Cables, Connectors (N/SMA/BNC) & Adapters with Measured Results for Comparisons

[rf-messkopf]

That's crazy. How does that work? Do I use the cable for the through, and then just run S21, or do I need to use one of those fancy features I haven't properly tried yet?

Yes, if your VNA supports that, you can use the cable as a thru, and then simply run a S21 measurement. That's what I did in the measurements I posted above -- the DUT is the thru. R&S calls that UOSM, but this method is also known as unknown thru or SOLR (short-open-load-reciprocal). It relies on the assumption that the 2-port device used as a thru is reciprocal, which is true for a cable. It was first published in 1992 in an IEEE Microwave and Guided Wave Letters volume by A. Ferrero and U. Pisani.

The method is based on the 8-term error model for VNA measurements (using error two boxes at each port, each described by ABCD-parameters), and this error model can be converted to the usual 12-term error model by using additional measured quantities, the so-called switch terms. These are the ratios of the incoming and outgoing wave amplitudes at one port while the other port is used as a source. This requires the VNA hardware to have four receivers (in a two-port VNA).

Sounds magical, but it works. At first sight, the math behind it is a bit confusing though.

[tautech]

Mine lives with member hendorog whom is away for a few days but I'll point him here for some comments to give you a better understanding of this VNA's capabilities.
Check out his Cal kits:
https://vnalab.net/

I've looked at his site before a few times, but I've never seen N stuff there. I know you prefer SMA, but I want N and SMA cal kits. I'd like a Siglent F504TS to magically appear on my bench.

Request a N type of the gender you prefer and he'll price something and give you an ETA.

hendorog likes to characterize every Cal kit to see what he's getting and know buyers will be satisfied for their spend.
BTW he discovered where F504TS is made and now buys direct to undercut what we can provide them for......I don't care and prefer he handles this RF stuff as it is pretty specialised and I'm not about to shell out for a 20+ GHz VNA like he has. 
I do know he'd prefer a SNA6134A but neither of our pockets are that deep ! 

[KungFuJosh]

That's crazy. How does that work? Do I use the cable for the through, and then just run S21, or do I need to use one of those fancy features I haven't properly tried yet?

Yes, if your VNA supports that, you can use the cable as a thru, and then simply run a S21 measurement. That's what I did in the measurements I posted above -- the DUT is the thru. R&S calls that UOSM, but this method is also known as unknown thru or SOLR (short-open-load-reciprocal). It relies on the assumption that the 2-port device used as a thru is reciprocal, which is true for a cable. It was first published in 1992 in an IEEE Microwave and Guided Wave Letters volume by A. Ferrero and U. Pisani.

The method is based on the 8-term error model for VNA measurements (using error two boxes at each port, each described by ABCD-parameters), and this error model can be converted to the usual 12-term error model by using additional measured quantities, the so-called switch terms. These are the ratios of the incoming and outgoing wave amplitudes at one port while the other port is used as a source. This requires the VNA hardware to have four receivers (in a two-port VNA).

Sounds magical, but it works. At first sight, the math behind it is a bit confusing though.

Thank you! That was very helpful. It looks like the advanced cal options make it so I can do SOLR and then change the R later, or do SOL and add R later too. Noice.

How do the attached charts look now?

Thanks,
Josh

[KungFuJosh]

[BTW he discovered where F504TS is made and now buys direct to undercut what we can provide them for...

Oooh, now that's interesting. I'm gonna have to email or message him. Thanks, Rob!

[rf-messkopf]

How do the attached charts look now?

Looks absolutely great. Even though there's a bit more trace noise than with the R&S. But I guess that's the difference between a € 6.000 and a € 60.000 VNA.

[KungFuJosh]

I wouldn't mind having an R&S FPL to play with also. Both are way out of my league though.

[KungFuJosh]

Looks absolutely great. Even though there's a bit more trace noise than with the R&S. But I guess that's the difference between a € 6.000 and a € 60.000 VNA.

What are your other settings, like RBW set to? Averaging? How many points are you testing at? Do you have any smoothing applied? I have mine set to 20001 points with no smoothing.

[tautech]

How do the attached charts look now?

Looks absolutely great. Even though there's a bit more trace noise than with the R&S. But I guess that's the difference between a € 6.000 and a € 60.000 VNA.

Recent discussions with hendorog have been about this.....
Tests of what the HW can do for him with HPAK, CM, SNA5004A and his X-R VNA's have been very interesting and spurred him to add a smoothing feature (not averaging) in the VNA SW package he's been developing.

