Why do I need a VNA rather than a SA with directional coupler to tune antenna?

[PA4TIM]

Network analysis is very complex. Not something that can be explained in a few posts. My tutorials are just the basics and written for the beginner in this field. If you want to know more, buy the book about this subject from R&S. Be prepared for an awful lot of very complex math. Keysight has a free impedance handbook as pdf. That is more practical.

Buying a VNA is easy, making measurements too, making good measurements is harder, reading the data correct even more. 

You need:
- a VNA

- if not build intern you need directional couplers and/or bridges depending the frequency. They are not universal usable from DC to daylight. I have a bridge that goes upto 100MHz, a bridge upto 500MHz and a coupler that goes from 2GHz to 4GHz. 

- a calkit, do not under estimate that, they are expensive upto extreme expensive and only useful if they come with data. Dave Kirby is a good source for N and Rosenberger has very affordable sma kits. See my tutorial about calibration.

- For some older analog VNA's you need things like powersplitters and adjustable extenders (telescopic rigid coax tot match phaselength of both paths). A vna has two receivers and one transmitter. Both receivers must get exact the same signal magnitude and phase or know the difference. That is the way they measure. They measure the difference between both receivers. And that is why you calibrate.

-decent cables and connectors, not the chinacrap. I got some new and calibrated N cables from a friend. The difference in measurements above 1 GHz is significant (my VNWA goes to 1300MHz but this cables and a good calkit make it usable to 1500). Those cables are 260 dollars a piece !!! You do not need them but just to let you know what is used in the VNA world. Calkits over 10k dollar are not extreme.

-fixtures for things that do not have a connector. And if your VNA does not do port extention then you need a special calkit you can connect instead of the DUT. But you can make those and calibrate them with your good cal kit (I often use a working kit that I check against my good kit because SMA connectors wear and I do not have a moment wrench for them so they wear maybe more)

I do not know how the normalisation of the modern cheap SA's are but my old ones only do the amplitude for S21. A VNA calibrates at the point you connect the DUT, so on the output of the coupler. My SA's can not calibrate that.

A SA gives you the impedance, not the phase, 50 ohm without phase does not tell you if there is resonance (so the reactance/imaginairy part/jX  is zero)
For adjusting a commercial 2 meter antenna no problem, but for building one from scratch at the frequency you mentioned and then also embedded I do not think a SA will be the right tool. It can be usable if you are lucky. There is a reason VNAs are so expensive 

[tautech]

I do not know how the normalisation of the modern cheap SA's are but my old ones only do the amplitude for S21. A VNA calibrates at the point you connect the DUT, so on the output of the coupler. My SA's can not calibrate that.

Normalise on the Siglent SA accounts for coupler and cabling right to the DUT.

See the linked pdf that explains this process.
https://mediacdn.eu/m/media/downloads/Siglent-RBSSA3X20-Reflection-Bridge-Manual.pdf

Thanks for your reply.

[G0HZU]

The 20dB spec for directivity isn't that great for the Siglent bridge. Is it better than this in reality? At the other extreme you could consider something like the Anritsu/Wiltron 6N50 return loss bridge if you can find a used one at a decent price. This offers about 40dB directivity up to about 2GHz and you could make fairly good scalar measurements with this and the Siglent analyser perhaps?

I think this is the external RLB that was originally shipped with my old Advantest TR4172 spectrum analyser. This special spectrum analyser had a factory option fitted that meant it could be used with the tracking gen and the external bridge and it could measure impedance like a VNA. But it was very fiddly to set up and calibrate. However, I wouldn't recommend using the TR4172 as a VNA, it was much better as a spectrum analyser. Sadly, the 6N50 bridge was long lost before I bought the analyser. But it still has the impedance option fitted internally.

[tautech]

The 20dB spec for directivity isn't that great for the Siglent bridge. Is it better than this in reality?

The info in the link is all that we have for it AFAIK, whereas it would be better to have a proper datasheet or better still a proper characterization chart like is provided with the Siglent SA cables.
I haven't got the Siglent bridge in stock to check one and probably don't have the skills and knowhow to do so.

I'd like to have shown the tiny SMA coupler performance hendorog gave me in a screenshot but alas I was missing an adapter. Tempted to use a snippet of soft copper wire to get connection between the SMA females. 

