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

[NeverDie]

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

As you are working with JLCPCB are you using them for assembly? I did so last year with different project and had really good results. I did have some difficulty finding compatible components, so I suspect that they won’t be able to resource all your radios… but that was some time ago and I have much to learn. I look forward to doing more assembly with JLCPCB. It was also really inexpensive and a great way to get around my poor home soldering quality.

Good idea. I haven't tried it, because I've been locked in a mindset of using radio modules, in which case soldering things is easy enough that I can do it myself manually without much effort. But for small, fine pitched chips (the kind I've been avoiding for just that reason), it sounds like a worthwhile thing to try. Maybe I could use better components than are typically found in aliexpress modules as well. :grinning:

[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?:

If you know of chips that you're keen to try, let me know

Since you asked, maybe look at the good-old cheap 433 MHz transmitter/receivers? Maybe you are sticking to transceivers so this wouldn’t fly. Or maybe you are staying away from ASK modulations. I’m sure these cheapo radios won’t pass your fundaments concrete/cross-house test, but it would be another data point. This season I’ve been slowly reworking my limited 433 MHz home network of motion detectors. Having failed to transmit through concrete/cross-house pathways, I built some 433 repeaters that successfully worked on a line-of-site pathways that allowed me to look around the corners. I have found these TX radios & SR501 motion detectors just sip energy and my 4-pack of AA batteries last for, in some cases, over 3 years.

Interesting idea. I'll have to give that one some thought. A lot of the cheap wireless temperature-humidity sensors seem to use ASK, and I've never been clear as to why. e.g. Oregon Scientific, and similar companies. There are a number of hacker-type projects aimed at receiving and decoding those signals precisely because the wireless sensors are so cheap. And those cheap wireless key-finder fobs I think may use it as well. Maybe the receiver current can be extremely low? Those fobs always have to be listening for a wireless signal. All I know is that low frequency implies less energy consumption by whatever oscillator is being used.

[Larson]

@NeverDie: Yes, as you describe elswere, and taught me, the design trade-off parameters are frequency, bandwith, dwell time, bit transfer rate, power demand, and most importantly, range. I'm looking forward to your findings.

Regarding the low reciever current: don't know about the FOB's but I have found the transmitter power to be key. My design has the "always on" reciever connected to mains. so I don't worry about that power. But the battery-limited transmitter devices are brought to life and triggered by the signal pin of the SR501 sensors that is high enough, long enough to complete a transmission, or several. The HT12E (encoder) chip is connected to the SR501 pin is fired and sends the settings of the dip-switch pin hi/low settings that is recieved by the always-on HT12d (decoder) that the Atmega 328PU can understand. I think the HT devices are MCUs in a different form, like what I imagine a FPLC to be like... don't know for sure. Why do I drone on? Because of the HT12D and HT12E pariings are very specific and very low on power demand due to their design. When I built my configuration (6 years ago) I knew more than I do now. But I'm relearning more. Andreas Spiess, Great Scott, Big Clive, and YOU, explore this stuff extensively, and I thank you all.

In summary, my repeaters and base RX station are mains powered so I don't worry about power & batteries. The primary focus for conservation of batteries is on the TX/Motion devices. Everything downstream of TX is mains powered or can be suitably backed up in the short term.

[NeverDie]

I think I'll pass on the ASK, and the reason is that it tends to be noise/interference sensitive. The usual workaround for noise/interference by ASK garage door openers, for instance, is to send the same packet multiple times, with the hope that at least one of them will get through. Well, I suppose that strategy works "good enough" in some sense for that application, but I'd say that all those extra packets being sent is a waste of energy for a battery powered sensor application that's transmitting, say, once every 5 minutes. On the other hand, if it doesn't really matter if you lose packets from time to time (e.g. a wireless TH sensor), maybe you don't need to send multiple redundant packets in the hopes that one gets through. That said, IIRC, Oregon Scientific TH sensors send their packets in triplicate. Well, that's my off-the-cuff take on it. Feel free to disagree if I'm missing something important.

