Let me start with a confession. In my first year as a hardware procurement lead (2017), I assumed all Semtech wireless solutions were basically the same family of parts. You know, pick a chip, hook it up, and you've got connectivity. That assumption cost us about $3,200 in rework on a single order—and a 1-week delay that made my boss question my judgement.
Here's the thing: Semtech isn't just "the LoRa company" anymore, and even within their LoRa portfolio, the right choice depends heavily on what you're building and how many you're building. There's no universal "best" Semtech wireless solution.
The Key Question: What Kind of IoT Device Are You Building?
The way I see it, there are three main scenarios I've dealt with, and each points to a different Semtech product family. Getting this wrong is the classic mistake I see engineers make—they fixate on the radio chip and ignore the network infrastructure, the power budget, and the final certification costs.
Scenario A: The Low-Power, Long-Range Sensor (Battery-Powered)
This is the classic LoRa use case. You're building a sensor—maybe for agricultural soil moisture, or a smart parking lot sensor, or a blood pressure monitor for telemedicine. The device needs to run for years on a coin cell battery and report small amounts of data across a kilometer or more.
For this, you're looking at Semtech's dedicated transceiver chips: the SX1276, SX1272, or the newer SX1262. I've used all three. The SX1262 is my go-to now for new designs, because it's slightly more efficient and supports TCXO options that make it easier to hit regulatory limits. But here's the catch: an SX1276-based sensor costs about $1.50 to $3.00 more per unit in BOM than a comparable wireless MCU. But you're paying for the specificity. The radio performance is frankly better for long-range.
From my perspective, the mistake people make is using a full LoRaWAN stack when a simple point-to-point protocol would do. If you're controlling a single valve on a farm, you don't need a gateway. You just need the transceiver and a simple microcontroller.
A cautionary tale: The question everyone asks is "what's the range?" The better question is "what's the range in a real building with concrete walls?" In September 2022, I approved a design using an SX1262 in a metal enclosure for an industrial pressure sensor. That enclosure acted like a Faraday cage. We had to add an external antenna—raised BOM by $0.40 per unit and hurt the aesthetics. The lesson? Test the antenna in the final housing, not on a dev board.
Scenario B: The High-Density Industrial Gateway (Power-Present)
If you're building a gateway or a concentrator that needs to listen to hundreds of end-nodes simultaneously, you're in the realm of Semtech's SX1301 baseband processor. This chip is basically a mini-base station. It can demodulate up to 8 channels (64 SF5+8 SF7-12) simultaneously. It's complex.
I recall a project in Q1 2024 where we tried to design a gateway for a smart building using a standard LoRa module. The unit couldn't keep up with the SF7 transmissions from 40 door sensors in a high-latency environment. We had to redesign around the SX1301, which added about $15 to the BOM and required us to redo the RF layout entirely. That was a costly mistake born from assuming "all LoRa chips are the same."
The reality is that if you have more than 50 end-nodes within range, or if your network requires higher throughput (like air quality sensors that report every 5 minutes), you need the SX1301. The SX1276 simply isn't a base station. It's a transceiver. To put it another way: you can't use a fishing rod to drag a cruise ship.
What's often missed: The SX1301 itself is just the baseband processor. You still need a separate radio front end (like the SX1255/57). And you need to license the LoRaWAN standard if you plan to sell it as a compliant gateway. That adds engineering cost and legal fees. I've seen startups blow their budget because they thought "buying the SX1301" was the final step. It's the first step in a much longer road.
Scenario C: The Cellular-Connected Asset Tracker (Cellular IoT)
This scenario is where Semtech's acquisition of Sierra Wireless becomes relevant. If your device needs to be mobile—a fleet tracker, a shipping container monitor, a rental equipment tag—and it crosses areas without a local LoRaWAN network, you need cellular.
Semtech now offers cellular IoT modules based on Sierra Wireless technology (like the HL78xx series for LTE-M/NB-IoT). This is a different ballgame. You're trading the simplicity of LoRa for the ubiquity of cellular.
In my experience, the number one underestimation is the data plan cost. I've had clients say, "the module is only $25!" Yes, but the monthly cellular IoT data plan (even for a few megabytes) can add $5-$15 per device per year, depending on the provider and roaming agreement. Over a 5-year lifespan, that connection cost exceeds the module price. For a fleet of 200 trailers, that's a $15,000 + operational cost over 5 years.
Here's where the transparency principle applies: the vendor who lists the module price and the estimated connection cost upfront—even if the total looks higher—is more trustworthy than the one who just quotes a low module price and lets you discover $5/month in hidden roaming fees later.
To be clear: Cellular IoT is fantastic for mobile assets. It is overkill for a stationary soil moisture sensor in a field where you have a LoRaWAN gateway within 5 km.
How to Decide Which Scenario You're In
Here's the decision framework I've used to avoid the $3,200 mistake I mentioned earlier. Ask yourself these questions:
- Is the device mobile or stationary? Mobile → Cellular IoT (Scenario C). Stationary → Proceed to Q2.
- How many devices will be within range of a single gateway? Fewer than 50 → Scenario A (single transceiver). More than 50 → Scenario B (multiple units or a gateway).
- What's the power source? Battery for 5+ years? → Scenario A (LoRa transceiver). Mains power or high-capacity battery? → Scenario B (gateway) might be feasible.
- Do you control the network infrastructure? Yes? You can choose between A and B. No? You need to ensure there's a LoRaWAN network available (like The Things Network) or fall back to Cellular IoT.
I'd argue the most common trap for IoT hardware manufacturers (like us) is jumping straight to Scenario B because they think they need a "network" of LoRa, when really Scenario A with a simple SX1262 and a repeater is actually cheaper and simpler for a small cluster of devices.
One more thing I should mention: If you're dealing with a device like a cordless phone or a simple remote control, you're actually overcomplicating things. Semtech's wireless portfolio includes simpler sub-GHz modules that are not LoRa—they're just basic RF transceivers. Don't buy a SX1276 for a garage door opener. It's like using a sledgehammer to crack a nut. Use a cheaper, simpler module that just does on-off keying.
The cost of a SX1276 ($3-4 in volume) versus a basic FSK transceiver ($0.80-1.20) adds up quickly on a 10,000-piece order. That's $20,000+ in unnecessary BOM cost. The lesson? Know exactly what protocol and data rate you need before you pick the chip.
Honestly, if you take away one thing from this: don't be like me in 2017. Don't assume the product family is a single entity. Semtech has a broad portfolio for a reason. The SX1301 is not a faster SX1276. The cellular module is not a replacement for LoRa. They are tools for different jobs.
(Note on pricing: The BOM figures I quote are approximate and based on orders we placed in Q3 2024. Since semiconductor pricing fluctuates, I recommend checking current distributor quotes. The structure of costs—module price vs. integration cost vs. connectivity cost—is what remains constant.)
