Kirkby calibration kit alternatives?

[Noy]

Thx for all the clarification.
I think i will go with the Mouser Parts and will not get Male-Male thiny + box.
I don't know for what i need this Male-Male Adaptor. Everything i use (cables are Male - Male, PCB connectors everytime female..) so i don't know for what i can use it.. So i will go for single buy (especially i have to order some parts from mouser already..).

And if i need such a adapter (maybe i will calibrate without a cable at the VNA Port with the N->SMA Female Adapters (https://www.delock.de/produkt/89983/merkmale.html)) i can use a "cheap" chinese and calibrate them out or? Its the same like using a very short cable?

[Mechatrommer]

with a "pogo plug SMA"...

oh no, dont! ..
1) slight wiggling will change characteristic, and where the measurement plane will be? thats why they will ask you to buy another expensive torque wrench. if its thats easy everybody is already doing that and skip the torque wrench.
2) what do you want to mass produce cal kit for? not everybody buying it. you will compete with $4 kit anyway that people buy because they think its good enough.

your provided SDR links are not deep enough... i remember i saw a website at characterizing few SDR-kit, i cant find it right now... the closest thing you should read is something like this..
http://hamcom.dk/VNWA/
https://www.sdr-kits.net/downloads/2014-Rosenberger-Fairview-male-female-Cal-standards.zip
from https://www.sdr-kits.net/Female-12%20GHz-Kit

some sort of hand tuning cal kit, or to see how much batch to batch differ in characteristics. i dont read because i dont own one and afaik they dont include Ln and Cn effect into consideration. you may read yourself and decide if its good enough or what and then tell us something about it. cheers.

[switchabl]

imho no, except with extra effort to screw unscrew for connection and reduce usable life of the connectors. with good continuity/connectivity and good quality 50 ohm Zo thru, connecting to Open and Short will only increase its offset length, the rest of parameters are still the same. connecting to a good Load will still appear 50 ohm to the VNA. but well, this is true given the CAL set is of descent quality, if not, even a SMA cable or the sacrificial adapter connected to your VNA can look funny. this can quite visible beyond 3GHz and much lower with nonsense hunglow grade. ymmv.

Yes, in theory, an ideal adapter will add only delay. And to be fair a good 3.5mm air-line f-f will get reasonably close to that. Your average SMA adapter may not be that ideal. Minicircuits specifies a return loss of only >23dB @<8GHz for theirs. Now imagine connecting your nice 40-45dB calibration load on the other side and calibrating with that. Yes, that is a bit extreme, the adapter will usually be much better then the spec (especially at low frequencies), but you get the point. In particular, unless you have a cal-kit for the other side, you will not be able to check how good it actually is. Sadly so, because it would be much simpler and more economical if we could just have one cal-kit with a bunch of adapters.

Now to be clear, for none-critical measurements you can get away with a lot. On a good lab-grade VNA, you can skip the cal-kit entirely, attach your cables and adapters and run the automatic port extension if you don't care about a few dB. But if you are worried about the 35dB residual mismatch you might hope for with the cheap Rosenberger kit, you should probably worry about adapters.

[rubidium]

Has anyone tested out the Applied EM Innovations kit?

[小太]

radar_macgyver mentioned they bought an Applied EM kit while having access to a professional 3.5mm calibration kit... I wonder how the comparison went given that it's almost 2 years later?

For reference, current prices with conversion rate £1=US$1.263

Item	Kirkby £GBP	Kirkby $USD	Applied EM $USD
6 GHz			599
7 GHz	485	578
8 GHz	513	648	649
Torque wrench	150	189	(included)

[radar_macgyver]

I have the 10 GHz version of the Applied EM cal kit. I made some measurements on a FieldFox 9917B that was first calibrated with a Rosenberger RPC-3.50. The Fieldfox was set to full span (30 kHz - 18 GHz) with 1601 points. I then measured the female standards from the Applied EM kit and saved the s1p files. Also grabbed an s1p from the RPC-3.50.

I plotted the impedance, and it seems like the AEM kit does well up to ~10 GHz.

I quit after tearing my hair out for a good 2 hours trying to figure out the METAS tools, so if someone else can use it to generate a model I'd be grateful (and maybe post a link to a tutorial?)