Anyhow, I think that's why one tends to see ASK being used at 433Mhz: it's a relatively clean channel, so the noise/interference isn't as big a factor as it would be at 915Mhz or certainly 2.4Ghz.

At the end of the day, I think ASK was popular because, historically, it was easy and cheaper to make. Think 1940's, 1950's, and maybe 1960's, before integrated circuits. Now that FSK, its main competitor, is widespread and cheap anyway, I just don't see much benefit in reverting to old school ASK unless maybe you're manufacturing millions of something and want to save every penny possible. Because of the possible noise issue, you're also limited to slower bit rates with ASK than the alternatives, which, in turn, also lead to longer transmit times.

P.S. Build time at JLCPCB is now running 2-3 days. Out of the 10 boards I submitted on Sunday (China time), 9 are still awaiting "data preparation." Only 1 of the 10 is currently in production. I think they're manpower limited right now, maybe because of China's zero-tolerance covid policy.

[Larson]

@NeverDie Yes, you are correct: the HT12D datasheet indicates that multiple signals must be recieved before the enable pin is activated. I think the 'multiple' is a count of three and there is probably a preamble in advance of the main message of high/low settings. The transmitter, again I think, just keeps sending the message until the SR501 drops power (2 seconds on the minimum settings). I get alot of double hits because of this. Good to know about the noise and history. Thanks.

[NeverDie]

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

@NeverDie Yes, you are correct: the HT12D datasheet indicates that multiple signals must be recieved before the enable pin is activated. I think the 'multiple' is a count of three and there is probably a preamble in advance of the main message of high/low settings. The transmitter, again I think, just keeps sending the message until the SR501 drops power (2 seconds on the minimum settings). I get alot of double hits because of this. Good to know about the noise and history. Thanks.

Worthy of note is, as you may already know, you can skip the HT12D and HT12E altogether if you're driving it from an Arduino. You can do all the encoding/decoding in software, and all you need on the transmit side is a mosfet to control the on/off of the transmit on the transmitter. I did this once by rolling my own software to do it, but I think it may also be handled in the wire library, IIRC. On the receiver side, I got better results by upgrading to an Arduino Due, because it has 12 bit instead of 10 bit Analog-to-digital converter, and a faster speed to do the processing. However, maybe I could have done it all digital and skipped that. I was just trying to eek the most reception I could out of possibly weak signals.

Fun fact: on a 900Mhz Raspberry Pi (Version 1), you can make a 900Mhz radio transmitter just by putting a wire of appropriate length in one of the GPIO pins and then toggling PWM to it at the 900Mhz clock cycle. Of course, you'll need something more sensitive on the receive side. This would be an ASK scheme, by the way.

Also, fun fact: the speed on the Due is fast enough that light (radio waves) doesn't move all that far per clock cycle, so you can detect when signals arrive directly between transmitter and receiver from an on-pulse, and you can also detect fainter echoes of that same pulse arriving later because they've bounced off a wall or something and took a longer path between the transmitter and receiver. I thought that was kinda cool, because I ordinarily think of the speed of light as being more or less instant in such close quarters. Anyhow, I forgot the name for this kind of temporal transmit "smearing", but it's a very real effect where the transmitter kinda interfers with itself, so to speak, because of the layout of reflecting surfaces in the environment. [Edit: just now looked it up. "Multipath interference" is the term. MIMO thrives on multipath signals, but it's a problem for regular radio to deal with it--though, as you'd expect, there are all kinda of ways to handle it; the easiest being just putting a long enough time interval after an ON pulse for the echoes to die out to background noise levels. ] :-)

P.P.S. All 10 of my JLCPCB boards are now in production, though it is now technically early Tuesday morning in China. Looking at the time stamps, the production process itself appears to run around the clock, not just during daytime business hours, so that's good. Hopefully that means there's no complicating factor left standing in the way of completing the production and shipping it.

