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

[Rubo]

Nothing to add to what gf said already, it pretty much covers everything.

I don't have any SMA cal kits, I standardized on N connectors with all my stuff, otherwise I could have mailed you mine so you could check if what you have is good or not.

Thanks for all the info. I plan on getting a proper kit + adapters, probably from Roger. I do want to learn how to characterize stuff eventually.

Check out what MiniCircuits have in their catalog (not as a kit but as stand alone part numbers), then look around on ebay.
I have bought a few N loads, shorts, etc. from there. brand new, for a lot less than other places. And checked with calibrated VNA's they were all perfectly following the datasheet.

[gf]

If metrology-grade standards are not required, the Premium 12 GHz Calibration Kits from SDR-Kits offer a viable budget option.
While assembled using high-quality Rosenberger components, these kits cost only a fraction of official factory calibration kits.
https://www.sdr-kits.net/calibration-information-for-DG8SAQ-VNWA-3-3EC
Unfortunately, they are currently listed as unavailable/sold out on their website.

EDIT: Even the Amphenol kits from SDR-Kits are likely better than NanoVNA bundled kits. While NanoVNA standards are completely uncharacterized No-Name parts, the Amphenol kits use genuine industrial-grade components and come with documented model coefficients (delay and fringe capacitance). Mechanical repeatability is likely better, too.

[KungFuJosh]

Thanks! Those are some great prices. Looks like I'll get some kits from them too, eventually.

[KungFuJosh]

This one is @mawyatt's fault.

36" Gore Phaseflex N to N cable...from eBay. Of all the eBay cables, this one seems to be tougher in the sense that it maintains the results shown regardless of screwing around with the cable. I can push it up or down, left or right, slightly loosen the connector, and there's almost no difference if anything.

So, anyway... What do we think of this cable? Does it live up to the Gore name?

Thanks,
Josh

[Rubo]

This one is @mawyatt's fault.

Glad I didn't get blamed for that one 

Much better. On S11 the worst is -33/-32db which is still quite good, but there's also a lot less variation with frequency (going from best case to worst case).
And a lot less loss too, .464db for 3ft that comes out to .155db/ft @ 3GHz, I can live with that.

[KungFuJosh]

I retested the M&P Sahara 10 (S10, 3 meters) and Stealthflex 7 (SF7, 4 meters) cables for comparison with the Gore (36").

Am I correct that for a comparison between the 3, S11 looks best on the Gore cable, both in terms of level and shape?

S21 @ 3GHz:
Gore: -0.464dB
S10: -0.76dB
SF7: -1.541dB
Can we directly compare these, or not? If it's acceptable to multiply them all and pretend they're the same length, then wouldn't the S10 be the best for this measurement? Or is that now how we go about comparing S21 with different lengths?

Thanks,
Josh

[Rubo]

You are correct, S11 looks best on Gore cable.

You are also correct in assuming you can compare S21 between the cables.
Every manufacturers datasheet gives you attenuation per (however many) feet/meters for a particular frequency.

So @ 3GHz your Gore has -0.464db for 1M cable.
For 3M cable you just multipy -0.464 x 3 = -1.392db (which is more than S10 @ the same lenght).
Same for 4M: -0.464 x 4 = -1.856db (which in turn is more than SF7).

For VNA you don't need the lowest loss coax (within reason obviously), what you need is very good phase stability and a linear S11 (not sure how to phrase that one).

S11 of SF7 looks better than that of the S10, but I still don't like it 

[KungFuJosh]

Thanks, that's helpful.

S11 of SF7 looks better than that of the S10, but I still don't like it 

That's what I expected you to say. I will retry S10 again. It might be PEBKAC related. SF7 looked similar then I did something and it was flatter. 

In any case, getting good results with that Gore cable took no effort. I think that's partially because it's much shorter and easier to setup without worrying about the extended length and whether I want to coil it or try and lay it out some way.

I did also get a lower quality Gore SMA cable (not Phaseflex) but it was not good. Rolling the dice on feeBay can be interesting.

Thanks,
Josh

[KungFuJosh]

What's the deal with SRL? I want to try and setup the test the same way M&P probably did for the attached tested cables.

Thanks,
Josh

[Rubo]

In any case, getting good results with that Gore cable took no effort. I think that's partially because it's much shorter and easier to setup without worrying about the extended length and whether I want to coil it or try and lay it out some way.

That's the whole point of phase stable cables (which Gore & Junkosha are one of the few companies that make them), that they are very stable with flex & temperature.
So you don't have to "freeze" everything in time to perform measurements. But they cost arm & a d**k new 

[Rubo]

What's the deal with SRL? I want to try and setup the test the same way M&P probably did for the attached tested cables.

I honestly have no idea what kind of VNA they used, but it definitely doesn't have enough points in the sweep.
But even in those tests you can see the impedance being very unstable.

Looking at the S10 test, if they had used as many points in the sweep as you did, they might have a similar result.

[KungFuJosh]

Hmm, so it sounds like you're saying I should drop it down to 2 points, and then get a nice linear result? 🤔🤔🤣

[Rubo]

If you want to get results similar to what M&P have, yup. 

[KungFuJosh]

lol, I'm gonna do it just to try and match their test results.

[KungFuJosh]

I dunno if I did the setup for this correctly, but...at least it looks cool.

This is the Sahara 10. I should drop the points down to 51 to match M&P's lol.

