Wi-Fi issue fixed. I restored and verified the original firmware afterward.

I also finished the new firmware builds for RX2, RX3 and RX4. All three include the reconnect fix and diagnostics. RX2 and RX3 also include the packet-age fix, while RX2 gets proper UDP socket locking. These builds are tested offline, but aren’t installed yet.

On the host side, I fixed a race in the logger handoff and tightened the time checks so an installation can’t start without enough time for qualification and a possible safe rollback. The complete installation-to-measurement workflow passes offline, but the real baseline and two-way peer measurements are still missing.

Next: install and qualify RX2–RX4 one at a time, then run the short capture and full baseline/peer comparison.

0
0
37
Open comments for this post

9h 56m 46s logged

DEVLOG #45

The setup recorded 5,165 raw frames in a short test, but the run still failed because its rejection log overflowed. I expanded that bounded log and moved slow startup hashing out of the packet-processing locks.

I also installed RX1’s new memory-diagnostic firmware and checked it for about 15 minutes. The old heap dip didn’t reappear, although that doesn’t prove its cause is fixed.

Wi-Fi is now the main headache. TX1’s access point restarted internally, and the receivers didn’t reliably reconnect. Restarting them helped temporarily, but the problem returned. I found a firmware path where an immediately failed reconnect schedules no further attempt. The corrected candidate is built and tested offline, with extra UART diagnostics to capture what happens during a real failure. The actual outage cause is still unproven.

Next: check the reconnect candidate on hardware, capture the failure, then retry the short recording and full baseline/peer comparison.

1
0
76
Open comments for this post

10h 19m 40s logged

DEVLOG #44

The offline evaluation finally fits within its three-minute limit. About 155 seconds on the maximum-rate test data, with the same results as before.

A real recording had stopped when RX1 packets arrived out of order. I added a small 10 ms reordering buffer, while stale or invalid data is still rejected. That makes the host more tolerant, but doesn’t explain what caused the original reordering.

I also fixed a recording-start issue where an old packet received before the start could abort the new capture. Rejection details and file closing are more reliable now, too, so failed runs leave better evidence.

Next: verify a clean 30-second recording start, then run the full baseline and two-way peer comparison.

0
0
56
Open comments for this post

10h 29m 2s logged

DEVLOG #43

I built a first offline position pipeline that predicts x/z coordinates instead of just fixed zones. It can save and reload a model, but I’ve only tested it with synthetic data so far. Real tracking accuracy is still unproven.

I also traced more of the CSI data path. A diagnostic capture contained 5,005 valid CSI frames from all four receivers, and a fresh offline replay accepted all of them. The running host still wasn’t accepting several receivers correctly, though, so the live issue isn’t fixed yet.

On the measurement side, I tightened capture timing and replaced the oversized blanket storage reserve with a budget based on the actual recording limits. The real storage medium still needs checking.

The main blocker now is processing speed: the maximum-rate synthetic run still exceeds the three-minute evaluation deadline. Recording and the return to the default state work, but that doesn’t make the complete run a pass.

Next: speed up evaluation using the saved test data, then recheck the complete measurement workflow before another live run.

0
0
45
Open comments for this post

10h 6m 15s logged

DEVLOG #42

RX3 and RX4 have passed their individual checks on the latest firmware. A complete OTA update, the LED pattern test and ten minutes of stability monitoring each, with no errors recorded.

The false SOS warning on RX1 is also fixed and verified on the board now. I corrected concurrent access in the shared UDP sender and transport statistics, too. Those are real software fixes, but I still can’t claim they explain the original RX3 heap corruption or watchdog resets.

I also improved CAD profile handling. Profiles can use a separate store, the UI shows which revision is loaded, and refreshing no longer overwrites unsaved edits or brings back an older saved version.

The rollout isn’t finished, though. The next RX1 update stopped because its device identity couldn’t be confirmed, so RX2’s update to this latest version hasn’t started either.

Next: resolve that binding, finish updating RX1 and RX2, then move on to the actual two-way peer comparison. Tracking acceptance is still open.

0
0
39
Open comments for this post

9h 58m 57s logged

DEVLOG #41

RX1 finally passed a full three-minute TX1→RX1 test. 6,407 qualified CSI frames, around 35.6 per second, and all 173 measurement windows passed. The firmware had been rejecting our setup because TX1 is also the Wi-Fi access point. That’s corrected and installed now.

