0.1-200MHz antenna analyzer

[xaxaxa]

I am in the process of designing my next product which is a 0.1 to 200MHz one port VNA/antenna analyzer. It is estimated to be sold for around 30 USD.

The overall architecture is similar to the xaVNA, but with a si5351 as the synthesizer and one receiver switched between the reflection channel and reference channel. The ADCs+FPGA is replaced with a STM32 with built in ADC. I have a working prototype and the accuracy looks reasonable, not noticeably different from the xaVNA.

The big design decision I am facing now is the infamous "whether to put a display on it" dilemma. There is already a USB interface to the PC and the existing xaVNA software works with it, but since the main market for this lower frequency antenna analyzer is ham radio, it would be far more convenient if it can be used on the field (since antennas are big at VHF and below).

The downsides of adding a display:
* Enclosure construction will be far more complicated since I can't use off the shelf aluminum enclosure anymore
* Increased chance of shipping damage
* UI takes a lot of development effort and I am already very busy with many other projects

All the items above add cost and I estimate the total cost will probably double. Offering two versions (one with a display and one without) does not solve the problem because they are mostly one-time costs.

What are your thoughts? Maybe I can offer it without a display but with a footprint for a standard ILI9341 display, and users can do the UI programming themselves?

[markus_jlrb]

Please take a look at EU1KY AA 300-kHz to 270MHz at 1. harmonic of the SI5351
and up to 800 MHz by using the 3. and 5. harmonic.

http://www.wkiefer.de/x28/EU1KY_AA.htm

Portable construction (stand alone without the need of a pc) should be a must for an AA.

A power level of 0dBm  should be a must too (the above AA have only -20dBm
what is a disadvantage measuring big antennas.

Markus

[Theboel]

Hi Xaxaxa,
The cost for ILI 9341 3,2 inch display with touchscreen about 12USD but how to program it thats totally different thing.
Anton

[profdc9]

You should check out the VNA I designed.  Sounds pretty similar...

http://www.github.com/profdc9/VNA

Dan

[xaxaxa]

You should check out the VNA I designed.  Sounds pretty similar...

http://www.github.com/profdc9/VNA

Dan

Yes I know about this one. I might use your UI code 
I found a few other open source VNA projects also using a STM32 and have a menu UI, but yours is the only one using a F103 which I'm also using.

[Yansi]

Originally, I decided not to comment in this thread, but I couldn't resist. Please, do not take this as offense, just a thought:

There is already a metric sh!t ton of "cheap", "simple", "stupid", or other not-that-usable tools.

Why not trying to make a real instrument, while keeping the price low? I mean - harmonic free test signal, proper signal filtering, settable amplitude, proper PCB design (no kits slapped together)...

The cost will not be as low, but if one wants any better than a cheap unusable toy, there is nothing to buy. There are just toys costing couple tens of bucks, then toys costing hundred bucks or more (because you know, it is a brand name), then LOOOOOOOOOOOONG way nothing and the next step thousand(s) dollars for a proper VNA.

Look for example, this is what I mean by at least a proper try at achieving something useful:
http://hforsten.com/cheap-homemade-30-mhz-6-ghz-vector-network-analyzer.html
http://hforsten.com/improved-homemade-vna.html

(Thank you Henrik for sharing with us!)

[xaxaxa]

Why not trying to make a real instrument, while keeping the price low? I mean - harmonic free test signal, proper signal filtering, settable amplitude, proper PCB design (no kits slapped together)...

Already exists. The original xaVNA has all the things you mentioned (harmonic conditioning, adjustable amplitude, proper DSP, etc) and is a ready-to-use solution. I have already done comparisons with the ~$500 level instruments and the S11 measurement is more accurate than those.

However, from sales data I can tell I have not tapped into the ham/hobbyist market at all, and market analysis heavily suggests pricing (and frequency range) is the reason. I think there is a big market for something at the $30 level, ready-to-use, and gets the job done without unnecessary bells and whistles. Harmonics in the test signal is not a problem for antenna and passive network measurements, and filtering is nonviable over 3 decades of bandwidth. A DDS solves that but at great cost.

Since you mentioned the hforsten VNA, I can reveal that I'm also working on productionizing the full two port xaVNA which has accuracy and dynamic range specs exactly on par with the hforsten one (other than frequency range) but is ready-to-use out of the box. It was offered to kickstarter backers at around $500.

[xaxaxa]

I will add that this is going to have proper DSP and real receivers, not yet another VNA-on-a-chip based design. The S11 accuracy will be on par with real instruments.

[richnormand]

Hi xaxaxa,

I have your original xaVNA and I like it a lot. 