Thanks,
Josh

ETA: Oh, I think this has to be wrong. It's a 3m cable.  Eh, I'll figure out the right way eventually. 

[Rubo]

Keep trying, you'll get it to look like M&P eventually 

[gf]

What's the deal with SRL?

Researching the literature a bit, it seems that SRL measures the signal reflections caused by internal manufacturing inhomogeneities inside a cable, calculated as the deviation from the cable's own fitted characteristic impedance as a function of frequency. It is not the S11 you measure. It neither includes the connectors, nor reflections from the impedance mismatch between the cable's characteristic impedance and 50 Ohm.

Typically, it's measured for whole reels (say 100m). If the DUT is too short, SRL may not 'see' periodic manufacturing inhomogeneities introduced by the machine. While some mechanical vibrations cause variations in the centimeter range, slower cycle times can also introduce periods of several meters.

IEC TR 61156-1-5 defines the standardized mathematical techniques to calculate SRL from S-parameter data. I don't think that the Siglent VNA can calucate it directly, so you would need to export measured S-parameters and calculate offline.

The best freely available description I found was chapter 4.3 in the following paper (here it is called Fitted Returns Loss):
https://www.aesa-cortaillod.com/wp-content/uploads/2024/03/Paper-2020-IWCS-Procedures-to-overcome-spurious-effects-in-meas-of-RF-com-cables.pdf

[KungFuJosh]

Typically, it's measured for whole reels (say 100m). If the DUT is too short, SRL may not 'see' periodic manufacturing inhomogeneities introduced by the machine. While some mechanical vibrations cause variations in the centimeter range, slower cycle times can also introduce periods of several meters.

Thanks. That's interesting, considering they charge extra to test the cable. 🤔

I asked them how they performed the test last year, and this was their response:

Hi Josh, the cable must be connected to the ports of a Vector Network Analyzer, the basic test measures the parameters S1-1 Impedance and Return loss.
Also, it ensures that the cable is working properly and the contacts with the connector are compliant (absence of standing waves).
On request, the parameters S1-2 (cable losses) can be measured only if this has a length > 5 meters for a 5 meter cable, 8-10 meters for cables
with a larger diameter.

From their answer, it sounds basic. Maybe something lost in translation?

Thanks,
Josh

[KungFuJosh]

Keep trying, you'll get it to look like M&P eventually 

I dropped it down to 51 points last night, and it's pretty ugly. ...still the wrong length though. I need to learn the correct way to use a cable as a DUT and do the SOLR cal specifically for TDR.

[gf]

I dropped it down to 51 points last night, and it's pretty ugly.

In order that you don't need to guess:

Lowpass TDR requires a harmonic frequency grid, i.e. f_start = step, and f_stop = N_points * step.
If non-harmonic frequency points are measured, the VNA would need to inter/extrapolate them for lowpass TDR, which is suboptimal.
[ DC unfortunately needs to extrapolated anyway as it can't be measured. ]

For good time resolution, use a large f_stop. The higher the frequency, the better the resolution: Δt = 1 / (2 × f_stop).

The frequency step (Δf) determines the maximum range of the time axis. A long cable requires a long time axis and therefore a small frequency step. For example, a 10 MHz step results in a time axis with a total duration of 1 / 10 MHz = 100 ns, which is typically displayed from -50 ns to +50 ns.

Keep in mind that the time axis resulting from the TDR (-> IFFT under the hood) is circular and wraps around. To avoid wrap-around (aliasing) of the data, make sure the time window is long enough for the DUT, and ideally even long enough to capture multiple reflections bouncing back and forth. If in doubt, it is always better to make the time axis longer than necessary by using a smaller frequency step.

Examples:

f_start = step = 10 MHz, N_points = 300, f_stop = 3 GHz (-> Δt = 167 ps, time axis = -50 ns ... +50 ns)
f_start = step = 1 MHz, N_points = 3000, f_stop = 3 GHz (-> Δt = 167 ps, time axis = -500 ns ... +500 ns)

[KungFuJosh]

Because TEAS is real, and I'm an idiot, I bought a couple more Gore cables. First, here's a comparison of the original 36" cable with the new one also.

[KungFuJosh]

Here's a 48 inch 3.5mm Gore cable using some crappy (free with nanoVNA) N to SMA adapters. I'm curious to see how this compares when I get some better adapters from @hendorog.

Thanks,
Josh

[Rubo]

Here's a 48 inch 3.5mm Gore cable using some crappy (free with nanoVNA) N to SMA adapters. I'm curious to see how this compares when I get some better adapters from @hendorog.

Thanks,
Josh

Even with crappy adapters the S11 still looks better than the M&P cables you were testing previously 

[KungFuJosh]

True, but they were attached during R calibration, so they shouldn't hurt it tooooo much, maybe?

It will be interesting to compare it with some good adapters that will be properly calibrated out before R though...and also with the same good adapters calibrated the same way I did with the crappy ones...whenever I get them from across the globe.

[KungFuJosh]

I got a 2nd Gore 3.5mm 48" cable in today. c1 is the original, and c2 is the new one.

Still with the crappy adapters and no torque wrench.

Playing with this crap has really made me look forward to getting better adapters and a torque wrench...and for that matter, a 3.5mm cal kit. Why is @hendorog on the other side of the planet??

I really want to see what these cables look like when I have that stuff. My cal kit only goes up to 4.5GHz, so I also want to see how they look up to 8.5GHz with the new kits.

Thanks,
Josh