SOS is the warning for missing or stale usable CSI from TX1. It doesn’t automatically mean the board is broken. Ironically, the warning itself had a bug, RX1 reported stale data even though a matching frame had arrived just 8.7 milliseconds earlier. A timing race between the LED task and CSI reception caused the false alarm. The fix is built and tested offline, but isn’t installed yet.

Host startup and recording also work now. RX2 is still a problem, though: its upload timed out while CSI kept arriving, and the cause isn’t confirmed.

Next: install and verify the LED fix on RX1 before continuing with the other receivers. The actual two-way peer comparison and tracking acceptance are still open.

0
0
23
Open comments for this post

10h 0m 36s logged

DEVLOG #40

I cut the recording footprint from 4.54 to 2.67 GB in the full synthetic test, about 41% less. The saving comes mainly from removing duplicate exports, not throwing away raw data or evidence.

I also corrected timing and CSI rate-limit bugs offline, plus a cleanup issue that could prevent the recorder from stopping its own processes when disk space got tight.

The frustrating part is that the peer comparison still hasn’t recorded anything. One attempt stopped on the timing bug, after that was fixed, the next hit a missing signing key during host startup. No peer firmware was uploaded and no capture started in either attempt.

Next: fix host startup and prepare a separate 35-minute pilot. The actual peer comparison and tracking acceptance are still open.

0
0
41
Open comments for this post

10h 15m 21s logged

DEVLOG #39

All four receivers finally made it through a full 30-minute CSI run together. RX1, RX3 and RX4 are now on C1, while RX2 stays on C. Each delivered around 17–20 matching frames per second, and 99.61% of the shared measurement windows passed the CSI checks.

The RX1↔RX4 peer firmware and recording workflow are also implemented and tested offline now. The actual radio comparison hasn’t happened yet, so I still don’t know whether those extra links will improve positioning.

One annoying bug kept rejecting good data after a 128→64→128 format change. That’s fixed in the offline analysis, along with separate format rules for TX1 and peer links and a recorder completion check.

I also added RX and radar exports that keep raw data, timing and recording boundaries together. That path works in offline tests; it isn’t a new live hardware result.

Next: check both mmWave sensors individually and then together at known points. The shared CSI run is a milestone, but tracking and full-system acceptance are still open.

0
0
23
Open comments for this post

9h 24m 53s logged

DEVLOG #38

RX2 is finally delivering a lot more matching CSI data. After installing the new firmware, the rate went from about 1.9 to 31.9 frames per second. A longer, half-hour capture averaged around 30. The firmware now fixes Wi-Fi bandwidth at 20 MHz and guards against invalid CSI callbacks.

One annoying catch: the long capture only recorded RX2 because some device endpoints were outdated or mixed up. I’ve prepared a stricter preflight so an incomplete group can’t silently become a four-receiver test again.

I also corrected the offline window analysis. A valid frame with a different format no longer ruins an entire otherwise usable window. The UI code now separates format exclusions from unexplained sequence gaps instead of calling everything “loss”. The host fixes are tested locally, but aren’t active on the running server yet.

Next: complete a proper simultaneous four-RX capture, then implement the RX1 <-> RX4 peer profile for testing. Tracking and full-system acceptance are still open.

0
0
51
Open comments for this post

9h 48m 15s logged

DEVLOG #37

I narrowed down the RX2 and mmWave2 issues from the failed run. Still no hardware fix, but the picture is getting clearer.

The surprising part is that RX2 was still sending valid CSI packets during its apparent dropout. They just weren’t the 128-subcarrier frames the current evaluation needs. mmWave2, on the other hand, really did have gaps in packets reaching the host. I’m looking into its send-and-acknowledgement path, but haven’t proven the cause yet.

I also found a parser mistake. A supposedly damaged RX packet was actually a valid warning message. That’s corrected in the isolated diagnostic tools, which now also distinguish device timeouts from requests the collector skipped or cancelled.

New diagnostic firmware candidates for TX1, RX2 and mmWave2 are built locally to help trace where things stop moving. They aren’t installed yet, and the TX send errors are still unresolved.