This one would indeed overlap the frequency range (0.1 to 200MHz and 137MHz to 3.5GHz)  and be most welcome. There are a few times I would have liked to go lower.
Here are some comments as you requested:

0)
using the same box/form factor would make it easy to tape the two units together. Perhaps some software support to integrate the running of the two simultaneously?
1)
display: I already use my laptop for the original. So the same layout on the computer would be fine with me.
2)
not having a display also would make it easy to integrate the two machines and update the programs as you are doing for the original on GitHub.
3)
if someone is desperate for a display for field use, and does not want to use a laptop or tablet, you might consider an app that runs on a Raspberry Pi.
There are many small touch screens that provide all the functions you might want that are readily available for the Pi. This includes boxes and such too.

My 2cents worth.

PS: Hope you will go forward with it. You might consider sending a note  to your past supporters to let them know too.

[Yansi]

if someone is desperate for a display for field use, and does not want to use a laptop or tablet, you might consider an app that runs on a Raspberry Pi.
There are many small touch screens that provide all the functions you might want that are readily available for the Pi. This includes boxes and such too.

I cryyyyy just whyyyyyyy.    Like if nothing else existed, than crapduino and that whatever fruit 3.14.

Okay xaxaxa, I see you can properly engineer. Then I'd step it a notch further. Make it real dual port VNA (not this 1.5 port one), up the frequency range at least so it can go lower, make it with better DR and properly characterize it, not in just 5 sentences summary. I wouldn't bother making more cheap junk. There are plenty of ways of making money, if that is what bothers you.

PS: Use proper USB instrument type B connector too. MicroUSB is the dumbest connector ever invented. 

[ogden]

Antenna analyzer shall be fully portable with battery & display, thou most likely such is not compatible with target 30$ price

Portable STM32-based VNA project with very nice UI:
https://www.eevblog.com/forum/rf-microwave/portable-2-port-1-mhz-to-6-ghz-vna/

Project page:
http://www.yl3akb.lv/content/vna_v2/vna_v2.php

The overall architecture is similar to the xaVNA, but with a si5351 as the synthesizer and one receiver switched between the reflection channel and reference channel.

You may use square wave for LO of receiver, but definitely not for transmitter of antenna analyzer which shall be considered as swept frequency *transmitter*. Unless you use pure sine wave - don't even consider that build.

[richnormand]

Fair point ogden and Yansi as it was called an antenna analyzer,

I guess, at the below $100 price point, given his resources and timeline I would much better appreciate the resources to go towards a solid preforming device that is easy to update in the future, as is his xaVNA unit.
So, for me, battery operation and a self-contained unit is not my first priority as most of the stuff is done in the shack. There are lots of other portable tuners around. I just pointed out there would be easy ways to make it portable should you want to.

[Yansi]

Antenna analyzer shall be fully portable with battery & display, thou most likely such is not compatible with target 30$ price

Portable STM32-based VNA project with very nice UI:
https://www.eevblog.com/forum/rf-microwave/portable-2-port-1-mhz-to-6-ghz-vna/

Project page:
http://www.yl3akb.lv/content/vna_v2/vna_v2.php

The overall architecture is similar to the xaVNA, but with a si5351 as the synthesizer and one receiver switched between the reflection channel and reference channel.

You may use square wave for LO of receiver, but definitely not for transmitter of antenna analyzer which shall be considered as swept frequency *transmitter*. Unless you use pure sine wave - don't even consider that build.

That is indeed impressive one!

Regarding the TX signal purity - if one would not want to use a DDS (as a 200MHz DDS may cost a small fortune), it could be solved likely the "old way" by mixing two high frequency oscillators, and then all gets low pass filtered. Simple enough, output quite clean.

I have long ago thought about using a couple of seriously cheap ADF4350 chips (from china) to make a 0-1GHz frequency synthesizer from them. But this idea was long forgotten since I bought a proper 3GHz Agilent genny for my home lab.

[profdc9]

You are welcome to use the touchscreen code or any other code you find in my VNA project.  It is all zlib licensed, which is like MIT licensed.  I am also working on an automatic antenna tuner for those interested:

http://github.com/profdc9/ModularTuner

The si5351 generates a square wave with harmonic content up to 1 GHz and beyond.  This is very useful as the harmonics of the fundamental may be used to measure reflection/transmission at the third harmonic to get up to 450 MHz, and above, though the SA612 VHF mixer I use has poor efficiency above 500 MHz.  The harmonic signals generate harmonic intermediate frequencies.    This is what the EU1KY analyzer does, and my VNA does this as well.  So square waves are not at all a problem as long as you use a band-pass filter to filter only the harmonic of the IF needed.

