What VNA to get?

[G0HZU]

Hi G0HZU,
how can I contact you and hear some tips.

Not sure what you mean but the best thing to do is start a thread on the forum. Virtually everything I say on RF forums gets dismissed by at least one or more 'experts' despite me spending all my career doing RF work. So you are probably asking the wrong person anyway  

[kg4arn]

@ G0HZU
I, for one, follow your posts and appreciate your insights.
Thank you

[KE5FX]

They aren't as popular these days because of the grainy and gloomy VGA display

Those Sony color displays sucked when they were brand new.   It's hard to exaggerate how awesome an 8753C/D/E model looks with the SimmConn 1024x768 LCD.  It is worth replacing a perfectly-good CRT with one.

[hendorog]

Hi G0HZU,
how can I contact you and hear some tips.

Not sure what you mean but the best thing to do is start a thread on the forum. Virtually everything I say on RF forums gets dismissed by at least one or more 'experts' despite me spending all my career doing RF work. So you are probably asking the wrong person anyway

FWIW, you are one of the guys that I take notice of when you post, and I value your comments highly. You generally post something I didn't know before or didn't think of. If it is something I did know already or worked out for myself then I feel quite chuffed

I think it is a bit on the nose for people to be asking for personal help by PM. The whole point of this forum is to post openly.

[Omicron]

One thing to watch out for with the ENA E507x series of VNAs from Agilent is that they generally don't have bias tee ports at the back of the instrument.

I'm assuming that if you don't have internal bias tees and use external ones you run into similar issues with S11 and S22 measurements then when you use external attenuators? Would it be as big a deal breaker not to have them when selecting a VNA as the source attenuation? It's interesting to know because all these things tend to be options you have to pay extra for so there is some kind of price/usefulness consideration. Do the internal bias tees even make sense on low end devices like the PicoVNA and the TTR506A or are they more of a gimmick considering the performance limitations of the VNA itself?

[G0HZU]

First, thanks for the kind comments from kg4arn and hendorog.


Getting back to VNAs and bias tees I can only offer my own experience of what I do and what I see at work...

I'm assuming that if you don't have internal bias tees and use external ones you run into similar issues with S11 and S22 measurements then when you use external attenuators? Would it be as big a deal breaker not to have them when selecting a VNA as the source attenuation? It's interesting to know because all these things tend to be options you have to pay extra for so there is some kind of price/usefulness consideration. Do the internal bias tees even make sense on low end devices like the PicoVNA and the TTR506A or are they more of a gimmick considering the performance limitations of the VNA itself?

At work, built in bias tees don't get used very often unless there is a doubt about a manufacturer's S2p data or the data doesn't cover enough frequency range. Then they can be useful if a transistor or MMIC needs to be measured. They can also be useful when testing something as simple as a biased diode. However, it is something that all RF engineers should try and do now and again just for the experience. I've also used external bias tees many times and calibrated them out. Other times it's possible to be sloppy and just bias the device directly inside the test fixture. I see this a lot at work and it's often OK to do this especially if the same bias network is being used in the real design anyway and the aim is just to get an s2p model of the circuit rather than the active device alone.

I have usually had very good results with (fairly basic) external bias tees up to a few GHz. I think it's more important to have the source power control than internal bias tees but I guess it depends on what type of device gets tested most often and over what frequency range. External bias tees are usually going to misbehave up at many GHz especially if the cable and bias tee are moved during calibration etc. So this will degrade the results.

I don't have any experience of using the USB VNAs like the Tek or the Pico so I don't know how well they perform when measuring active parts. The specs for the Tek don't look that good to me and the Pico VNA doesn't have much range on the source power control. They are both probably fine for most investigative tasks, however and it would be interesting to see how well they can measure s2p data of active devices using the bias tees. I expect them both to struggle with critical measurements but I don't think they are designed for stuff like that. It's all about compromises in terms of size, power, weight and cost as I'm sure we will all agree.

[Omicron]

Well these are low cost devices after all so they are certainly not meant for critical measurements in an RF lab. If there is an interest, I have access to both a PicoVNA and a R&S ZNLE and could perform some measurements to evaluate their performance. I've only used these for antenna measurements though and transistor level RF amplifiers are a bit outside my comfort zone so I'd need someone to volunteer a DUT and some guidance as how to best set things up.

[G0HZU]

Edit: I think we double posted but I had a go at making a quick and dirty amplifier that would have similar properties as the Crystek MMIC amp (back in post #36) and I've listed the results below.

The aim was to simulate and predict the performance using the manufacturer's S2p data for my MMIC (in my case I used and old SBB-5089Z) and a simulation of the PCB layout using Genesys and Sonnet. Then compare it (using a VNA) to a measurement of the real circuit built on a PCB. It's possible to design the circuit and PCB in Genesys and then export it all to Sonnet to simulate the PCB layout. I find this setup works really well.

The Genesys plot below shows the simulation data for S21 and S12 and also the imported data of the real PCB with a real 5089Z MMIC on it. i.e. the VNA was used to compare/confirm the simulation.

There's a plot showing the input VSWR for both the real PCB and the simulation and you can see they agree quite well. The two green traces almost overlay perfectly However, I did cheat a bit because my simulation only goes to 5GHz and not 6GHz. At 6GHz my bias choke has a slight resonance and it spoils the comparison. I don't have an accurate model of the bias choke at 6GHz so I cheated and cropped the data at 5GHz.