[alan.bain]

Are you trying to use VNA Tool?  I had previously had issues with this and the process I got was

Right-click on the file of measurements, then select Database / Fit Calibration Standard Model (as per the screenshot)

You then choose which type of standard it is and which parameters you don't want to adjust (I would suggest that initially you reduce the set size e.g. remove L2/L3 from the fit for a short) and hit "Start Optimization".   If you try and fit too many it can get into "Not Responding" state which is a bit poor!

[TheSteve]

I have the 10 GHz version of the Applied EM cal kit. I made some measurements on a FieldFox 9917B that was first calibrated with a Rosenberger RPC-3.50. The Fieldfox was set to full span (30 kHz - 18 GHz) with 1601 points. I then measured the female standards from the Applied EM kit and saved the s1p files. Also grabbed an s1p from the RPC-3.50.

I plotted the impedance, and it seems like the AEM kit does well up to ~10 GHz.

I quit after tearing my hair out for a good 2 hours trying to figure out the METAS tools, so if someone else can use it to generate a model I'd be grateful (and maybe post a link to a tutorial?)

Can you give the expected/specified Z0 offset and offset delay for the Applied EM kit. As was mentioned for it to work best you should know some of the parameters, or at least enter what they should ideally be to reduce the variables.

[小太]

I have the 10 GHz version of the Applied EM cal kit. I made some measurements on a FieldFox 9917B that was first calibrated with a Rosenberger RPC-3.50. The Fieldfox was set to full span (30 kHz - 18 GHz) with 1601 points. I then measured the female standards from the Applied EM kit and saved the s1p files. Also grabbed an s1p from the RPC-3.50.

I plotted the impedance, and it seems like the AEM kit does well up to ~10 GHz.

I quit after tearing my hair out for a good 2 hours trying to figure out the METAS tools, so if someone else can use it to generate a model I'd be grateful (and maybe post a link to a tutorial?)

Can you give the expected/specified Z0 offset and offset delay for the Applied EM kit. As was mentioned for it to work best you should know some of the parameters, or at least enter what they should ideally be to reduce the variables.

They posted values in this post as an attachment.
TL;DR for the female load is offset Z_0=54.16Ω and delay=4.77ps

I'll try to derive calibration parameters myself with the data too in a few days, if nobody else has done it by then 

[radar_macgyver]

Per the PDF included with the kit, Z0 offset is 54.16 ohm, delay is 4.77 ps, and loss is 10 Gohm/sec for the female load.
I ran the IN3OTD Octave scripts with the max frequency set to 10 GHz, and got Z0 offset = 54.5 ohm, delay 5.37 ps and loss of 20 Gohm/s. I suspect the optimization is not converging on a good solution since it hits various bounds (in the case of the run I included the results from, the upper bound for loss was 20 Gohm/s).

[小太]

Here are the fitted values by METAS VNA Tools for the entire 0~18GHz range.
The Applied EM kit is only specified up to 10GHz though, so this is just for educational purposes and not a commentary about the kit quality
(I'll do a proper 0~10GHz version in my next post)

Results have been plotted in Mathematica with equations from Keysight application note 1287-11 "Specifying Calibration Standards and Kits".
"Given" values are those included with the Applied EM kit (as reported here), while "Fitted" are those VNA Tools generated

Short F	Offset Z0 (Ω)	Offset delay (ps)	Offset loss (GΩ⋅s-1)	L0 (10-12 H)	L1 (10-24 H⋅Hz-1)	L2 (10-33 H⋅Hz-2)	L3 (10-42 H⋅Hz-3)
Given	51.56	83.61	2.57	-889.07	9989.25	9983.11	-1413.8
Fitted	51.524	81.139	6.2800	-909.96	111076	-14204	340.48
Open F	Offset Z0 (Ω)	Offset delay (ps)	Offset loss (GΩ⋅s-1)	C0 (10-15 F)	C1 (10-27 F⋅Hz-1)	C2 (10-36 F⋅Hz-2)	C3 (10-45 F⋅Hz-3)
Given	50	76.37	1.96	-351.35	-3947	1326.61	-507.08
Fitted	53.549	67.442	2.7871	46.597	-90050	10138	-358.16
Load F	Offset Z0 (Ω)	Offset delay (ps)	Offset loss (GΩ⋅s-1)
Given	54.16	4.77	10
Fitted	54.253	8.4764	-11.561