Apparently single pixel traces are on the way ! 

It's a totally private development that I get to test a little as it can report if there is a new version and autoupdate on command.
It's in a fairly advanced stage now and primarily collects the raw data from a VNA over LAN or GPIB and crunches it on a PC to offer more capability than what a VNA can provide.

A wee introduction and screenshot is here:
https://www.eevblog.com/forum/testgear/siglent-sva1015x-1-5ghz-spectrum-vector-network-analyzer-(coming)/msg6243401/#msg6243401

[rf-messkopf]

What are your other settings, like RBW set to? Averaging? How many points are you testing at? Do you have any smoothing applied? I have mine set to 20001 points with no smoothing.

From memory, I think it was set to 200 Hz RBW and 601 points, 0 dBm source power, with no averaging or smoothing filters applied. But the ZVB generally has an extremely low trace noise, and a dynamic range well above the published specs.

[KungFuJosh]

From memory, I think it was set to 200 Hz RBW and 601 points, 0 dBm source power, with no averaging or smoothing filters applied. But the ZVB generally has an extremely low trace noise, and a dynamic range well above the published specs.

Here's two for comparison. One with 20001 points, 30kHz RBW, and one with 601 points, 200Hz RBW. Both with a touch of smoothing applied. I'd say that's good enough for me.

Thanks,
Josh

[rf-messkopf]

Here's two for comparison. One with 20001 points, 30kHz RBW, and one with 601 points, 200Hz RBW. Both with a touch of smoothing applied. I'd say that's good enough for me.

Looks very good. Also for comparison: Both channels set to 601 points. Channel 1 is at 100 kHz IFBW (which yields 54.09 ms sweep time), channel 2 is at 100 Hz IFBW (which yields 12.06 s sweep time). No averaging or smoothing is applied. Both channels are calibrated with the unknown thru method, the DUT is again the Sucoflex 104 cable.

You can see a bit of trace noise with 100 kHz IFBW if you zoom in, but at 5 dB/div that would be invisible. Also you can see a small discrepancy between the two S21 traces. That might be because the two calibrations of channel 1 and 2 were done with separate connections of the cal standards. At 0.08 dB/div things like temperature and connector torque start to matter (I didn't use a torque wrench and only tightened them hand-tight.).

[rf-messkopf]

That might be because the two calibrations of channel 1 and 2 were done with separate connections of the cal standards. At 0.08 dB/div things like temperature and connector torque start to matter (I didn't use a torque wrench and only tightened them hand-tight.).

Yep, if you carry the calibration of channel 2 over to channel 1, the two S21 traces agree much better, so it very likely is connector repeatability when connecting the cal standards. But you do get some ripple from transplanting the calibration.

[KungFuJosh]

Thanks! Your posts are informative and helpful for learning this stuff.

Why do you select 601 points in this range?

Thanks,
Josh

[mawyatt]

Josh,

You need to get real cables from Gore!!

https://www.gore.com/products/vna-microwave-rf-assemblies?xcmp=hdr-all_aero_ppc_aad_google_NA%7C%7Cgore-cables-branded

Best

[KungFuJosh]

You need to get real cables from Gore!!

Quit it!

I have real cables from Messi & Pauloni, and I won't be spending anything higher than that. I've already spent more on cables from them than my previous two VNAs cost combined.

The cable I'm testing is a shitty free cable, but that's on purpose. I do have some more stuff coming from M&P also though, including a nice N to N cable they're making for me. I'll test that when I get it. I'll also make a couple N to N cables myself for comparison with solder and crimp versions. The nicer cables I have are N to BNC.

Thanks,
Josh

[rf-messkopf]

Why do you select 601 points in this range?

When you only want to look at, say, a plot of the S21 magnitude and things don't vary rapidly with frequency, using more frequency points doesn't help much but increases sweep time. After all, your screen resolution is limited. This may be different when things like the phase changes rapidly with frequency and you want to plot that, or look at the Smith chart, and the curve becomes angled. But that's not the case with a piece of cable, which should have a flat group delay.

You may also need a lot more points when you want to do some specific post-processing of the S-parameter data, like converting to the time domain. Then the number of points is directly related to the maximum ambiguity-free electrical length of your setup.

[KungFuJosh]

This is a different short N to N cable. What do you think about this one? How do you think this cable compares to the other one?

Thanks,
Josh

[rf-messkopf]

This is a different short N to N cable. What do you think about this one? How do you think this cable compares to the other one?