I've also included an image of the real PCB when it was being tested on the VNA. This is a quick and dirty test PCB milled on some 0.031" FR4 material.

The results agree very closely and this is partly due to the accuracy of the manufacturer's s2p data for the 5089Z MMIC.

[Omicron]

That kind of CAD software is well out of my reach, I see why you would want a VNA with matching capabilities

It does provide a good baseline for comparison though. Would you be willing to send me the test board? I could measure it on the low end devices and send you the touchstone files. We could then see how they fare.

[Antonio84e]

consider MegiQ for affordable and easy to use VNAs: https://www.megiq.com/

[G0HZU]

That kind of CAD software is well out of my reach, I see why you would want a VNA with matching capabilities

It does provide a good baseline for comparison though. Would you be willing to send me the test board? I could measure it on the low end devices and send you the touchstone files. We could then see how they fare.

Sadly, the board is a bit of a scruffy mess and it has some expensive connectors on it so I don't want to part with it. Plus I want to try and scavenge a decent wideband choke at work next week and some lower loss caps to get it to work with a smooth response up to 6GHz.

The circuit is really simple and the results should be predictable. If you look at the datasheet for the SBB-5089Z there is an application circuit and PCB layout. My results are very similar to the datasheet. Your S11 plot for your Crystek amplifier looks a bit ripply. Maybe try doing just a 1 port cal directly at the VNA port 1 and try measuring S11 without a cable? Simply terminate the output of the amp in a decent SMA 50R load. I'd expect the S11 result to be a bit like my MMIC board in that the response curve and the dip should be smooth (rather than ripply) but then I can only guess at what is inside the Crystek module.

 

[borjam]

We had a 40GHz HP VNA that ran WinNT on a really old embedded PC. While you can connect it to the network. It was basically a huge honey trap of unpatched vulns. We could have upgraded it to WinXP, but the involved replacing the motherboard and the front panel.

So, connect it to a firewall? For about 100 EUR there are mini PCs that have 2 or more Ethernet ports, and any semi-competent network administrator can setup something like that.

Of course. Now think that it's not the only untouchable sacred cow Managing that can be a nightmare.

[Omicron]

Your S11 plot for your Crystek amplifier looks a bit ripply. Maybe try doing just a 1 port cal directly at the VNA port 1 and try measuring S11 without a cable? Simply terminate the output of the amp in a decent SMA 50R load.

Yes, the ripples in my previous measurement are caused by my carelessness in doing the measurement and possibly because I don't have really good cables. I re-measured as you suggested using only an adapter to connect the amplifier. I measured without the 20dB attenuator first and stored that to memory (the blue trace). This trace is quite similar to your results. Then I repeated with the 20dB attenuator in place (after recalibration obviously). This results in the yellow trace. So you can definitely see the difference. Some waviness is added that I couldn't get rid of with calibration. When I measure the distance between the ripples this is about 2.1GHz. This, I believe corresponds to roughly 5...6 cm which is indeed in the ball park for the length of the adapter + attenuator I used.

In full disclosure, I used a ZNL for these measurements as that's what I have access to. The hardware is identical to the ZNLE though. The service manual for these is available from R&S (you do have to register). Looking at it both the ZNL and the ZNLE use the same RF board, exact same part number. In fact there doesn't even appear any provision for the optional source and receiver attenuators that you can get for the ZNL. I'm guessing these are already on the board. Curiously they can't be activated by just a software key, you need to send the device back to R&S for re-calibration. I think that might also be the only difference between the ZNL and the ZNLE: the way they calibrate the instrument (slightly better tolerances on the ZNL?).

[Wolfgang]

Hi,

I have an Anritsu MS2036A portable and a Keysight E5071C 2-Port 8.5GHz. The Anritsu is small, but absolutely no fun at the bench. 

The E5071C is an old-fashioned boatanchor, but
- It works without a PC and has a large display
- Its more accurate and has better dynamic range that a lot of other VNAs
- You can measure frequency translating devices, too (paramps, mixers, ...)
- SCPI controllability is very good
- Its not buggy like newer stuff.
- You can write extensions to the installed software, because its Windows based. Visual Basic is old-fashioned, but powerful.
- There is a lot of support on the web, docs and utility software.

The downside: you have to negotiate *a lot* to get a reasonable price for it. Something in the range of 40% discount should be possible. 

Some measurements with frequency offset mode:
https://electronicprojectsforfun.wordpress.com/parametric-rf-amplifiers/a-parametric-upconverter-for-the-40m-band/

[clockspectrum]

The desktop units can be large, but I sure love mine! Ive used a E5071C at a time where all I had was an older 8753ES. It was a large improvement in usability compared to the older stuff. I think its a great analyzer for the money. Ive ended up upgrading to a N5221A and love it. I will say that you should make sure you invest in a good calibration set and possibly cables depending on what kind of measurements you are making, otherwise you will be pulling your hair out.

Matt

[Wolfgang]

I agree. Without a proper cal kit and reasonably phase stable cables all the possible accuracy is gone, especially above a GHz or two.
I use an N cal kit from Anritsu, an SMA cal kit from Dr. Kirkby, and finally a Keysight e-cal module (I got too lazy).
Good cables are Sucoflex 104 or the thick, big as a hotdog-sausage-type phase stable ones from Gore or Keysight. They are unwieldy but accurate.