# Inner life of a LoRa module - the E22-400M33S _Build Guides · started by aetherlab on Tue, Sep 22, 2026 3:01 PM_ --- ## Original post **aetherlab** · Tue, Sep 22, 2026 3:01 PM Hello to all! Some time ago i failed to properly desolder an E22-400M33S from an aethernode board, damaging the tracks in the process. I kept it and opened it to see what EBYTE did in it and do some hardware path resolution for the purpose of learning and just plain curiosity food. ![IMG_20260912_131531_792.jpg](/storage/forum/JEtVpjExPWdntmOmFgzCQN5iYw3g16dYesvNDMg3.jpg) For those who don't know, the E22-400M33S is a mid-band (there are also 120MHz and 800/900MHz units) SPI interfaced LoRa module with output power of 33dBm or 2W, and a built in LNA or Low Noise Amplifier in the receive path. I use those for the aethernode and aethernodeS3 base nodes in Sofia. After cutting off the shield, this is what we are met with: ![IMG_20260922_155909_313 (2).jpg](/storage/forum/6Hntq4HFMS9o3CqElw0qsBIAiIQqbQFc30aOYX7k.jpg) Let's take a closer look and see how this module is built. It employs the SX1268 by Semtech, which on it's own can muster up to 22dBm of output power, around 158mW. So for us to get to 2W we need a power amplifier, a transmit/receive switch to connect the antenna to the respective path, and some other stuff. All my explanations are transferable to the low and high-band modules, with their respective frequencies, and if you note empty SMD pads along the ways, those are used in the other models, so there is only one PCB design to produce for universality. ![IMG_20260922_155909_313.jpg](/storage/forum/cd7vHNtIrUeGGraIMXbOBIBAmRelBNQwq5HzchTO.jpg) In **blue** here i have depicted the antenna path. It leads from the IPEX antenna coaxial connector, trough a filter network, to the transmit/receive switch (for short TX/RX from now on). The filter network actually does two tasks simultaneously so it has to be carefully designed and executed. First it forms a band-pass filter, that will attenuate anything below 410MHz and above 493MHz, the working range of the SX. This will not only keep away unwanted signals, that might be strong and interfering, but will also cut local harmonics, created by the power amplifier (PA for short from now on). For example at 433MHz, the PA will generate a formidable signal at 866MHz, and a smaller one at 1299MHz and so on, up to 10-th order. Sometimes amplifiers, especially when pushed too hard and in clipping (the amplitude of the amplified signal reaches the power supply voltage and cannot rise anymore, creating a "plateau" in the shape, that is very bad signal-purity-wise), generate sub-harmonics, in our example they will be around 216.5MHz and below, or might not even be harmonically related (not a divisor by two). All these possibilities are addressed by this filter, and other within the module. Second, the filter is an impedance transformer. The power amplifier expects a certain load impedance, onto which to project it's power in full and with minimum distortion. The SX also expects some load impedance both on it's inputs and outputs. For the sake of simplicity, let's take the classic - 50 Ohms resistive load. The 50 is widely accepted, and whilst other impedance systems and components exist, it is kinda the defacto standard. Why? To put it simply - efficiency in power transfer, because of geometry of waveguides (cables). But this is simply put. No, I am not getting in this rabbit hole, sorry, do your own homework if you wanna know, haha... So the filter also transforms, as all on this board do, all the impedances to around 50 ohm resistive for the 410-493MHz range, so the whole thing can work smoothly. Next stop is the TX/RX switch. It is an analog multiplexer, that switches the antenna port either to the PA or to the RX path, depending on if we transmit or receive at any given time. We have only one antenna, so kinda reasonable. The switch is controlled by the RXEN and TXEN ports of the module. They are separated for...reasons. I dislike this choice in the face of an RNode, but this allows for a third state to exist - isolation. Meaning you can kinda mute the antenna input. This might be useful if you have other powerful TX's around that you need to close your ears for, when they transmit, so you don't damage the SX's RX input, or the LNA input. Or if you have an algorithm to measure internal noises in the system, you can use this to actually hear only them, but meh. To control the TXEN switch we normally use the SX itself - it has DIO2 output settable to TXEN control, and I always use this, because of all things in the system, what knows better and best WHEN transmission will occur, than the base-band