P.S., Equations from Keysight application note 1287-11 converted to Mathematica form

Code: [Select]

Zr = Quantity[50, "Ohms"];
(* Equations 1.1 and 1.4 *)
\[Gamma]l = \[Alpha]l + I \[Beta]l;
\[CapitalGamma]1 = (Zc - Zr)/(Zc + Zr);
\[CapitalGamma]T = (ZT - Zr)/(ZT + Zr);
\[CapitalGamma]i = (\[CapitalGamma]1 (1 - Exp[-2 \[Gamma]l] - \[CapitalGamma]1 \[CapitalGamma]T) + Exp[-2 \[Gamma]l] \[CapitalGamma]T)/(1 - \[CapitalGamma]1 (Exp[-2 \[Gamma]l] \[CapitalGamma]1 + \[CapitalGamma]T (1 - Exp[-2 \[Gamma]l])));
(* Equation 1.10 *)
\[Alpha]l = (offsetLoss offsetDelay/(2 offsetZ0)) Sqrt[f/Quantity["Gigahertz"]];
\[Beta]l = 2 \[Pi] f offsetDelay + \[Alpha]l;
Zc = offsetZ0 + (1 - I) (offsetLoss/(4 \[Pi] f)) Sqrt[f/Quantity["Gigahertz"]];
(* Equations 1.12 and 1.13 *)
ZS = I 2 \[Pi] f (L0 + L1 f + L2 f^2 + L3 f^3);
ZO = 1/(I 2 \[Pi] f (C0 + C1 f + C2 f^2 + C3 f^3));

(* Example usage with OPEN parameters *)
\[CapitalGamma]i /. ZT -> ZO /. {
   C0 -> Quantity[46.597 10^-15, "Farads"],
   C1 -> Quantity[-90050 10^-27, "Farads"/"Hertz"],
   C2 -> Quantity[10138 10^-36, "Farads"/"Hertz"^2],
   C3 -> Quantity[-358.16 10^-45, "Farads"/"Hertz"^3],
   offsetDelay -> Quantity[67.442, "Picoseconds"],
   offsetLoss -> Quantity[2.7871, "Gigaohms"/"Seconds"],
   offsetZ0 -> Quantity[53.549, "Ohms"],
   f -> Quantity[f, "Gigahertz"]
};
S11 = % /. Cases[Variables[%], v_ /; Head[v] =!= Quantity -> v -> Quantity[v, "DimensionlessUnit"]] /. {
    Quantity[v_, "DimensionlessUnit"] :> v,
    Quantity[v_, 1/"DimensionlessUnit"] :> v,
    Quantity[v_, Sqrt["DimensionlessUnit"]] :> v
};
Plot[Abs[S11], {f, 0, 18}, AxesLabel -> {"Frequency (GHz)", "|S₁₁|"}]


[小太]

Same thing as above, but fitting was limited to just 0~10GHz

Short F	Offset Z0 (Ω)	Offset delay (ps)	Offset loss (GΩ⋅s-1)	L0 (10-12 H)	L1 (10-24 H⋅Hz-1)	L2 (10-33 H⋅Hz-2)	L3 (10-42 H⋅Hz-3)
Given	51.56	83.61	2.57	-889.07	9989.25	9983.11	-1413.8
Fitted	52.319	85.325	1.0887	-965.37	-61723	29349	-2997.3
Open F	Offset Z0 (Ω)	Offset delay (ps)	Offset loss (GΩ⋅s-1)	C0 (10-15 F)	C1 (10-27 F⋅Hz-1)	C2 (10-36 F⋅Hz-2)	C3 (10-45 F⋅Hz-3)
Given	50	76.37	1.96	-351.35	-3947	1326.61	-507.08
Fitted	52.277	77.447	1.2253	-309.12	-12284	219.52	-456.68
Load F	Offset Z0 (Ω)	Offset delay (ps)	Offset loss (GΩ⋅s-1)
Given	54.16	4.77	10
Fitted 1*	77.775	0.76737	340.06
Fitted 2*	54.142	6.2391	10.000

