DIY Antenna Analyzer

End Devices (Nodes) Hardware
44

Ok some more useful results. A couple of quick n nasty gnuplots

r3bazz.pdf (97.2 KB)
r3duck.pdf (155.1 KB)

Both are the results of sweeping an RFM95W from 818 MHz to 918 MHz in 5k steps, at a power setting of 2dB ie the lowest it will go.

Unfortunately I can not up load the csv files for the sweeps as the forum software does not like that my files are csv’s.

The plots are for one of the small rubber ducky type antennas that come with the 868 ESP32 Lora modules from aliexpress and from a vertical bazooka I made from a length of RG174 attached to an SMA connector.

The plots look plausible I will have to wait to compare them with a noise bridge and SDR spectrum analyser till all the parts have arrived though. The data sets are very noisy. Some extra decoupling, screening and a clip on ferrite or 2 on the USB lead should help a lot here. I can probably do some extra things in my code to filter each measurement too.

45

In summary then it appears to work despite the poor RF construction at the intended frequencies and on the lowest power setting of the RFM95W.

So the principle is useful and would bare having a bit more time put into cleaning up the signal.

Big thanks to LorTracker for the initial sig gen code, which sped up working out how to generate sweeps a lot.

46

5 khz steps is a bit ‘overdone’ imho and to meassure reflected power you need to transmit some power…
also SWR 1.033 @ 850 and 1.044 @920 … extremely good antenna :wink:

it is probably more usefull to do 1 Mhz steps but then meassure ten times and take the average on the same freq.

47

The LoRa device as a transmitter is putting out ‘power’. Does it make any differance when measuring antenna reflections etc as to whether the power is 2dBm or 17dBm ?

5 khz steps is a bit ‘overdone’ imho

If your refering to the 5khz steps in the scanner program I wrote, then I would agree that 5khz steps is a bit overdone. However, if you try to increase the steps beyond this I found that the output breaks up. There is a comment to this effect in the program.

48

well, measuring only the antenna reflection is not the total picture.
the cable plays an important role to, so yes I’ve tested cable+antenna with at least 6 db because I noticed that lower generator power gave wrong results.

sorry I didn’t see your scanner program (but I will)
I test a F range from 100-200 mhz in steps of 1 mhz normally, that’s fine enough (and fast :wink:)

I never meassured this (did you maybe ?) but the RF output of that rf95w when set to a certain level… how consisted is that level over the whole range ?
so you set it to 2 db but , is it really 2 db on 820 and 920 ?

49

Good question, no idea I don’t really have the kit to test this.

I think Vs could be back calculated and with a 50 Ohm terminator in place on the test port (Is can be calculated from Vs and Rs) there is the possibility of getting something useful answer wise out of it. Without any reactive components Vs is pretty much constrained by Vcc.

TBH 100 MHz is a wide range. I have done a sweep with a 50 Ohm SMA terminator on the test port. I did expect it to be flat-ish but there was a noticeable gradient over the full sweep range. Not too much, about an Ohm or two of calculated impedance, but enough to see the response is not completely flat.

Given that the construction, lead lengths and parasitic impedances have not been attended to I can not claim that the gradient was due solely to the terminator. Gut feel says it is as much about the construction as anything else.

The SWR should be ignored at the moment. It is a bodged calculation and not a full calculation. It is indicative at best. More work on the math needs doing. The Nuts & Volts article is missing a bunch of important equations and my math is not good enough.

I am quite gobsmacked that it appeared to work at all. But there you go. Nothing ventured nothing gained. I was leaning rather heavily on it being sufficiently resistive so as to damp out a bunch of reactive components. UHF heading towards microwave can be a bit piccy.

First cut is definitely all about testing the idea in principle. If it was no where near there would be no point in attempting to refine the design. As it is there looks to be some benefit to a redesign based on the same principles and maybe a cheaper compact microcontroller, say a 32u4.

I used a teensy as I had one available, but if the measurements were offloaded as is. there is no reason why the match could not be done on the host. Would prefer to be able to show are sult on a cheap LCD attached locally though, or even use an ESP8266/32 and offload the results via wifi to say a mobile phone, post sweep.

On the sweep resolution, yes 5KHz is overkill but it shows up the noise rather nicely. Having too much information makes re-binning possible. If it is not there in the first instance then re-binning is not sensibly doable. Re-binning is, or can be, in effect a low pass filter. I did also go with LoRaTrackers initial advice/code. Life is too short and I am all for making the most out of it through expediency.

50

Good question again. I think that more power should increase the measured voltages and make them proportionately larger than the noise floor. Increasing SNR is usually a good thing. The down side si the amount of interference we would be pumping out of the connected antenna system whilst we were sweeping. I also need to turn off the RF when no sweep is taking place. The end calculated results (as they are the result of factors and relative measures) should not vary appreciably unless the circuitry is adding to much noise/variability.

