Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?

[Larson]

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

Also, I'm really curious as to how sensitive (or not) the ESP8266 2.4Ghz ESP-NOW communications will be to interference in a home environment that, at least so far, seems to have shredded 2.4Ghz LoRa.

My ESP-NOW (8266) 2.4 Ghz experience has been pretty uninterrupted that I know of. For about 3 years, I've been monitoring the water pressure of my home system. The sensor/ESP transmitter is at the water heater and the receiver is remote. To give some sense of the density of my WiFi environment: my browser scan detects 7 networks and I live in a medium-density neighborhood (about 4 homes/acre), so it is not too bad. Funny, the recharge curves of the system water pressure, after use, looks just like a capacitor voltage in a RC circuit, only much slower.

Unlike the low-power theme of this thread, my ESP’s drink as much as they need as they are always on and mains powered.

[Larson]

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

It does a nice job of measuring milliamps. I wouldn't say the PPK2 is good at measuring <1ua currents though, such as sleep currents. If that matters to you, you'll want a Current Ranger, or TinyCurrents or similar.

Out of gratitude I'd like to buy a TinyCurrent for you. Saw the Donate button on the openhardware.io. Does that go to you?

The TinyCurrents device would be interesting to own. But if the granularity of the PPKII (still on backorder) is down to 1uA, then anything more exceeds my practical interest. For my hypothetical 2000mAhr battery, a burn rate of 1uA would last 228 years... again, if I did the math right. That exceeds my interest, though it is fun to chase down the details.

Been re-reading this thread for days. I want to help and hope to, if I can keep up with the blog entries. I've got a shopping list running - some of it already ordered!

[NeverDie]

@Larson That's very generous of you, but by no means feel obligated. The TinyCurrent-R would be the one to get. Same price as the TinyCurrent, but it has the BNC connector facing the correct direction for connecting directly to an oscilloscope's input terminal.

I do this stuff because it's both challenging and fun. I enjoy the precision of it. I just keep chipping away at it and eventually make progress. If my posts are of any use to others, then all the better: it all works best when everyone helps everybody else, because collectively we know more than any one of us individually. We've barely scratched the surface of what can be done. Synchronizing clocks among all the different motes and doing time division multiplexing among them would offer ultimate efficiency I think. The technology to do that is already known but the details of how to do it are not on display in a manner that's widely accessible.

[Larson]

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

We've barely scratched the surface of what can be done.

Openhardware.io gift coming. Please use it to replace your missing uCurrentGold with the proceeds. Someday I’ll contribute my own projects to these splendid forums, I hope.

[NeverDie]

@Larson Thanks! I purchased it off Ebay, which says I should receive it by July 15,

though my guess would be probably sooner than that. I'll post a picture after it arrives.
In combination with an o-scope, It should be good for those occasions where there's a need to capture very fast action in fine detail. :-)

[Larson]

@NeverDie Super. It will be fun to see what you do with it. Glad you have the OC scope to leverage the tool. Cheers.

[NeverDie]

@Larson You're doing the right thing by getting the PPK2. It will do 99% of what you want, and much more easily than an oscilloscope.

By the way, in case anyone is interested, there are a couple of reasons why a direct-to-oscilloscope connection beats hooking a typical 1X/10X oscilloscope probe up to, say, a uCurrent Gold or TinyCurrent. The first is that, obviously, there's less noise involved because of the short connection. The second reason is far less obvious, though, which is that the typical 1X/10X oscilloscope probes make for fairly lousy 1X connnections. For measuring small voltages on a 1X connection, you're better off using regular coax or, better still, a direct connection and eliminate the coax altogether. Why? Well, the TL;DR is that oscilloscope probes are optimized for use as 10X (which really meas divide by 10), as explained in the EEVLOG video:
https://www.youtube.com/watch?v=OiAmER1OJh4

---

Meanwhile.... I've been trying to get Channel Activity Detection to work on the SX1280, but so far no joy. In some sense it's its own mode, because it "supposedly" times out automatically and returns to Idle mode after a fixed number of symbol periods that's set by SetCadParams. So, I'm puzzled as to what's going on and will have to take a deeper dive to figure this one out. Unfortunately, the library does not support CAD, so there's no working example to start from. There have been some alternative libraries posted on github, so I'll see if they shed any light on how to get CAD working.

[NeverDie]

Well, this is a good example of how having a PPK2 helps to troubleshoot the CAD problem. We can see from the PPK2 that the CAD actually is activated, once per second, as per my code:

So, just knowing that narrows down the problem considerably. The problem now is no longer that but instead why it isn't setting the IRQ flag to indicate that it has finished its detection cycle? The Rx completion flags aren't being set either.

[NeverDie]

@Larson You'll want to add some jumpers to your BOM if you don't already have them:

The above are two motes. The one on the left is used for measurement, and the one on the right isn't.

[Larson]

@NeverDie Super, thanks for pointing out the jumpers - very clear to see how they would be used.