* The "Load F" fitting decided to produce stupid values initially, so I ran it again but forced the offset loss to 10, and it gave more reasonable values. Both results are shown


I'm not really qualified to make a conclusion here (I'm too inexperienced), but if I were to make one anyways:
 - Assuming the VNA was calibrated correctly with the Rosenberger kit (and kit was within spec)
 - Applied EM's short and open have calibration coefficients that seem fitted to a max of 6~8GHz rather than all the way up to 10GHz
 - Applied EM's load coefficients seem to be way off
 - Ignoring the load coefficients (I haven't worked out how to apply them to measured data yet), the load only has a return loss of ≥36dB up to 4GHz, and it progressively gets worse until 29dB at 10GHz, which doesn't sound like it's within spec

[EE-digger]

You can also ask yourself how good your results have to be.  I've done a lot of chip antenna layout and tuning work.  After cal to the end of my VNA cable with SMA standards, I still have to extend it over an SMA to U.FL to another U.FL on the pcb.

In a case like this, you're better off creating your own open, short load on the pcb itself.  Then, just put an RF short on the last connector and run auto or manual compensation for the short.

added - and, of course the frequencies you're working on will affect this big time

[joeqsmith]

I quit after tearing my hair out for a good 2 hours trying to figure out the METAS tools, so if someone else can use it to generate a model I'd be grateful (and maybe post a link to a tutorial?)

Sorry for the necropost.  I had watched the METAS video and they did not go into any details on how to accomplish this.  Going through this thread, I too am curious how to derive the coefficients for an unknown set.   

The last time I did this, I used one of the local members Matlab scripts. I would like to try this with METAS.  Any pointers to save my hair are appreciated.   

Also:
https://www.eevblog.com/forum/testgear/kirkby-calibration-kit-alternatives/msg1437115/#msg1437115
SUJ, nice job with the waveguides.  I am on a similar path.   

[shabaz]

Not read the entire thread (too long!) but I was in a situation where I had an OK-ish quality, new OSL Tee, but without the coefficients. I asked Dr. Kirkby nicely, and he was kindly willing to measure it and provide the coefficients for me. The price was extremely reasonable (this was about five years ago). So, that might be an option for the cases where someone's ended up with a cal kit (say from eBay) which might accidentally be missing the data.

Recently I've been using the SDR-kits cal kit (PDF doc), but have not exercised them beyond 1 GHz yet. The kit is not bad for what it is, however, the supplied 50-ohm load in my case had 48.52 ohm DC resistance. That's normally not an issue (there's another thread on the topic elsewhere) and some VNAs allow for entering that value, but if that's a concern (personally, I wanted closer to 50 ohm because I also tend to use it for various purposes at low frequencies, from DC to tens of kHz) then the very easy solution is to just replace that with your own 50-ohm load, e.g. Huber & Suhner ones are often much closer at DC.

[joeqsmith]

A much better thread:
https://www.eevblog.com/forum/rf-microwave/calculating-correction-data-for-an-unknown-vna-calibration-kit/

[小太]

I quit after tearing my hair out for a good 2 hours trying to figure out the METAS tools, so if someone else can use it to generate a model I'd be grateful (and maybe post a link to a tutorial?)

Sorry for the necropost.  I had watched the METAS video and they did not go into any details on how to accomplish this.  Going through this thread, I too am curious how to derive the coefficients for an unknown set.   

The last time I did this, I used one of the local members Matlab scripts. I would like to try this with METAS.  Any pointers to save my hair are appreciated.   

I just followed the instructions by alan.bain posted a bit further up:
(You have to open a measurement file first)

Are you trying to use VNA Tool?  I had previously had issues with this and the process I got was

Right-click on the file of measurements, then select Database / Fit Calibration Standard Model (as per the screenshot)

You then choose which type of standard it is and which parameters you don't want to adjust (I would suggest that initially you reduce the set size e.g. remove L2/L3 from the fit for a short) and hit "Start Optimization".   If you try and fit too many it can get into "Not Responding" state which is a bit poor!

[vsilves]

Novice here and recently found this thread and looked at Dr. Kirkby statement back in 2018 with respect to DC resistance (DCR) of 50Ω loads, which I trust to be "correct."