The UI is less misleading now, too: an ESP being reachable doesn’t mean radar data is arriving, and old readings no longer keep disconnected devices looking active.

Next: review the diagnostic candidates, then use a short, controlled hardware run to investigate the remaining errors. The last full acceptance is still a FAIL.

0
0
24
Open comments for this post

9h 31m 34s logged

DEVLOG #36

The first full TX1 acceptance run is done now and it failed, but at least I finally have a much clearer picture of what is still going wrong…

The common setup passed: all seven expected participants were visible, and 22 complete snapshots kept the shared basis valid for more than 20 seconds. Then Scope A ran its three 120-second windows. All three failed. There were 14 new TX send errors, and the diagnostic counters placed them in the wifi_handoff ENOMEM stage. That tells me where the error was recorded, but not yet why it happens or whether it is fixed. MMWAVE2 also became stale in parts of the run, and several RX requests were missing, so none of the historical comparisons could count as a valid PASS.

Scope B stopped itself at the safety gate when the source binding became inconsistent and RX2 was already 5.044 seconds old. This is frustrating, but the guard did exactly what it was supposed to do, it refused to turn an incomplete 120-second window into a result.

Next: decide how to continue the ENOMEM and source-binding issues before starting another acceptance run.

0
0
55
Open comments for this post

10h 9m 20s logged

DEVLOG #35

I fixed two concrete TX1 firmware bugs. An authentication check that ran too late and an HTTP task with too little stack space.

The status endpoint could lose the request headers before checking authorization. I moved that check ahead of the response. I also increased the HTTP task’s stack from 4 KB to 8 KB after confirming that some call paths could exceed the old limit.

Two independent builds matched byte for byte. After taking fresh backups, I installed the candidate and verified the complete 16 MiB flash against the expected image. That part passed.

The UDP/ENOMEM errors are not yet proven fixed.

Next: investigate the reset path, complete a clean controlled boot, then check the matching server setup, authentication, all participants, and the measurement windows.

0
0
79
Open comments for this post

10h 7m 39s logged

DEVLOG #34

The TX1 migration is now 100% verified, but the project is still only around ≈ 70% complete. The third and final boot attempt did not give me the clean result I was hoping for.

Recovery happened within the allowed time, but UART0 showed two app starts. After the first start, an RTC_SW_CPU_RST appeared and the app started again. It eventually reached BOOT_OK, but because of the duplicate start I still cannot accept the boot trace or move on to server activation and measurements. I am currently documenting the exact restart sequence and trying to figure out what caused it.

I also put together a source collection about Wi-Fi sensing and RF-based person detection. GitHub

The separate mmWave freshness test did not reproduce the old MMWAVE2 delay problem, but that does not prove it is fixed.

Next: get one clean TX1 boot, then activate the server and start collecting real measurements.

0
0
92
Open comments for this post

10h 4m 28s logged

DEVLOG #33

The TX1 finally made it all the way to BOOT_OK. After the CHECK_COUNTRY problem from the last update, the new candidate passed the Country20/Power44 and Wi-Fi checks, booted into test mode, and passed the local TX/RX4 checks. I also confirmed the amber LED and both mmWave transport paths.

The previous candidate had stopped because one sampling gap grew to 10.2 seconds. I traced that back to the timing bookkeeping and changed the sampler to schedule from the last real sample completion. The corrected run stayed within the sampling limits. That is a real improvement, but it is not the final acceptance yet. The full coverage result is still blocked by the clock-binding requirement, and the radar window only showed no_target, not a position result.

I also switched to the new diagnostic server in a controlled run. It stayed ready and a separate 61-second capture passed the coverage gate, but the active CSI position setup is still missing, so this does not count as a complete TX-7B or system pass.

The CAD UI got a pretty big upgrade too. MMWAVE1 and MMWAVE2 can now be edited as separate sensors with their own colors, 6 m fields of view, direction arrows, and yaw handles. I added a button to aim them at each other, plus a wall-angle tool that draws the real beam/wall intersection and lets me set the angle directly. Marker labels also move out of the way automatically, and the snap setting survives a refresh.

Next: fix the remaining coverage and clock evidence, activate the sealed CAD/CSI setup, and complete the full TX-7B validation.

0
0
93
Loading more…

Followers

Loading…