[Larson]

@NeverDie : Very interesting stuff. Even at the lowly 8MHz, the speed of processing pretty much blows away the human concept of time. Fun to hear of the Raspberry and Due. I’ll have to play with those.

HT12D/HT12E: Yes, I learned that the processing could be done on a MCU and I did that on the repeaters I built. I put a Pro-Mini in between a 433 Mhz RX and TX. This repeater sits and listens in the RX mode until it is pinged with a transmission from one of the 433-HT12E detectors. Once a message comes in, the MCU then switches to TX mode and passes the message along, then reverts to listening again. The repeaters and the base station are mains powered as I figured the always ON state would be a battery killer whereas the 433-HT12E’s are very low power and outfitted with batteries.

Rather than using the IIRC library I used RCSwitch – probably because I found that first. RadioHead was another option. The biggest challenge for me was learning which protocol to use for the 433 radios. The command mySwitch.setProtocol(11); did the trick.

[NeverDie]

@Larson The Due has an 84Mhz clock speed, making it easier to spot the effect. These days you could save some coin by using an ESP8266 (80Mhz) or even faster ESP-32, not to mention the crazy fast Teensy 4.1 (600Mhz). At 600Mhz, light moves only 20 inches per clock cycle. Pretty cool for cheap parts, isn't it? :-)

[NeverDie]

I received the Dorji parts I had ordered. They arrived very fast and very well packed. I got them on ebay, but for comparison I'm still waiting on everything else that I had ordered from aliexpress.

It turns out it took 3 days for JLCPCB to build my simple 2 layer PCB's. Either they're overwhelmed with business or understaffed. With any luck they'll ship out tomorrow, so probably another week to get here.

Meanwhile I received a Nord PPK2 Power Profiler, as I found a seller who briefly had some in stock. So, I'll be putting that to use to accurately compare current consumption of the various test builds after the PCBs eventually arrive from JLPCB. Meanwhile, I may try it out on an ESP8266 to finally get some solid current consumption numbers on how much power it takes to wake-up into ESP-NOW mode from a cold start. If it's good, then maybe so-called trigger boards will bring the ESP8266 back to life as a low power competitor. 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. According to the labeling on the tin, they come with 25dBa power amplifiers, so I expect they'll do far better than a bog standard nRF24L01, which has a maximum of only just 0dBa transmit power. Of course, when you add in the cost of a trigger board, then total cost may (?) turn out to be a wash compared to other mcu/radio combos that don't require a trigger board. We'll see!

[NeverDie]

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

I'm hoping (but haven't confirmed) that by eliminating the oscillator on pins D20 and D21, I can use those pins to drive two LED's kinda "for free" since nobody uses those pins for anything.

Reporting back: I have the answer. It turns out that the standard arduino core for atmega328p that's baked into the standard Arduino IDE does not support Arduino pins 20 and 21 as digital GPIO pins for driving LEDs. However, the good news is that there's an even better Arduino core, called MCUDude MiniCore which does support exactly those pins for such purposes. Here's the TL;DR:

This core gives you two extra IO pins if you're using the internal oscillator! PB6 and PB7 is mapped to Arduino pin 20 and 21.

https://mcudude.github.io/MiniCore/package_MCUdude_MiniCore_index.json
It's very easy to use. You can install it into the regular Arduino IDE, pick from among the MiniCore "boards" in the board manager, select the 8Mhz option and a few other obvious options, and then you're done with instalation. From that point on your code will automagically compile using MiniCore. Just to be sure, I gave it a try myself, and I'm now blinking a blue LED off of Ardino Pin 20. It works!

[Larson]

@NeverDie More good information - thanks for posting. Good work on finding the alternate core. I have a couple of questions:

1. By “trigger board”, are you referring to the TPL5010 discussed above? (Adafruit carries the TPL5110, I think.) Is the idea here to power and de-power the radio, or mcu, to discover the rock-bottom minimum current requirement?
2. What are you sacrificing when not using the external crystal? And to do this, are you setting fuses to tell the 328 to use the internal crystal? I've found setting fuses to be a complicated affair.