<Dr. Kirkby QUOTE>
I feel that's irrelevant, as it does not indicate the RF performance at all, because the real part of the resistance is not independent of frequency.
...
S11 and S22 measurements can't be expected to be very accurate at low frequencies, as the return loss of the attenuator is very good. A high return loss means a high measurement uncertainty. As the frequency rises, so the return loss of the attenuator deteriorates, and the S11 and S22 measurements are more meaningful. But the S21 measurement is useful at all frequencies.
</Dr. Kirkby QUOTE>

I evaluated all my chinisium 50Ω loads using a converted SSA3021X+ for S11 return loss (RL) after performing an OSL calibration with the presumably better Mini-Circuits ANNE-50L+ DC-12 GHz Precision 50 SMA-M load --see attached PNG files--.  Used a Fluke 289 with a female SMA connector to banana plug DIY rig to measure DC resistance (DCR) relative to one of my NanoVNA F-3 "shorts" as offset --file name includes the DCR measurement for each load--.  Number 9 file is actually one of a set of 5 terminators, all with similar RL and DCR profiles (could not upload more than 10 files to this forum).

After getting the Mini-Circuits load and reading the comment about RL, I consider the job of 50Ω loads to absorb energy and return as close to 0 as possible (opposite of open and close, required to return as much as possible, 1 ideally).  Based on this, the ANNE-50L+ (mostly better than -40 dB RL), and my Long1 and Long2 (better than -35 dB) are the only loads I plan to use for calibration of NanoVNA and Siglent SVA. Dot1, dot2, dot3 could be used.  Most of these arrived with NanoVNA F2, F3, TinySA, or Ultra, I think, but lost track of which is which.

All the other 50Ω loads I have appear to be made of copper-chinisium alloy and show appalling RL, reflect significantly more energy than the 50Ω loads in the previous paragraph.

These results are very surprising to me, and made me change my prior assumptions.  I previously focused exclusively on DCR and considered my L1, L2, L3 the best loads for OSL calibration based on their proximity to 50.0Ω DCR.  Now I consider them unusable based on their poor RL.

I purchased an N Type OSL set from Amazon and it was similarly 2% off on DCR, but I may have misjudged it, and now I need to measure how it responds to RL by observing its S11 and completely disregard its DCR.

Eventually, I will find time to OSL calibrate the Siglent SVA using one or more poor RL 50Ω loads and then measure all the others to see if I draw different conclusions.

[vsilves]

Follow up post.  The N-type load in the Amazon cal set turned also horrible with respect to RL (only better than -20 dB RL at low frequencies). Had to remove N male to SMA female (calibration plane) to attach the N-type load.

Decided to look at all my loads in the Smith chart as well (after recalling previous calibration, re-confirming previous RL values, using same N to SMA calibration plane --no cables--, etc.).  All my good loads (Mini-Circuits, Long1..2, and Dot1..3) had exceptional single dot on the center of the Smith chart all the way up to 3.2 GHz.  Dot2 distinguised itself as-good as better ones.  Dot1 and dot3 a bit worse but also great.  All others (L1..3, Long China1..5, and N-type): oh my gosh, big arc, nothing close to constant impedance (some with inductance and some with capacitance). The Long China even got me an open from time to time (it's not a bug, it's a feature, still on return window but keeping as examples of what loads should NOT be).

Attached one example each for each type of load name tested (rest on same group had similar profiles). These results sort of confirm which 50Ω loads are good and which ones are junk.

[hendorog]

Comparing loads is a bit tricky - it ultimately just a comparison against the load you calibrated with - along with added noise and drift which are introduced by the VNA you use.

The first measurement, of the MC ANNE load, which is what you used for the cal, tells you something about the stability of the VNA but it tells you nothing about the load itself.

Attached is a screenshot showing:

ANNE-50+ load in yellow
ANNEF-50+ in blue

The cal was SOLR (with isolation as a test), and the cal kit was a combo:
Female cal parts: 85033
Male cal parts: 85052C




The second screenshot with the blue trace only shows the Female cal load re-measured after being disconnected and re-connected again.




The next screenshot is from METAS, and is a NanoVNA cheapo load (measured a while ago using an adapter to protect the VNA connectors)




The next screenshot is the best INMET load I have measured




The final one is the Load from a mid priced Taobao cal kit.