It was my intention as well to make a very low cost VNA.   I wanted to make it a relatively easy to build kit for new hams as an antenna analyzer and for more advanced work, hence having both through-hole (which is good in practice up to 200 MHz) and SMD versions.  But the parts are very low cost and widely available, and it can be used with USB or a touchscreen.

The biggest disadvantage of the VNA currently is that its dynamic range is low, about 50 dB reflect and 60 dB transmit.   But even this limited range can be quite useful for many purposes.  Minimizing the direct coupling between transmitter and receiver is very difficult when both are on the same PCB using the same oscillator chip.   It would probably be better if there were two SI5351A, one generating RF and the other LO, because the RF generating chip needs to only be at the transmit port, and then a mixer could be used to get the IF.  But I think those who have a fancy laboratory instrument forget how useful it can be to have even limited capability, which is why xaxaxa's work is so important.

[xaxaxa]

I have attached the schematics of the current prototype.

Measuring a coax stub:



I used ad8342 mixers instead of sa612 because receiver linearity was found to be the most important factor in S11 error. The baseband detection is also done coherently in software and there is a reference channel. The USB power rail is stepped up then regulated down to get rid of LF noise. The main design objective was to come up with the most accurate 1 port VNA within the tight budget because the vast majority of hobbyist (including me when i was starting out) can not afford more than $50 for an antenna analyzer.

I have looked into mixing two synthesizers to create a sinewave signal, but it seems like a very small value add for a big cost. As mentioned earlier test signal harmonics make no difference when measuring passive/linear networks provided the baseband detection is narrowband and does not respond to harmonics. Are there other reasons to prefer a sine wave output?

I think for the trial run I'll just include a ILI9341 touchscreen footprint and allow the user buy the display from aliexpress, while the software side should work out of the box (i.e. just solder the display on and be greeted with the UI). Battery is a problem though because of logistics issues, but I don't think USB powered would be too problematic because USB power banks are so ubiquitous these days.

I plan to also include an e-cal port on the same board.

[xaxaxa]

Make it real dual port VNA (not this 1.5 port one), up the frequency range at least so it can go lower, make it with better DR and properly characterize it, not in just 5 sentences summary. I wouldn't bother making more cheap junk. There are plenty of ways of making money, if that is what bothers you.

PS: Use proper USB instrument type B connector too. MicroUSB is the dumbest connector ever invented. 

I'm also doing that (the full two port VNA), and actually starting from the next batch all xaVNA will be able to go down to 35MHz because I replaced adf4350 with adf4351. I can better characterize it once I can afford higher end test equipment 

I originally settled on micro USB because it's the most ubiquitous cable type and I didn't want to include a cable (to be environmentally responsible). But I will note your suggestion and do some research on how common type B cables are.

I started this business not only to make money. There is a reason I don't accept outside investment and will not have any shareholders other than myself. When I started out in this hobby I realized everything was beyond reach (because test equipment is way overpriced). Thanks to manufacturers like Rigol and countless anonymous aliexpress engineers (look at e.g. the $20 LCR meter thread) the hobby is a different place now.

[rahandford]

Hi

Very interesting project, I was also looking at something like this.
How about installing/providing a small footprint Bluetooth module, and some type of (Android?) UI that can use the display of a phone?

[profdc9]

You might want to put a ferrite bead on VDDA, and make sure analog ground and digital ground are tied as close together as possible under the microcontroller AND ONLY UNDER THE MICROCONTROLLER.  It can be tricky to get 12 actual analog bits out of an ADC on the same chip as a bunch of digital noise.    The Bluepill doesn't do this, because it's cheap, but if you're starting from a clean slate, pay attention to this.

Similarly, you may want to add ferrite beads to the power pins of the AD8342.

Also, one of the problems is that if you don't sample precisely at a multiple of your IF, you will get phase error because the microcontroller clock is different than the SI5351A clock, and so you can not integrate the IF signal for a long time to accumulate signal.  Perhaps try to derive a clock signal from the SI5351A that is harmonically related to the IF you can use with a timer to trigger the ADC acquisition.  For example, say your IF is 50 kHz.  You have RF at 100 MHz, LO at 100 + 50 kHz, and then perhaps have the SI5351A output a third clock at some multiple of 50 kHz such as 1 MHz (CLK1 is unused right now).  Then you put this to an input counter-timer pin of the STM32F103 (e.g. PA0 or PA15 for T2C1E) and then have it count a certain number of pulses (for example 5 pulses), which is the multiple of the IF, and then fire an interrupt each time the counter rolls over.  Then you can for example sample at 200 kHz which is 4 times the IF.

