@bluejedi is very busy writing a summary topicstart for part 2 
1 Like
No expert on soldering smd components either. I will use my magnifying glass, multimeter and small tipped iron.
Youre wright. Forgot to place the brackets there.
Send a message to my seller and got a very fast reply. They only have a generic schematic but said one can be made if i want to. Strange answer but ok asked for one and will see what shows up.
I think some (many?) chinese pcb sellers just copy & paste application reference designs found in developer kits of chip suppliers.
The schematic from “LilyGo” (see above) seems to be very close to that in this Semtech Document
The question then is, if those sellers hardly do any kind of engineering and quality assurance…?
I checked surface temperatures of a “Heltec” labelled board (white pcb) with infrared thermometer after ~20 hours of continous operation in wifi sniffing mode. Hotspot is surprisingly not the ESP32 itself (~35 °C.), but the flash RAM chip which has ~39 °C.
The Heltec ESP32 WiFi LoRa I have here has circa 3.5dBm worse receiver performance than a more EMI quite ATMega 328P @ 8Mhz powered receiver.
Tested at 434.4Mhz, bandwidth 125000, spreading factor 8, code rate 4:5.
Details of how the test was done here;
The IPEX to SMA cable on the Heltec is probably a major factor in the 3.5dBm difference.
Yes he is but a ranking between boards will not be part of it. 
1 Like
Possibly, but not a lot you can do about it.
What loss would you expect in that length of the IPX cable ?
Can answer that;
http://www.rfconnector.com/series-data.php#ipx
So around 0.5dB loss in the 15cm cable.
1 Like
I’m not sure how to interpret the numbers on that rfconnectors page for calculating the total attenuation.
It says nominal attenuation for 1.13 mm coax is 3.73 dB/m. Several other references specify attenuation between 1.9 and 2.3 dB/m. But that is for the cable. What about the attenuation of the IPX/U.FL connecors?
The page specifies “Insulation Resistance: 500 m ohms min.” and “Contact Resistance: 20 m ohms max. (center contact) and 10 m ohms max. (outer contact)”. I assume “m” in “m ohms” stands for ‘milli’ for contact resistance and ‘mega’ for insulation resistance (but it is not clear).
Also of importance is the following:
“Female U.FL connectors are not designed with reconnection in mind, and they are only rated for a few reconnects (approximately 30 mating cycles [3]) before replacement is needed. The female U.FL connectors are generally not sold separately, but rather as part of a pigtail with a high-quality 1.32 mm doubly shielded cable, which allows for a low-loss connection.” Source: Wikipedia / Hirose U.FL
This implies that the quality of the connection will degrade with each connect/disconnect cycle (with substantial impact after 30 cycles).
I did some simple experiments lately and noticed a difference of about 3 dBm (which is an estimate). This is similar to your findings of 3.5 dBm.
(A minor difference with your tests is that I only used my nodes as ‘end-device’. So my results are based on the transmitter part of the nodes only.)
I did the tests to get a rough indication of how the Heltec LoRa antenna/signal compares.
My test setup (868 MHz):
A Heltec Wifi LoRa 32 node and a second node consisting of Arduino Pro Mini with a separate RFM95W module on a HopeRF RFM95 adapter. It has a SMA connector so the LoRa antenna can be connected directly to the module (similar to your setup).
I first tried the 5cm whip antenna that was included with the Heltec and then tried a 19,5cm (¼λ) 868 MHz whip antenna. I used the same antenna’s on both nodes. RSSI of messages coming from the Heltec were about 3dBm lower than from the other node. The difference between the 5cm antenna and the 19,5cm (¼λ) antenna was also in the order of several dBm (around 3 dBm) .
The gateway used is a RAK831 based multi-channel gateway. The TTN Gateway console (traffic tab) shows the RSSI value for incoming messages. This allowed me to compare the RSSI of both nodes.
The reported RSSI values vary several dBm per message. I averaged the values from multiple messages.
I tested only one distance between the nodes and the gateway (5m, which includes a concrete floor).
IPX connectors woulf not be my choice at all.
In the case of the test I did, there had been 4 or possibly 5 ‘connect/disconnect’ cycles,
But it will be difficult to hack an SMA connector directly onto the Heltec Wifi LoRa 32.
Thanks for testing and sharing your results with us here!
Question now is: how can the heltec board be improved?
So, further investigation would be to check were the 3.5 dBm get lost:
- is it EMI on the board, and if yes, from which chip is it emitted (ESP32, CP2102 or Flash-RAM)?
- on the passive RF path between SX1276 and U.FL connector?
- on the pigtail cable?
The improvements I would suggest are very simple, extended the board be a few mm, and you could then have the option of fitting an edge SMA.
Also put the battery connector on the PCB top, if you fit the Heltec into a project you cannot then access the battery. Better still would be to put the battery line on one of the pins, which you would then have space for if you extended for a proper SMA socket.
Ok - but what exactly causes the 3 dBm loss in RF receive path? Sure that this is as simple as changing the U.FL?
I doubt the u.fl is the problem.
I did use a different test method to compares a bare ESP32 module versus an ATMega using the exact same LoRa module for both tests, the results are here;
http://www.loratracker.uk/?p=611
Around 3dB in that case too, I suggest the loss of sensitivity is due to EMI from the faster processors.
Interesting.
You say you used a bare ESP32 versus ATmega using the exact same LoRa module for both tests.
Is your bare ESP32 a shielded (like ESP-WROOM-32) or unshielded version?
What LoRa module/board did you use? How is it connected and positioned in respect to the MCUs?
In case of using a separate LoRa module I would expect it to be relatively easy to shield the LoRa module from EMI (RF) caused by the MCU boards by using metal shielding.
Testing with and without the shielding should reveal the effect of EMI per MCU board.
Another possibility for lower RSSI of the Heltec Wifi LoRa 32 could be sub-optimal LoRa RF circuitry design (including PCB lanes and lack of proper shielding). Looking at the relatively poor Wifi RF performance of the Heltec (poor antenna/placement and an unconnected leftover Wifi antenna on the bottom) that would not come as a surprise.
The antenna on the TTGO (both V1 and V2) appears to be (an imitation of) a better Wifi antenna, but placing it on the bottom side is sub-optimal which will result in similar Wifi performance.
The TTGO V2 appears to have improved LoRa circuitry because it now uses a separate LoRa module which also includes metal shielding.