[trevwhite]

Tautech, what is the worst that can happen?

[hendorog]

The 20dB spec for directivity isn't that great for the Siglent bridge. Is it better than this in reality?

The info in the link is all that we have for it AFAIK, whereas it would be better to have a proper datasheet or better still a proper characterization chart like is provided with the Siglent SA cables.
I haven't got the Siglent bridge in stock to check one and probably don't have the skills and knowhow to do so.

I'd like to have shown the tiny SMA coupler performance hendorog gave me in a screenshot but alas I was missing an adapter. Tempted to use a snippet of soft copper wire to get connection between the SMA females. 

That typical measurement quoted has to be at 2Ghz - i.e. a worst case. I'll bet its much better than that at the lower end.

Sounds like you have RP-SMA adapters? Damn annoying, just bridge it with a small bit of wire if you haven't got anything else. That little Narda coupler is a 4-8 GHz one, so it will be out of range, and didn't have great directivity in the first place so don't expect too much

[vk6zgo]

I am wanting to do a bit of work tuning antenna. Be that PCB, helical or other. I have an SA with TG and a directional coupler on the way but someone told me I really need a VNA for this kind of work. Anyone explain why to me or point me in the direction of some useful article(s) that explain why?

Thanks

Trev

You don't!

Until VNAs became relatively inexpensive, Spectrum Analysers with a Tracking Generator were widely used to tune antennas, as were RF sweepers used with a detector.
Hell, Hams often tune  antennas with just an SWR meter!
Experience teaches various "work arounds" to get as close as possible.

The advantage of a VNA is that your don't need work arounds---the thing tells you the whole story!

In the past,such esoteric devices as vector impedance bridges were used if it was necessary to look at it the phase angle of the impedance.( a long drawn out experience, as a lot of spot frequencies needed checking.)

If EEs & Techs could have had a VNA that was relatively cheap & a usable size back in the day,they would have jumped for joy.

[tautech]

Tautech, what is the worst that can happen?

Well I hummed and harred when I bought mine not knowing anything about SA's and asked here on the forum....I think I've pointed you to those posts, right ?

In the end I decided that my early usage cases weren't likely to challenge datasheet spec or the 50v DC input protection so I just shelled out for the Siglent UKitSSA3X accessory kit......but I can buy these at cost. 
http://www.siglentamerica.com/prodcut-fjxx.aspx?fjid=1558&id=5113&tid=227&T=2
Most of what I do is looking/identifying frequencies of interest and much can be done with just a coax sniffer loop where there's little chance of exceeding input spec. Even with the strongest of signals from a coax loop the internal attenuation available can easily suppress it.
I've needed little more in the way of accessories to date, only adding some 6" SMA patch leads and straight SMA connectors for hendorog's little directional coupler.

Direct connection need be thought of exposing you to two risks: overvoltage and/or signal strength.
Internal attenuation can handle most needs and the overvoltage risk can easily be managed if we use that thing on top of our shoulders.....we should anyway.
High power/voltage is where things get tricky and you must shell out some $ to get the job done. Period.
Go looking for attenuators on eBay and you'll see there's all sizes and shapes that can handle/do what you need.

Of course none of us want to be the first to bust our SA  but to start with simple stuff and work forward as you gain knowledge is the path I've chosen. Understanding dB and what that means to the input ratings is important.

When in doubt, dial in a heap of attenuation and if the signal's not there, start reducing it until you see what you need. And keep reading whatever you can find.....btw, in the SSA3kX thread, rf-loop has posted many links to great SA documents....grab them all. 

[PA4TIM]

Be careful if you need to measure big antennas, like a ham dipole for 80 meters or 25 meters endfed wire etc, there can be a lot of signal on there and static can cause very high voltage. (like kV's)

[PA4TIM]

Just found some time to test normalisation om my Marconi  2380/2383 with factory standard TG. Goes from 100Hz to 4,2GHz  http://www.pa4tim.nl/?p=5587
The TG output is already flat without normalisation but normalisation is the cherry on the pie 