[ogden]

As mentioned earlier test signal harmonics make no difference when measuring passive/linear networks provided the baseband detection is narrowband and does not respond to harmonics. Are there other reasons to prefer a sine wave output?

Regulatory reasons. Antenna analyzer together with antenna under test is RF transmitter unless one uses anechoic chamber. Every antenna tuned for specific frequency will perform quite well at it's harmonics.

According to ETSI EN 301 783-1 radio amateur equipment shall not transmit out of band emissions above -57dBm. Other RF devices have strict out of band radiation limits as well. As soon as one user will get RF regulator such as FCC knocking on his doors, your product may get loads of bad press in a no time Putting square wave into antenna in 0..200MHz band is very, very bad idea that could end very badly in many countries. That freq range includes aviation and military bands BTW.

Regarding the TX signal purity - if one would not want to use a DDS (as a 200MHz DDS may cost a small fortune), it could be solved likely the "old way" by mixing two high frequency oscillators, and then all gets low pass filtered. Simple enough, output quite clean.

Right. BTW there are low cost RF synthesizers in form of RF transceiver ICs. They are not pure "continuous PLL" synths because VCO is tuned only at frequency setup and will drift away somehow, but for antenna tuner application it's ok. One to mention - SI4460, ~2.2$. Obviously those need low pass filter as well - because output is square wave. 850–1050 MHz band of SI4460 seems good this project.

Also, one of the problems is that if you don't sample precisely at a multiple of your IF, you will get phase error because the microcontroller clock is different than the SI5351A clock

It is no brainer to clock ADC (thus microcontroller) and RF part from the same clock. It is how it shall be done in RF freq/phase-sensitive applications. Also I would like to discourage STM32F103 use - it is old & venerable. Newer chips such as STM32F303 have better ADC and other peripherals such as PLL with lower phase noise. Actually due to inherent phase noise of PLL, I would underclock STM32 and run it straight from 24MHz or 32MHz external clock w/o internal frequency multiplier.

[xaxaxa]

It is no brainer to clock ADC (thus microcontroller) and RF part from the same clock. It is how it shall be done in RF freq/phase-sensitive applications. Also I would like to discourage STM32F103 use - it is old & venerable. Newer chips such as STM32F303 have better ADC and other peripherals such as PLL with lower phase noise. Actually due to inherent phase noise of PLL, I would underclock STM32 and run it straight from 24MHz or 32MHz external clock w/o internal frequency multiplier.

Not so easy when the si5351 is constantly being updated and will have breaks in the clock output. The STM32 seems to have no way to clock the ADC independently of the system clock, but I'll try the timer trigger suggestion above. I worked around this issue by switching between reference and reflected path often so that only relative phase is sampled and phase error can not accumulate. Either way this isn't the main source of noise, since as you can see from the above graphs the noise on S11 is frequency dependent and only becomes visible at >100MHz. I'll put a bit more power supply filtering on the TCXO, clock gen, and mixer.

I'm not so sure about the harmonic emission thing, since when you connect a VNA to an antenna you will have unwanted emission either way. I think the best that can be done is to keep output power low (-10dBm). So far the hforsten VNA is the only one I have seen that has any harmonic conditioning and it isn't perfect (there is just a bank of 4 lowpass filters, each with rejection of 20dB). -57dBm is a spec that even the $3000 grade instruments are not likely to meet.

[ogden]

It is no brainer to clock ADC (thus microcontroller) and RF part from the same clock.

Not so easy when the si5351 is constantly being updated and will have breaks in the clock output. The STM32 seems to have no way to clock the ADC independently of the system clock.

Easy. Clock both from same clock source, 25MHz crystal oscillator. I think 24MHz will work for si5351 as well. If you need USB at the same time while CPU runs at incompatible with USB peripheral frequency,  then you may use STM32 with USB PLL like STM32L151.

when you connect a VNA to an antenna you will have unwanted emission either way.

You will have in-band emission. When HAM is testing/tuning his HF antenna and sets VNA sweep to 7.0-7.2 MHz HAM band, other HAM's may not like if he does it for long time, but he do not radiate out of band interference into FM or for god's sake, AIR band. I do not care - you change your mind, ideas or beliefs or not. Just saying.

[edit] Never mind. Seems like HAM's are not concerned about regulations For example FA-VA5 uses square wave, most likely si5351 or similar used.

https://youtu.be/K4iy_az6pBQ?t=1325

This product: https://www.sdr-kits.net/VA5-Antenna-Analyzer-Kit

[Kalvin]

Apparently a new, portable antenna analyzer 60 kHz - 35 MHz (230 MHz) will be soon available:

https://qrznow.com/rigexpert-stick-60khz-35mhz-analyzer/

I have no further info on price or performance etc.