Watching Where LoRa Silicon Is Headed — SX126x and LR11xx

two radio chip icons comparing the SX126X and LR11XX chipsets

Radio silicon isn't something I think about every day — once you've picked a transceiver and built a product around it, the temptation is to stop paying attention. But every so often it's worth resurfacing and looking at where the underlying chips have moved, if only so I'm not caught flat-footed the next time we're evaluating parts. I spent some time recently going back through what's changed across Semtech's LoRa transceiver generations, from the older SX127x-era parts through SX126x and into the newer LR11xx family. This is me tracking the silicon landscape out of professional curiosity, not a statement about what's in any current Siliqs product.

The generational trend: better sensitivity, lower power

The consistent theme across successive generations of LoRa transceivers has been incremental but real improvement in two dimensions that matter enormously for a battery-powered, long-range sensor node: receive sensitivity and power consumption. Better receive sensitivity means you can close a link at a lower signal-to-noise ratio — which in practice means more range for the same transmit power, or the same range with more margin against interference and multipath. Lower power consumption, particularly in receive and low-power listening modes, matters just as much as transmit-side efficiency for a node that spends most of its life listening or sleeping rather than actively transmitting, since airtime is a small fraction of a LoRa node's duty cycle compared to time spent idle or in low-power receive windows.

I'll be honest that I don't have precise dBm sensitivity figures or exact current-consumption numbers I'd want to put in writing here — those depend heavily on specific configuration, spreading factor, bandwidth, and test conditions, and I'd rather not quote numbers I'm not fully confident are representative. But the qualitative direction is well-established and consistent with how RF silicon generally improves over successive generations: newer process nodes and refined RF front-end design tend to buy you meaningfully better sensitivity and lower current draw than the generation before, and that's held true moving from SX127x-class parts to SX126x and onward.

LR11xx's headline addition: geolocation without a full GPS

What actually got me spending an afternoon reading datasheets, though, was the LR11xx family's built-in geolocation capability — specifically, integrated GNSS scanning and passive Wi-Fi scanning, both aimed at low-power location without needing a dedicated GPS receiver chip. The idea is clever: instead of powering up a full GNSS receiver (which is a real power draw, especially for a battery-powered node), the transceiver itself can do a quick, low-power scan — capturing enough satellite or Wi-Fi access point signal information to either compute a rough position on-device or send the raw scan data off to a cloud service that resolves position from it. Passive Wi-Fi scanning for geolocation is a neat trick in particular — you're not associating with an access point or transmitting anything, just listening for beacon signals from known access points and using that for a location fix, similar in spirit to how some smartphone location services work indoors without GPS.

For a mobile or intermittently-relocated asset-tracking use case, that's a genuinely appealing feature — location awareness without the battery cost of a dedicated always-on GPS module, using silicon you might already have on the board for the radio link itself.

Where this leaves me

None of this changes what we ship today — our current nodes are fixed-position industrial sensors where geolocation isn't the primary problem we're solving. But it's the kind of feature that makes me want to keep an eye on where the chipset roadmap goes next, because "the radio chip you already need for the mesh link can also give you a low-power location fix" is a genuinely elegant piece of integration, and if we ever look at more mobile or asset-tracking-flavored use cases, it's exactly the kind of capability I'd want to already understand rather than discover cold. For now, it's just me satisfied to have spent an afternoon catching up on where the silicon's been going.

兩個無線電晶片圖示比較 SX126X 與 LR11XX

無線電晶片不是我每天都會想到的東西——一旦選定了收發器並圍繞它做出產品,就很容易停止關注後續發展。但偶爾還是值得重新浮上來看一下底層晶片走到哪了,至少下次我們要評估零件時不會措手不及。我最近花了些時間回頭梳理 Semtech LoRa 收發器歷代的變化,從較早的 SX127x 世代,經過 SX126x,再到較新的 LR11xx 家族。這是我出於專業好奇心在追蹤晶片產業的走向,不是在陳述任何現行 Siliqs 產品裡用了什麼。

世代趨勢:更好的靈敏度、更低的功耗

歷代 LoRa 收發器一直有個一致的主題:在兩個對電池供電、長距離感測器節點極為重要的維度上,持續但確實地進步——接收靈敏度與功耗。更好的接收靈敏度代表你可以在更低的訊噪比下把連結建立起來——實務上代表相同發射功率能拿到更遠的距離,或是相同距離能有更多對抗干擾與多路徑的餘裕。功耗降低,特別是在接收與低功耗監聽模式下,對一個大部分生命都在監聽或睡眠、而不是主動發射的節點來說,重要程度不亞於發射端的效率,因為對 LoRa 節點來說,空中傳輸時間相對於閒置或低功耗接收窗口所佔的時間比例其實很小。

老實說,我手上沒有精確的 dBm 靈敏度數字或確切的電流消耗數字想寫在這裡——這些數字高度取決於具體設定、spreading factor、頻寬,以及測試條件,我寧可不要引用一些我沒有十足把握具代表性的數字。但這個定性方向是相當確立的,也跟 RF 晶片歷代改進的一般模式一致:更新的製程節點與經過改良的 RF 前端設計,通常能換來比前一代明顯更好的靈敏度與更低的電流消耗,從 SX127x 級的晶片走到 SX126x 再往後,這個趨勢一直成立。

LR11xx 的招牌新功能:不用完整 GPS 就能定位

不過真正讓我花一個下午讀 datasheet 的,是 LR11xx 家族內建的定位能力——具體來說是整合的 GNSS 掃描以及被動 Wi-Fi 掃描,兩者都是為了在不需要專用 GPS 接收晶片的情況下做到低功耗定位。這個想法很巧妙:與其開啟一整顆完整的 GNSS 接收器(這對電池供電節點來說是實打實的耗電),收發器本身就可以做一次快速、低功耗的掃描——擷取足夠的衛星或 Wi-Fi 熱點訊號資訊,要嘛在裝置端直接算出大概位置,要嘛把原始掃描資料送到雲端服務去解算位置。被動 Wi-Fi 掃描用於定位特別是個聰明的技巧——你並沒有連上某個熱點或發送任何東西,只是在監聽已知熱點的 beacon 訊號,並用這個做定位,精神上有點類似某些智慧型手機在室內沒有 GPS 時做定位服務的方式。

對於行動或間歇性移動的資產追蹤使用情境來說,這是一個確實吸引人的功能——不用付出專用常開 GPS 模組的電池成本就能有位置感知能力,而且用的還是你可能本來就已經在板子上、為了無線電連結而存在的晶片。

這讓我得出什麼想法

這些都不會改變我們現在出的貨——我們目前的節點是固定位置的工業感測器,定位不是我們在解決的主要問題。但這種功能讓我想繼續關注晶片路線圖接下來會往哪走,因為「你本來就需要用來做 mesh 連結的無線電晶片,同時還能給你低功耗定位」是相當優雅的整合設計,如果哪天我們真的去看更偏行動、或資產追蹤性質的使用情境,這正是我會想事先搞懂、而不是臨時現學的能力。目前為止,單純就是我花了一個下午把晶片的發展方向補了課,心滿意足而已。