My HP coupler is a -30dB version made for 1,9GHz to 4GHz. It is a 4 port version so it can measure forward too. A thing to keep in mind, this is not the directivity but the attenuation so you can measure on higher power DUT without cooking your SA frontend. So be careful if you buy a DC. Be sure it is not fried and has a high attenuation. My Marconi was a high end SA and has a high enough dynamic range to still be usable with a 30 dB attenuation and the TG at -10dBm but I have seen modern Chinesiums that probably have a problem with that. If the normalisation corrects for dips of 20dB or more and the TG is -10dBm and your bridge takes another 30dB I do not think it will be very usable.
I had a Tek 2710 and that TG was far from flat by itself (10's of dBs)

My Marconi asks to insert a fixture/devise before normalization
If I normalize at the DUT port of the HP DC with a short and with an open. The calkit, cables etc I used  is all GR and have the same ref-plane. There was no real  difference between open and short. That is a good test, they must be the same. So if you cal with nothing on the DC the refplane is at that point. If you use a homemade short that shorts at the and off the barrel that reflplane is much further to the DUT. If I used very bad opens/shorts the formalization was affected a lot. So a good calkit is also here important. In the siglent case an open but testing after normalization with a short and load is a good test. all 3 of the 50 ohm loads I used gave a return loss over 20 dB at the worst frequency. The best return loss peaks where over 40dB. At that point the TG signal is -10dBm, the DC attenuates the respons another 30dB, the RL was 40dB so the RX sees a -80dBm signal. If the noise floor of your SA is at -60 your RL dissapears in the noise floor.

You have no correction options for the cal plane with SA's so try to do normalisation at the DUT. Refplane changes are "phase things". Problem there is that the phase transforming will mess up things.

If you adjust your antenna to a |Z| of 50 ohms you want resonance too. In that case the antenna shows like a 50 ohm real resistance to the generator. There is a 180 degrees phasejump and the phase is 0 degrees at that frequency. At that point RL is better as 25dB.  But a |Z| of 50 ohm does not mean there is always resonance. Often it is, and with a commercial antenna the magnitude alone will be usable.
Return loss is usable because it will never be around 20-25 dB if there is no resonance. But things become complicated if you are measuring high frequency stuff and embedded. You need to be far enough from the antenna so you are no part off it. You need to be at least out the reactive field but better also out the near field. For high frequency stuff more easy.

But in that case you need a cable and then you measure the impedance at the begin of the cable and that varies with length as long as the antenna is not 50 ohm and resonant. So try different lengths of coax. If the results on the SA changes the antenna itselve is not 50 ohm and resonant. A cable transforms impedance but never to 50 ohm if it is 50 ohm. If the RL is better as 20-25dB you should not see a change if you take a longer piece of coax.

A VNA will show you the phase and modern VNA's have advanced forms of error correction, port extention/de-embedding, calibration methods. You need that to measure things like on pcb antennas or other stuff without a coax connector.

I wonder what they normalise in SA+TG and VNA's , the output of the generator so it will give the same power over the whole sweep, or just do some math to compensate both receiver and TG aberrations.  In old analog beasts they only can flatten the generator but with digital stuff a lot can be done in software but in that case the DUT sees a changing amplitude. For antennas not a big deal but if you measure amplifiers a bad thing. On the other hand, modern DSS generators can be made flat by software control very easy

About antennas, the analyser (vna or SA) does not only see's the signal from its own source but also other signals and on bigger antennas (like a big shortwave HAM  used dipole or longwire) that can give problems with falls readings.

[trevwhite]

This is a fantastic thread, thanks for all the replies. Still re-reading and trying to understand half of it.

[hendorog]

I wonder what they normalise in SA+TG and VNA's , the output of the generator so it will give the same power over the whole sweep, or just do some math to compensate both receiver and TG aberrations.  In old analog beasts they only can flatten the generator but with digital stuff a lot can be done in software but in that case the DUT sees a changing amplitude. For antennas not a big deal but if you measure amplifiers a bad thing. On the other hand, modern DSS generators can be made flat by software control very easy

The SNA normalise function on my Signalhound is equivalent to the Response Cal on my 8753 VNA - which is itself the same as just using the Display function to save the trace and then display the measurement as a ratio to the saved trace. These are all done in maths without changing the source power across the sweep.
Only one calibration standard is used.

If you do that, it doesn't matter what you have connected to the ports (directional coupler or short), or what you are measuring (S11 or S21), the normalise function will give you a flat line at 0dB.

Like you said though, the older gear is better at providing a flat source to start with, but as long as the DUT is kept in it's linear range it won't matter will it?