[Larson]

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

At 600Mhz, light moves only 20 inches per clock cycle. Pretty cool for cheap parts, isn't it?

Yes, mind-blowing. I got into this hobby via the Parallax Basic kits. The second MCU I bought from them was about $60 and I thought that was stunning. So when I found the ATTINY85 I was addicted, if only because the access to this technology costs less than a cup of coffee. The progression to Arduinos and ESP's was a natural extension for this cheap hobby.

[NeverDie]

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

@NeverDie More good information - thanks for posting. Good work on finding the alternate core. I have a couple of questions:

1. By “trigger board”, are you referring to the TPL5010 discussed above?  (Adafruit carries the TPL5110, I think.)    Is the idea here to power and de-power the radio, or mcu, to discover the rock-bottom minimum current requirement?
   

2. What are you sacrificing when not using the external crystal? And to do this, are you setting fuses to tell the 328 to use the internal crystal? I've found setting fuses to be a complicated affair.

Good questions!

1. Yes, by "trigger board", the adafruit tpl5110, or something like it, is what I have in mind. The goal would be zero leakage current beyond the ~40 nanoamps of management overhead. I don't know if the p-channel mosfet design in adafruit's tpl5110 breakout board achieves zero leakage (or near enough),but it would be worth measuring. I think Kevin Darrah may be the one who coined the term "trigBoard". His design is a lot more elaborate and, I assume, costly: https://www.kevindarrah.com/wiki/index.php?title=TrigBoardV7

Is the idea here to power and de-power the radio, or mcu, to discover the rock-bottom minimum current requirement?

Yes. It's a little more complicated though, because generally the mcu/radio wake-up time is much longer from a cold start than from a deep sleep, so the real objective would be to find the breakeven point at which power-off is the better choice (i.e. how long does the duty cycle need to be).

What are you sacrificing when not using the external crystal?

Accuracy. If you were interfacing to something that had extremely tight timing requirements, the crystal would be more accurate. In practice, I haven't ever noticed a difference in anything that I've done.

And to do this, are you setting fuses to tell the 328 to use the internal crystal? I've found setting fuses to be a complicated affair.

Yes, I set the fuses. Currently I use the following fuse settings:

By running from the 8Mhz internal oscillator and sleeping, not only can you sleep the atmega328p at just 100na, but as a bonus the atmega328p can wake up in just 3 microseconds (I previously measured it with an oscilloscope, so that number is solid). For being such an old mcu, the atmega328p can be amazingly high performance when it comes to saving power. This becomes of greater relevance if you're getting your power by harvesting weak energy sources.

If you want to play around with setting fuse bits, the standard advice is to leave the Extended bits alone until you know what you're doing. Even then, there's probably no need to change them. It's the High and Low bits that are of relevance.

A great source of information is: https://www.gammon.com.au/power

[Larson]

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

sleep the atmega328p at just 100na

Wow, that is pretty close to zero in my view. Consider battery self-discharge in comparison: I use 18650 LiOn batteries in most of my projects. Say a 2000 mAh battery loses 1%/month in self discharge s(ome say it is much higher). Well, the math, if I did it right, shows a self-discharge of about 25 uA. So sleeping at 100nA is really impressive.

Thanks for the fuse guidance. I will employ this next chance I get.

[NeverDie]

@Larson Yes. The RFM69 can also be slept at just 100na, so taken together they are an awesome duo. If you're running from 2xAA batteries, then the the combined 200na are essentially zero. If, in contrast, you're trying to harvest energy from a small solar cell in an indoor environment that's dimly lit by LED lighting only rarely, then it becomes more relevant.

Anyhow, I've noticed that a lot of people measure very small currents incorrectly, so just trying to find answers on the internet, which is full of wrong measurements, isn't easy. Datasheets can provide some insight.