For a long time, I've wanted to make ESP programmer boards that would use springs to temporarily hold the ESP. I actually built one using pogo-pins and a 3-D printed housing pocket. It worked but was fiddly. I may try again on your radio carrier board but differently this time. I’d like to use stiff carbon steel music wire that would be mounted in drilled holes in the existing pads and create new pads on the back-side for mechanical-solder support. Some trace revision would be in order on the back-side, and Captain Tape could cover the new solder pads if needed. Think of it as a bed of nails. Carbon steel music wire is used in springs, I think, and that is what I'd like to use for its stiffness. Perhaps I’ll make the dimension between the rows of holes about 1 mm shy of the distance between the castellated indents. Then to use, push the SX1282 down into the field of springs getting 4 or 5 mm from the carrier board or until the springs start to splay out. If it works, the castellated indent would snuggly fit the wire if I can select the right gauge. One board could then serve different radios of the same class with the same pinouts. The downside may be that I have 16 new unintended antennae. Maybe I could trim these springs shorter after learning how it works.

The compelling reason for this is to preserve the radio for later permanent installation elsewhere, if needed. There may be some performance degradation as the effective trace length would be longer. I have no idea of the electrical properties of this wire. I'll find out. I’ve got too many ESP’s that are permanently attached to dead testing, or otherwise abandonded, boards. What do you think of the idea?

Atmega328P selection question: is this linked choice the right pin configuration for the footprint you used? There are so many other choices. Digikey is out of stock, but I'm presuming I'll find them elsewhere.

[NeverDie]

@Larson AFAIK, any atmega328p with a 32TQFP package should work. Yes, they are in surprisingly short supply right now. Because of that, some people (like Great Scott!) have taken to desoldering them from Pro Mini's, which aren't cheaper but are easier to find. If you go that route, I recommend you buy some ChipQuik to make your desoldering a whole lot easier: https://www.amazon.com/gp/product/B0019UZP7I/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1 But that's me. Maybe you have your own preferred method. Whatever works. If you have old projects you no longer use, now is a good time to recycle their parts.

[NeverDie]

@Larson said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

What do you think of the idea?

I think it would be easier and less iffy to simply solder it on as per usual and then use ChipQuik if you later wanted to desolder it and move it to another board. But again, that's me. That said, I somewhere have one of those springy jigs that I bought on aliexpress for programming virgin ESP8266's, and it seems to work fine--well, at least for the purpose of installing bootloaders. I never tried to do wifi with them in such a jig, so I don't know what the RF consequencies might be, if any. If you can tolerate some impairment, maybe it's good enough.

[Larson]

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

... then use ChipQuik if you later wanted to desolder it...

Okay. One more item for my BOM and skill set. Thanks for the 32TQFP tip also. I do happen to have several Pro Minis and Nanos on hand since I bought them by the dozen - but they are so perfect as they are. Okay, I'll risk one and see how it goes. I am very excited to test the power difference between fuse settings and excluding the external crystal. Andrew Retallack wrote a great series on Beyond the Arduino for power savings and programming in C. I'll have to go back and look but I think it all was based on the external crystal.

[NeverDie]

@Larson Here's all you need to know in one place on how to get to 100na sleep current on the atmega328p: https://www.gammon.com.au/power He covers all the different methods. TL;DR: burn your fuses to:
External: 0xFF
High: 0xDC
Low: 0xC2
and then run Sketch J from the Nick Gammon link. That's all there is to it, provided you have an external device/circuit to wake up your atmega328p.

[Larson]

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

Meanwhile.... I've been trying to get Channel Activity Detection to work on the SX1280, but so far no joy.

Is there a CAD register that can be polled? In the SX1280 datasheet section “11.6.8 SetCAD” there is discussion of a CadDone IRQ. Maybe that can be used outside of the library? Maybe with its own interrupt? It looks to be bit 13 of the IRQ Register per table 11-73. I’ve never played with these radios but have found other devices (simpler devices) to be register friendly. Most of the references to CAD are in section 11. You, no doubt, have already been there. If so, I leave the comment for others.

[edit: on second read I see you have already been to section 11 with your comment on SetCadParams. Hmmm... wish I knew more. I'll play with it when I get set-up.]

[NeverDie]

@Larson said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

Meanwhile.... I've been trying to get Channel Activity Detection to work on the SX1280, but so far no joy.

Is there a CAD register that can be polled? In the SX1280 datasheet section “11.6.8 SetCAD” there is discussion of a CadDone IRQ. Maybe that can be used outside of the library? Maybe with its own interrupt? It looks to be bit 13 of the IRQ Register per table 11-73. I’ve never played with these radios but have found other devices (simpler devices) to be register friendly. Most of the references to CAD are in section 11. You, no doubt, have already been there. If so, I leave the comment for others.

[edit: on second read I see you have already been to section 11 with your comment on SetCadParams. Hmmm... wish I knew more. I'll play with it when I get set-up.]

