Unit triggers ok but then seems to saturate and requires a ~60 second cool down time

[rugene76]

I've got the rcwl-0516 connected to an esp32/esphome and physically dangling in the air. I'm testing it to see how reliably it detects movement.
Initially it triggers well and within about a 10 ft radius. But after about 5 minutes of frequent triggers, it loses all sensitivity and seems to need about a 60 second cool down before it detects again. (times are approximate)

Feels like perhaps the 0516 difference detection gets saturated or reaches a new equilibrium that fails to trigger. I'm guessing that the ~60 second cool down allows for a decay of the difference signal - but who knows....

Anybody seeing this behavior? or can explain what's happening with a fix? seems like a deal breaker for me for controlling lights.
thanks!
rugene

[benjaminBoehme]

I starting to see a similar behaviour. It's hooked up to a rasberry and I wrote a simple python script which confirmed my observations. Imagine me standing 2m right in front of my sensor an continuously waving my arm. What I'm seeing on the output is:

┌──────────┬──────────────┬──────────────────────────────────────────────────────────────────────────────────────────────────┐
│          │              │                                                                                                  │
│          │ yes          │       ┌───────────────────────────────────────────────────────────────────────────────────────   │
│ Movement │              │       │                                                                                          │
│          │ no           │  ─────┘                                                                                          │
│          │              │                                                                                                  │
├──────────┼──────────────┼──────────────────────────────────────────────────────────────────────────────────────────────────┤
│          │              │                                                                                                  │
│          │ activity     │       ┌─────────────────┐      ┌────────┐     ┌───┐                           ┌───────────────   │
│   OUTPUT │              │       │                 │      │        │     │   │                           │                  │
│          │ no activity  │  ─────┘                 └──────┘        └─────┘   └─────────  ──  ──  ──  ────┘                  │
│          │              │       ▲                 ▲      ▲        ▲     ▲   |                           ▲                  │
├──────────┴──────────────┼────── │ ─────────────── │ ──── │ ────── │ ─── │  (15.2) ───────────────────── │ ─────────────────┤
│                         │       │                 │      │        │     │                               │                  │
│   Time (seconds)        │      (0)               (6)   (8.4)   (11.2) (13.5)                          (90)                 │
│                         │                                                                                                  │
└─────────────────────────┴──────────────────────────────────────────────────────────────────────────────────────────────────┘

In other words: although there's constant movement i can see a decline in the maximum pulsewidth on the output which finally falls slightly below the calculated value of Tx which is around 2.4 seconds (see this picture within this repository for calulation and definitions). After that the sensor becomes "blind" for a slightly variable amount of time (somewhere between 30s and one minute) and returns back with reduced or full sensitivity* (did not find a clear timing pattern here). At the same time the pulsewidths on the output are declining, I can observe a slight decline in Tiwhich should be around 2.4 s as well.

*I do not observe a reduced sensitivity in terms of range, but in therms of the maximum duration of the activitiy "pulse" on the output. The maximum "initial" pulsewidth I get is around 6-8 seconds with continuous movements.

To me this looks like there's some issue with at least one capacitance not (un-)charging sufficiently within the timeframe it is supposed to. Measurements of the SMD parts do not indicate a fault to me - at least their values align with the datasheet's values besides R1 which is 1kOhms on my board. So I guess it must be some issue within the IC.

[rugene76]

nice investigating. I guess the part is "too good to be true" for its price point.

I did find that conditioning the 5v supply can greatly reduce false triggers. I ended up putting a 10uf electrolytic followed by a ferrite bead and then a 100pf ceramic capacitor on the 5v feed to create a low pass filter over a wide bandwidth. Even though that helped a lot, I'm still getting weird behavior similar to what you describe. Unfortunately, I don't have an oscilloscope capable of delving into this more deeply.
So after much experimentation, I've given up on the part for use as a presence detector in an ESPHome environment.

[casesolved-co-uk]

I wonder if the chip has some kind of automatic gain control which isn't quite tuned correctly.

I bought 3 of these boards which were still panellised so must be identical: one worked with several meters of range straight away, one worked with about 10cm of range and the other seemed to have a pretty good range too the second time I tried it.

BUT if you leave it running the range seems to vary a lot. Sometimes it works great, other times you have to get within cms.

I also wonder if there's actually a fault with the design which is why they are so cheap? Shame.

[casesolved-co-uk]

Also see #57
Could be some kind of periodic interference with metallic objects close-by?