
Today I actually had a link to debug on the bench, not just reading — a couple of nodes were dropping packets, and my first instinct was to check RSSI. Wei pointed out that RSSI on its own doesn't tell the whole story, which sent me down a rabbit hole of finally understanding what RSSI and SNR actually measure and why you need both.
RSSI (received signal strength indicator) is exactly what it sounds like: how strong the received signal is, in absolute terms, usually reported in dBm. It tells you the raw power level arriving at the receiver — but it says nothing about what else is going on at that frequency. A receiver sitting right next to a noisy switching power supply or in a room full of other RF traffic could show a perfectly strong RSSI while still having a garbage link, because the noise floor around it is also elevated. Strong signal, strong noise — the receiver still can't reliably tell signal from noise.
That's where SNR (signal-to-noise ratio) comes in. SNR compares the signal power to the noise floor and expresses it in dB — how far above the noise the signal actually stands out. A high SNR means the signal is clearly distinguishable from background noise even if the absolute RSSI is on the weaker side; a low or negative SNR means the signal is close to (or, thanks to LoRa's processing gain, potentially even below) the noise floor, which is exactly the situation where packets start getting dropped or corrupted even though "the RSSI looked fine."
The two numbers together are what actually gave me useful information today. My weak link had a not-terrible RSSI but a pretty poor SNR, which told me the problem probably wasn't distance/path loss (that would tend to tank RSSI first) — it pointed more toward a noisy RF environment near one of the nodes. That's a genuinely different debugging path: instead of "move the nodes closer together or add a bigger antenna," I should be looking at what's near that node generating interference — I'm thinking through candidates like switching supplies, other 2.4GHz gear if it's anywhere close in band, or just a bad ground/shielding situation, though I haven't confirmed which yet.
The mental model I'm walking away with: RSSI tells you "how loud is the signal," SNR tells you "how much louder is the signal than everything else in the way." A link can fail with strong RSSI and weak SNR (noisy environment) just as easily as it can fail with weak RSSI and okay SNR (just too far / too much path loss, but a clean environment). Looking at either number alone would've sent me down the wrong troubleshooting path today — I probably would've started by trying to shorten the distance between nodes, which wouldn't have fixed anything.
Still want to build more intuition for what SNR numbers actually mean in practice for different spreading factors — I get that higher spreading factors tolerate lower (more negative) SNR before the link breaks down, but I don't yet have a good feel for where the real-world thresholds sit. That, plus figuring out exactly what's generating the noise near that node, is on my list for tomorrow and beyond.

今天在 bench 上真的遇到一條要排查的連結,不只是讀文件——有幾個節點在掉封包,我第一直覺是去看 RSSI。Wei 提醒我光看 RSSI 沒辦法看到全貌,這讓我一路查下去,終於搞懂 RSSI 跟 SNR 到底各自量的是什麼,還有為什麼兩個都要看。
RSSI(received signal strength indicator,接收訊號強度指示)就跟字面意思一樣:收到的訊號有多強,用絕對值來看,通常以 dBm 表示。它告訴你抵達接收端的原始功率大小——但完全沒說那個頻率上還有什麼別的東西在發生。一個接收端如果剛好在很吵的 switching power supply 旁邊、或是在一堆其他 RF 流量的房間裡,RSSI 可能顯示得很漂亮,連結品質卻可能一團糟,因為它周圍的 noise floor 也同樣被墊高了。訊號強、雜訊也強——接收端還是沒辦法可靠地把訊號跟雜訊分開來。
這時候就輪到 SNR(signal-to-noise ratio,訊噪比)上場。SNR 拿訊號功率跟 noise floor 比較,用 dB 表示——訊號實際高出雜訊多少。SNR 高,代表就算絕對 RSSI 偏弱,訊號還是清楚地能跟背景雜訊區分開來;SNR 低甚至是負的,代表訊號已經逼近(甚至因為 LoRa 的 processing gain,理論上可以低於)noise floor,這正是封包開始掉包或損壞的情況——即使「RSSI 看起來還好」。
今天真正給我有用資訊的是把兩個數字放在一起看。我那條弱連結的 RSSI 沒有很差,但 SNR 相當差,這告訴我問題大概不是距離/路徑損耗(那種情況通常會先讓 RSSI 掉下去)——比較像是其中一個節點附近的 RF 環境很吵。這其實指向完全不同的排查方向:不是「把節點靠近一點」或「換更大的天線」,而是該去看那個節點附近有什麼東西在產生干擾——我在想的可能對象有 switching power supply、如果頻段夠近的話也可能是其他 2.4GHz 設備,或者單純是接地/屏蔽沒做好,不過目前還沒確認是哪一個。
今天帶走的心智模型是:RSSI 告訴你「訊號有多大聲」,SNR 告訴你「訊號比周圍其他東西大聲多少」。RSSI 強、SNR 差(吵雜環境)可能讓連結掛掉,跟 RSSI 弱、SNR 還可以(單純太遠/路徑損耗太多,但環境乾淨)一樣容易讓連結掛掉。今天如果只看其中一個數字,大概會被帶到錯誤的排查方向——我可能會先想辦法縮短節點間距,結果根本沒解決問題。
我還想對不同 spreading factor 下 SNR 數字實際代表什麼建立更多直覺——我知道 spreading factor 越高,連結能容忍的 SNR 越低(越負)才會斷線,但實際上這個門檻大概落在哪裡,我還沒有很好的手感。這個,加上搞清楚那個節點附近到底是什麼東西在產生雜訊,是我接下來要繼續查的。