If the results are for real I'd call that cable defective. Around 3 GHz you have a S21 ripple with almost 2 dB amplitude (pp). That could be explained by two impedance discontinuities along the cable (e.g., at the ends) of about -9.4 dB return loss if you do the math, assuming they are both equal. With two such discontinuities at -10 dB return loss each, the reflections would add up to about -4 dB worst case (constructive interference), which corresponds roughly to the peaks of the S11 ripple near 3 GHz. So what you show is consistent with two discontinuities of about -10 dB return loss near 3 GHz.

You could check if the frequency Δf between two ripple peaks corresponds to the electrical length L of the cable, from L = kc/2Δf, where k is the velocity factor of the cable. With k = 0.7 and ΔF = 280 MHz (the 2 in the denominator accounts for the round trip of the reflected wave in the cable) this yields L = 0.38 m. Is that your cable length? That would mean the discontinuities are at the ends.

So either your calibration is off, or you have some serious defects along the cable, e.g., incorrectly mounted connectors. A good cable should have a |S11| below -25 dB and only minimal S21 ripple.

[KungFuJosh]

If the results are for real I'd call that cable defective. Around 3 GHz you have a S21 ripple with almost 2 dB amplitude (pp). That could be explained by two impedance discontinuities along the cable (e.g., at the ends) of about -9.4 dB return loss if you do the math, assuming they are both equal. With two such discontinuities at -10 dB return loss each, the reflections would add up to about -4 dB worst case (constructive interference), which corresponds roughly to the peaks of the S11 ripple near 3 GHz. So what you show is consistent with two discontinuities of about -10 dB return loss near 3 GHz.

You could check if the frequency Δf between two ripple peaks corresponds to the electrical length L of the cable, from L = kc/2Δf, where k is the velocity factor of the cable. With k = 0.7 and ΔF = 280 MHz (the 2 in the denominator accounts for the round trip of the reflected wave in the cable) this yields L = 0.38 m. Is that your cable length? That would mean the discontinuities are at the ends.

So either your calibration is off, or you have some serious defects along the cable, e.g., incorrectly mounted connectors. A good cable should have a |S11| below -25 dB and only minimal S21 ripple.

That's what I thought (crappy cable). It was a short one I made last year, but I definitely screwed up the connectors. Yes, your length estimate is correct as well.

Attached is another cable I made, but much longer, and also a little better. I redid one of the connectors yesterday to change the gender from F to M, and I assume that also improved the quality of the connection since I'm better at those now than I was a year ago.

Should I redo the other end, or do you think this looks good? Keep in mind it's Hyperflex 5, which they don't recommend above 2.4GHz.

Thanks,
Josh

[KungFuJosh]

This is that same "better" cable. I did a full SOLR cal before each of the screenshots, switching the direction of the cable for R.

I swapped the direction to see if this is a valid test of the connector quality at either end of the cable.

Am I correct that the S11 readings on "Old to new" are less desirable than those on the "New to old" screenshots?

If so, I assume I should redo the Old connector to improve the cable. Is that correct?

Thanks,
Josh

[Rubo]

Yes, the "New to Old" is slightly better, but not by much.

[KungFuJosh]

Yes, the "New to Old" is slightly better, but not by much.

Thanks, that's what I thought. Is it enough to bother redoing? Or do you think they should both be improved? Or leave them alone?

Thanks,
Josh

[Rubo]

Looking at their datasheet, they don't list any specs regarding return loss above 2GHz. And return loss between 1.2GHz & 2GHz is ">22db", whatever that means 
I personally wouldn't use it for anything above 1-1.2GHz to do any critical measurements. I've played with M&P coax in amateur radio setting, but would not call it "test equipment grade" cable.

[KungFuJosh]

They have different grades of cable, but their datasheets are a little weird. I think they might be at the top end of what I'm willing to spend on cables though. It's also worth noting that nothing I'm doing requires anything better than $20 generic cables.

I'm sure the $5000 Gore cables Mike suggested are a little better.

What do you use?

[Rubo]

Cables Mike linked sure are nice, but I'm not shitting $$$ 

I scrounge eBay for good deals on similar (not quite as good) cables. Also, some of the equipment I have bought, came with realy nice cables.

Right now there are couple of cables available (no affiliation):

https://www.ebay.com/itm/227309779780

If I have to make my own and they are not going to be very long, I use RG400. Even chinesium stuff is relatively consistent in performance and you can use good connectors for rg58 on it.