[Larson]

@NeverDie Yes, the experience and knowledge found by doing (measuring) is always deeper,

Forgot to mention: Kevin Darah. I have bought several of his TrigBoards. I find the price to be a good value for the design, components, assembly, and all the content he has posted. The quality of his boards is great. I am a Patreon of his. I just need to play with the TrigBoards some more.

[NeverDie]

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

@NeverDie Yes, the experience and knowledge found by doing (measuring) is always deeper,

Forgot to mention: Kevin Darah. I have bought several of his TrigBoards. I find the price to be a good value for the design, components, assembly, and all the content he has posted. The quality of his boards is great. I am a Patreon of his. I just need to play with the TrigBoards some more.

I'd be interested to hear your thoughts on Kevin's trigBoards after you've had a chance to get familiar with them. A compare/contrast of his board to the adafruit board would be even more awesome. The adafruit board is a very simple design and would cost less than $1 in parts if you were to source them yourself. I suppose its main limitation is 1. the maximum sleep time is two hours, and 2. not as accurate as a proper RTC. In its favor is that at 40na the sleep current consumption is next to nothing. On the other hand, some RTCs have a sleep current less than 1ua, so maybe that's good enough.

[Larson]

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

In its favor is that at 40na the sleep current...

I'll order several. Then compare to the TrigBoards. Adafruit's TPL5011 advertises a run current of 20uA. Is that acceptable for you? If so I'll buy several. Couple of limitations for me: 1. Time - I'm about a month out, 2. My Ammeter is a standard DMM from Harbor Freight. The smallest range is 200uA and I've found it very useful. Certainly fine enough to show the 20uA, but inadequate for nanoAmps.

[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?:

In its favor is that at 40na the sleep current...

I'll order several. Then compare to the TrigBoards. Adafruit's TPL5011 advertises a run current of 20uA. Is that acceptable for you? If so I'll buy several. Couple of limitations for me: 1. Time - I'm about a month out, 2. My Ammeter is a standard DMM from Harbor Freight. The smallest range is 200uA and I've found it very useful. Certainly fine enough to show the 20uA, but inadequate for nanoAmps.

Hmmmm... Something's wrong then with Adafruit's design if it's 20uA. The chip itself consumes only 35na according to its datasheet:
https://www.ti.com/lit/ds/symlink/tpl5110.pdf?ts=1652658923819&ref_url=https%253A%252F%252Fwww.google.com%252F
Maybe their p-channel mosfet has high leakage current with the circuit has they've designed it? I would pass if it really is 20uA. I'm sure I could do better than that.

You need something like either a Current Ranger or a MicroCurrent Gold to accurately measure microamps and nanoamps. Cost is around $100.

Also, I just now looked at one of Kevin Darrah's videos:
https://www.youtube.com/watch?v=WNH6tyQpwF4

where he has embraced ESP-NOW. He said his esp8266 still takes around 1.5 seconds to wake up and transmit, even though running ESP-NOW, which is a very long time, as the module will be burning current through out that wake-up interval. In the example he showed, the delay was quite noticeable (though definitely an improvement compared to before when he was using plain wi-fi). Of course, the devil is in the details, but seeing that I'm more skeptical now than before.

[Larson]

@NeverDie I recognized the KD image as soon as I saw it. I'll watch... again... after dinner. Yes, the battery economics depend on time and power as you say. One of the tricks I have learned is aggregating data at the MCU level in RTC memory. I'm using both a 328 and an ESP; each dedicated to it's task. When 20 samples (minutes of data) are gathered the MCU wakes the ESP so it can bulk-load data to Thingspeak. This economizes the cost of the ESP wake-up by a factor of 20. This is all for my water-meter monitoring project. It has been fun stuff and I've learned a bunch. And I'm not using ESP-NOW because I'm going directly from the ESP transmitter to my home network - so I could do even better if I injected an ESP repeater that would transmit to the network. I think my wake-up/network-connection is about 8 seconds.

We have strayed far away from your initial "anyone-useing..." post. I hope that is okay?