[tautech]

Decided to look at all my loads in the Smith chart as well (after recalling previous calibration, re-confirming previous RL values, using same N to SMA calibration plane --no cables--, etc.).  All my good loads (Mini-Circuits, Long1..2, and Dot1..3) had exceptional single dot on the center of the Smith chart all the way up to 3.2 GHz.  Dot2 distinguised itself as-good as better ones.  Dot1 and dot3 a bit worse but also great.

Changing the Smith chart scale can reveal some surprises and make the trace large enough to place markers on.
Good idea to capture at a 1U scale first then have something to compare zoomed in captures to.

[hendorog]

Here is the calibration load measured on the SSA3075XR with 100 averages.
No disconnect/reconnect done.
Male cal kit same as before.



Without 100 averages it looks about the same as vsilves

And the same ANNE load *** Edited initial image was not the same load ***

[joeqsmith]

Novice here and recently found this thread and looked at Dr. Kirkby statement back in 2018 with respect to DC resistance (DCR) of 50Ω loads, which I trust to be "correct."
...

I evaluated all my chinisium 50Ω loads using a converted SSA3021X+ for S11 return loss (RL) after performing an OSL calibration with the presumably better Mini-Circuits ANNE-50L+ DC-12 GHz Precision 50 SMA-M load --see attached PNG files--.  Used a Fluke 289 with a female SMA connector to banana plug DIY rig to measure DC resistance (DCR) relative to one of my NanoVNA F-3 "shorts" as offset --file name includes the DCR measurement for each load--.  Number 9 file is actually one of a set of 5 terminators, all with similar RL and DCR profiles (could not upload more than 10 files to this forum).
...

"presumably"  I assume you read and understood the datasheet.  So you know what they spec them to worse case. 

Data showing two of the best ANNE terminators compared with a few others.  Reference is vintage Agilent PNA calibrated with 85033e standards.  The other six ANNE loads I have were not near that good.  Sorting was a cheap way to get a decent load standard.
https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg3323942/#msg3323942

Making various DC measurements for loads.
https://www.eevblog.com/forum/testgear/nanovna-v24-sma-cal-load/

Recently I attempted to tighten up my measurements with the low cost VNAs, using a couple of T-Checks to verify the quality of the calibration.  This was a comparison between the low cost VNA and my PNA.  Even after adding support for databased standards, characterizing my low cost standards,  verifying my math against Agilents..... when it was all said and done, the VNA was the limiting factor, not the cheap standards I use with them. 
   
https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg5665155/#msg5665155

More discussion on the ANNE terminators.
https://www.eevblog.com/forum/projects/20db-rf-attenuator-seeking-feedback-to-improve/msg3093968/#msg3093968

[vsilves]

Appreciate the hints, measurements, and comparisons from more experienced folks.

Thanks @tautech for hint on zooming into Smith chart (have never done it yet, but makes sense, and should be easy to figure out).

@joeqsmith: Haha, I did "presume".  I had seen ANNE specs, but did not take into account or recall numbers during my measurements (complete gamble and bad move on my part, so thanks for pointing it out).  Looking at ANNE specs again I see min RL is only -26 dB (DC-4GHz) but that is, of course, for 4 GHz.  Typical is -30.84 dB (4 GHz), and closer to my upper bound -33.81 dB (3.05 GHz). Higher use I probably have is 2.4 GHz wifi so around -35 dB typical. Your two ANNE traces and hendorog's M and F ANNE traces are roughly in line with typical specs.  I see now that I got a not so bad ANNE, could be better, but could also be a lot worse. In the land of the blind (i.e., my basement), my one-eyed ANNE is king.

After yesterday measurements, I had asked ChatGPT what RL to expect for a good 50Ω load 0-3.2GHz and the answer was high-quality load (metrology-grade, precision loads) better than -45 dB up to 3 GHz.  Standard lab-grade (high-quality but not metrology grade) better than -35 dB to -40 dB across 3.2 GHz.  General purpose 50 Ω terminations (not for precision calibration) -30 dB or worst at higher frequencies. Based on this I keep my modest aim on getting loads -35 dB or better, if I can help it.