51

With the noise floor at -110dBm or so, do you think you really need to use signals at +10dBm, which 120dB above the noise floor ?

The fancy antenna gizmo (N1201SA) that many think is such a good device uses an RF power output of -18dBm. So 2dBm output from the LoRa device with a 20dB in line attenuator ought to be the equivalent.

52

@BoRRoZ the cable plays an important role to, so yes I’ve tested cable+antenna with at least 6 db because I noticed that lower generator power gave wrong results.

Stop and think about that for a moment. With your setup you got different results at different powers, I would suspect some issue with the measuring equipment.

The antenna and its cables are passive, they should react the same (in proportion) to a power level of 10dBm as for a power level of -10dBm. If it were not this way you would have a situation where things such as antenna gain, VWSR and cable loss would vary depending on the power level in use ?

I test a F range from 100-200 mhz in steps of 1 mhz normally, that’s fine enough (and fast

The fancy antenna gizmo (N12101SA) uses up to 1khz steps.

I never meassured this (did you maybe ?) but the RF output of that rf95w when set to a certain level… how consisted is that level over the whole range ?

When I measured it with my RF Explorer, the output was surprisingly flat, although as is the case with most types of vector type analyser there needs to be a calibration step so that a ‘flat’ output is achieved for the purposes of the test.

1 Like
53

if I have the time I will setup something to meassure the real output of the RFM95 will my old, but calibrated equipment, which I sometimes used to meassure cable/connector losses and which you can’t do with the rf explorer off course.

x857

.

54

I keep a 433Mhz and 868Mhz LoRa module here that I went to the trouble of calibrating their power output using a spectrum analyzer with a proven calibration record.

I have these LoRa modules specifically so that I can use them as a reference for checking devices such as SDRs, LoRa receivers or indeed my RF Explorer are displaying accurate power levels.

Since I know my 868Mhz LoRa module puts out a measured 15.9dBm carrier when set to 17dBm, checking attenuators or cable losses with an RF Explorer (or SDR) is easy enough.

55

Hi!
Why not build a graph from 800mhz to 1000mhz with the RSSI value between two lora module to check the antenna?

Something like this but for a range of two frequency:

56

The output of the LoRa module will of course vary across such a large frequency range, so you would be measuring the response of its output filter as well as the response of the antenna.

It would be intersting to see how well you get it to work.

57

I did do a write up of the simple antenna testing described in the Video, it uses a LoRa device as a simple RSSI meter;

Testing antennas is easy

58

The idea of testing an antenna with the actual Lora radio unit that will be used in the actual installation is an interesting idea. It means you are evaluating the whole system not just the antenna in a test environment. I have been looking at replacing the code in one of my nodes with an RF logging system.

59

Indeed so.

And there is a (small) benefit in tuning antennas for each individual module.

In the report I linked to above you will find a graph of how the optimum antenna length varied across 3 different modules and that the optimum length in all cases was less than the theoretical calculated length.

1 Like
60

Yes I think there is a lot to be said for real world practical tests, both in terms of using the actual tx/rx gear (or similar) that will be used and checking actual ERP and profiles.

The little RG45 & microbore copper tube based bazookas I have made up for portable Lorawan (868MHz) work very well indeed. I have pinged packets off gateways in Barnsley from the M18, and similarly to the far side of Sheffield from the junction of the M1 and M18. All of that with the pinger sat on the dash of the car whilst driving along. They are flexible enough to go in a back pack or pocket without getting unduly damaged.

The little plastic antennas though that come from aliexpress have been terrible.

My ad hoc tester works well enough for my needs, that it works at all is as much a surprise to me as it was to other posters here. Part of why I think it works is simply down to the fact that all the lead lengths are less than a 1/4 wave and the loading is for the greater part resistive, due to the bridge, even allowing for the unknown whatever on the test port.

If I were making a tester like this one again, I think I would use surface mount components to get the lead lengths even shorter and screen it in a can to stop noise pickup, between the Lora module and the diode samplers. Feed through caps would have been nice but I don’t have any. It was a junk box special mostly.

I think using one of those nice ESP32 TTGO units with an oled screen and being able to show the results, rather than post processing the visualisation on my laptop would be nice as well. Makes it all a bit more interactive.

1 Like
61

On theoretically calculated length.

The theory uses the speed of light through a vacuum, whereas the speed of light through whatever the antenna is made up of and parked next to will vary a lot from this.

So tuning to the kit and location will always be necessary, if you want to wring that last little bit of performance out of it. IMHO.

1 Like
62

I can recommend the report is really worth reading and spending the time to comprehend.

I must admit i keep coming back looking to build a slim light-weight co-linear that fits inside a watertight gateway enclosure that doesn’t need a ground plane as the radials penetrate and compromise the enclosure integrity.