Yeah, I'd say the Semtech SX1280 architecture is quite different than a lot of previous Semtech chips, which, you're right, in prior times were very register oriented. There still are some registers, but for SX1280 Semtech adopted a less transparent op-code architecture where you can issue commands--but in some cases there's no way to verify whether the command "sticks" like there is with a register oriented architecture, where you can always read-back the register to confirm that what you think you wrote to a register was actually written exactly as you wrote it. This is making troubleshooting relatively hard, because the chip has some internal states that you can no longer read like you could with a more register oriented architecture. Instead, you send an op-code over SPI, and maybe some parameters to go with it (the type and number of which can vary), which you "hope" works. I don't like it at all, because, among other things, it assumes that the documentation is 100% complete and flawless--and documentation never is. In particular, the SX1280 datasheet never gives the exact protocol for launching CAD, and whatever that protocol is, I seemingly haven't found it yet, despite testing numerous possibilities--though I may have gotten close. And in a system with more than one SPI device, what happens if there's some conflict in your SPI communication? How do you ever unravel what it is if there's no way to verify? Seems like heartache waiting to happen. I don't understand why the opcode system is thought to be better. Anyone know?

Anyhow, Stuart's Projects never implemented CAD, and internally I'm finding it makes, IMHO, some seemingly improper assumptions about which IRQ's are assigned to which DIO's in how it operates. That makes it brittle. That's something one just can't know when looking purely at a library's veneer. Anyway, the good news is that late last night I found a different, more general library which claims to have a working SX1280 CAD example, so I'll be giving it a test drive soon. https://github.com/jgromes/RadioLib It's not radiohead, but it's a similarly ambitious library project that spans multiple different kinds of radios and multiple different kinds of MCUs. If the CAD demo code works, then I'll adopt the new library and drop Stuart's Projects. If nothing else, I admire the authors ambition to be the one library to rule them all. As I hate changing libraries, I truly hope this is "the One" library that can do that. :-)

[NeverDie]

Well, the Radiolib CAD example compiles and runs, and it nominally looks like it should:

I just need to introduce a separate LoRa transmitter to the vicinity to confirm that the detection actually works. I'm guessing the LoRa parameters have to be an exact match (SF at minimum and maybe other parameters as well). Then, sending a long enough string--and maybe reducing delays in the CAD detector loop--should lead to a detection.

[Edit: Yup, I tried running the generic SX1280 Radiolib transmit example, with delay removed, and the CAD detector easily detects it. It works! :-) :-) :-) ]

[NeverDie]

Since the Barebones v001 is, after all, a test platform, I'm going to assemble a mote and SX1280 adapter that use Dupont headers. The reason is that the 8 sense pins on the PPK2 work with Dupont wires, and I don't see an easy way to make those work with machine pin headers. No change in the PCBs is required. @Larson You may want to assemble your motes this way from the get-go.

By the way, can anyone here recommend where I can buy "good" Dupont wiring and connectors? If I buy from China, I seem to often get very loose-goosey connections.

P.S. I could be wrong, but the RadioHead library seems to have stopped progressing. None of the newer LoRa chips are in any of its examples. Too bad, as it was a very progressive library in its day. That leaves RadioLib as the library to rule them all. However, it has a very different overarching paradigm than, say, StuartsLibrary, which was easier to grasp.

[Larson]

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

Dupont headers

My plan is to use female headers on the Barebones board, then to use doubled long-legged male headers on the SX1280 adapter board like these from Digikey. It all stacks together and bed of nails will stick out the top of it all. That way the PPK2 Dupont females can nicely plug onto the any of the protruding males in the stack that I select to use as a signal or interrupt. Again, I risk a bed of 16 inadvertent antennae with all that metal in the air. I can always snip-off the pins I don’t use. PPK2 is expected on Monday!

I spent the day shopping for parts at Aliexpress, Digikey, and Amazon. I found some 328P's at Aliexpress. Could be some time for the Aliexpress parts (SX1280 module). In the meantime I'll be learning KiCAD. I'm hoping that Eagle and KiCAD might allow file swapping as Eagle is familiar ground for me. Or ... I could just take the easy path and go directly to OSHPark since you have made it so easy.

[NeverDie]

@Larson said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

@NeverDie said in Anyone using/tried the E28-2G4M27S 2.4Ghz LoRa SX1280 27dB module?:

Dupont headers

My plan is to use female headers on the Barebones board, then to use doubled long-legged male headers on the SX1280 adapter board like these from Digikey. It all stacks together and bed of nails will stick out the top of it all. That way the PPK2 Dupont females can nicely plug onto the any of the protruding males in the stack that I select to use as a signal or interrupt. Again, I risk a bed of 16 inadvertent antennae with all that metal in the air. I can always snip-off the pins I don’t use. PPK2 is expected on Monday!

Somehow I had gotten it in my head to use extra long female headers (such as these, https://www.amazon.com/gp/product/B071GTP5V6/ref=ppx_od_dt_b_asin_title_s00?ie=UTF8&psc=1, or similar) on the adapter board and then regular female headers on the basic atmega328p PCB. Maybe it's a wash?