For my calibration/measurements, I had increased the trace point to 2000 for 0-3.2GHz.  Surprised hendorog uses the default 201 points for 0-7.5GHz. On the other screen covering ANNE-50+ and ANNEF-50+, S11 and S22 respectively, to 8.5 GHz I cannot see the number of points.  Should not make a hughe difference but couldn't that miss some peaks or valleys, if present?

Hendorog's female ANNE dropping roughly from -40 dB to -75 dB after removal and reinsertion is crazy, no?  When I measured the Smith charts I first looked at the RL (after power cycle, calibration recall, had even removed N-to-SMA and put back to restore same cal plane, etc) and the RL for all loads looked indistinguishable from the original uploaded measurements (to my uncalibrated eye).

Hendorog's cal load on SSA3075XR with 100 averages is showing roughly below -60 dB (around -70 dB even), and without averages roughly at -40 dB (similar to my measurement). Question.  Shouldn't we care about the worst case RL and engage max hold instead of any averaging? (I think I'm gong to try that, unless I hear it makes no sense).

In my original measurements for RL, I had 10 averages on, but at around 1-2 sec sweep time (cannot recall) I probably did not even wait all 10 runs before taking screenshots. For Smith I do not think I turned on averaging at all. On hendorog's measurement to 7.5GHz of the calibration load I see the same (lots of) jagged edges (be it noise or uncertainty??), same as the measurement of my ANNE cal load (immediately after calibration without removing it). Curious to know why measuring whatever cal load is used results in a more jagged trace than measuring loads other than the reference cal load, independent of overall badness of RL of each load. (Will try to cal with my Long1 and then measure my ANNE, expect the jaggedness to go away, even if I don't know why).

@joeqsmith I like your "the VNA was the limiting factor, not the cheap standards I use with them" comment and since I have a NanoVNA-F3 will try and experiment to see what I find now that I know a bit more about he quality of my loads (how good or bad the NanoVNA is compared with the Siglent, using same loads).

Just learning all this and, overall, I am very pleased nobody mentioned DCR in their responses, a topic that was very popular on older posts in this thread ;-)

[joeqsmith]

Your two ANNE traces and hendorog's M and F ANNE traces are roughly in line with typical specs.  I see now that I got a not so bad ANNE, could be better, but could also be a lot worse. In the land of the blind (i.e., my basement), my one-eyed ANNE is king.

Mine were below 40dB compared with a metrology grade standard on a name brand VNA.   Below 40 is pretty much in the noise floor.    How do you know what yours is?  You would need something as a reference to compare it with.

For my calibration/measurements, I had increased the trace point to 2000 for 0-3.2GHz.  Surprised hendorog uses the default 201 points for 0-7.5GHz. On the other screen covering ANNE-50+ and ANNEF-50+, S11 and S22 respectively, to 8.5 GHz I cannot see the number of points.  Should not make a hughe difference but couldn't that miss some peaks or valleys, if present?

After upgrading the Agilent for 9GHz and running through the alignment,  I borrowed a set of metrology grade standards that I transferred to my E-Cals.  I maxed out the memory on them using a high sample count.  IFBW was cranked way down.  I also tried two different power levels to see if there were any improvements.   That was what I used to then re-characterize the low cost standards.   

I think for the T-Check experiments, I limited the number of points to 1600.   The problem isn't the T-Check stability, its the low cost VNA has a lot of noise as I get beyond 4GHz.

Hendorog's female ANNE dropping roughly from -40 dB to -75 dB after removal and reinsertion is crazy, no? 

No.  See response below:

Interesting measurements. I would say this though, any return loss measurements below 35 dB are not terribly meaningful. The uncertainty of reflection measurement is entirely dependent on the return loss of the load standard. A load with 45 dB of return loss would allow VNA measurements of 35 dB with +/- 3.3 dB of accuracy and +/- 1dB at 25 dB. The best load in the world is at NIST and it's 52 dB. To get a better calibration one would have to use TRL where it is possible to get the directivity uncertainty down to 60 dB with a pristine air line.

Just learning all this and, overall, I am very pleased nobody mentioned DCR in their responses, a topic that was very popular on older posts in this thread ;-)

I did provide a link for DCR measurements.   That was the same thread where OP was talking about SWR three places beyond the decimal point.  There was a recent thread running from ham trying to use some low cost terminators for a standard.   You should try using the search engine.  There is a lot of useful information available.