chip itself? So if we externally (outside the module) connect DIO2 to TXEN, we can set this to be the perfect transmission timing. What remains is the caveat of separate TXEN and RXEN pins and my object of meh - who will tell the module to LISTEN? In this case we can either use another pin from the CPU to do this, it is in the RNode firmware, OR we can do something very simple and in pure hardware, like I have done on the aethernodes - if you check the top picture you will see a resistor and a transistor above the module on the PCB. I use this as an inverter, and whenever TXEN is low, it pulls RXEN high. This way no IO is being wasted on this purely internal for the SX base-band business. Next we venture in the RX path, which is in **green**. The first thing you can see is a 3-pin part, named PD, or Protection Diode. It is a set of two very fast diodes, having very low capacitance (so they do not interfere with the signals), that are connected between the RX path and ground. What they will do is actually limit the voltage on the path to +/-0,48V, kinda crowbar - protecting the rest of the path from too strong signals. The TX/RX switch can withstand a ton of abuse, but the LNA and SX not. After this we have the next set of filtering parts, that provide more bandpass goooody and match the path to our RX star - the LNA. It provides around 17-19dB of gain and allows for greater sensitivity. The price for this is reduced SNR (signal to noise) ratio, because we amplify all, good and bad. (more info on this [here](https://rns.recipes/forum/help/what-do-q-and-s-mean-on-the-rnode-interface#post-1159)) After the LNA we have even more filtering and impedance matching and we get our signals to the SX for receiving duty. For the TX path in **red** we will start off at the SX. The internal 22dBm PA of the chip has a balanced output, meaning it outputs the power not between one pin and ground (single ended), but between two pins deferentially. This is very useful in different applications and designs, but here it is not needed, so the filtering network not only filters and impedance-s, but also turns differential to single-ended. Than we have the star of the TX path, our PA. The gain of the amplifier varies, depending how hard we push it, meaning for 433MHz it compresses a bit. If we look in the datasheet, we have the following: ![image.png](/storage/forum/NqANDukxO2EWfiOhLzbI6gj5kUfIkef6BOJTpaKF.png) This tells us, that at 433MHz to get 33dBm out, we need to drive the amplifier with 21dBm from the SX chip. Or in other word, we need 0.126W from the SX to get 2W out of the module, which is 12dB of gain. After the PA we have more filtering and we get to the TX/RX switch and on to the world trough the antenna port. Lets now close with several words about the service part of the module. The TCXO or Thermally Compensated Crystal Oscillator, is a CRITICAL part of LoRa chirp generation, especially at spreading factors over 7. It is the beating heart of the SX chip and it's clock, and thus all timings within are measured in cycles of the TCXO. If the generated frequency is not exact or stable, all will fall apart and modules will not intercommunicate well. Above is a block of power supply and filtering, comprising of the Low Drop-Out stabilizers. The leftmost takes the 5V input and makes 3.3V for the SX. The second one, LDO2) powers the TCXO, separately, because all the digital switching action in the SX actually makes lots a noise on the powerline and will pull the TCXO around a bit, and we do not want this. LDO3, the one with the massive 2R2 inductor-based filter on the output, powers the LNA, it needs a freaking QUIET power source, because it will amplify like crazy all it senses and we will lose signal fidelity. The PA is powered directly from the 5V input line, with some small filtering, which I consider a bit on the cheap side and this is why I always have this filter network on my designs, that separates the module from the rest. And last, but not least, note how all digital I/O lines form the SX are fed trough black ferrite inductors, line filters? This is critical, as if ANY rf gets back onto these lines, the SX will just malfunction, and we have plenty of RF around, from our own PA. So that's it! Smaller modules, like the E22-XXXm22S have no PA, only RX/TX switch. And some cheap crap you should stay away from, like NiceRF, have no switch altogether, replacing it with a lossy combiner-splitter filter...meh... Bill --- ## Reply 1 **Mark** · Tue, Sep 22, 2026 3:29 PM Wow Bill, this was a very interesting and insightful look into the module! Thanks so much for writing it up and sharing the de-lidded closeups! Good stuff! ---