VNA Compact Cal Kit from Rosenberger

[sixtimesseven]

I'm looking around for a Cal Kit to ultimatly replace my DIY attempt (which is pretty bad at 6GHz).
So far all the used HP/Agilent options seem really expensive to me.

I looked at the Kirkby cal kits but I'm a bit reluctant to get SMA connectors at that price point.

However, I stumbled on an offering from Rosenberger's Compact Cal Kits: https://www.rosenberger.com/0_documents/de/catalogs/ba_tm/TM_Compact_CalKits.pdf
In particular: 03K30R-MSOS3
Datasheet: https://catalog.rosenberger.com/images/documents/db/03K30R-MSOS3-RMS.pdf

They are not cheap, 1250USD 850USD (Rosenberger quote) for 3.5mm SOL or 1450USD 1025USD (Rosenberger quote) for SOLT, but still a far way from Keysight Kits and pretty close to used HP/Agilent 3.5mm kits in dubious condition and they go up to 26GHz...
Specs look pretty good. But I'm still interested if sombody uses them already?

Andreas


EDIT:
Updated pricing with actual quoted prices from local Rosenberger distributor. Prices are significantly lower then Mouser/Digikey prices so ask!

EDIT II:
Received kit after about a month. I'm pretty impressed by it so far. I used VNA Cal Kit Manager http://www.vnahelp.com/ckman_new2.html to send the cal via GPIB.
A little quirk in Windows is that, if you see comma seperated values in the CKM, you have to change them to point seperated values.
Control Panel\Clock and Region -> Change date, time or number formats -> Additional Settings -> Decimal Symbol -> Change from "," to "."
I have attached the .ckm file to load directly via CKM.

[rastro]

That's a good find.  Rosenberger is a good brand/reputation.

I wonder if anyone knows who makes HP/Agilent/KS calibration kits?  I suspect they are re-badged.

rastro

[TheSteve]

Keysight makes there own 3.5mm kits such as the 85052x. Years ago and possibly still today they have rebranded Maury Microwave kits. I believe the all in one kits like the 85521a are rebranded.

Rosenberger should be good stuff. I've never used them though. Cal kit deals pop up on eBay all the time, but they usually sell pretty quick.

[sixtimesseven]

Cal kit deals pop up on eBay all the time, but they usually sell pretty quick.

But the prices are ... Rather high I found. I mean a as-is kit with how knows connector health for 600+ Euros 
https://www.ebay.de/itm/HP-Agilent-Keysight-85033D-3-5mm-Calibration-Kit/202660980808?hash=item2f2f892848:g:YUQAAOSwok9cwB9Y

[TheSteve]

Keysight makes there own 3.5mm kits such as the 85052x. Years ago and possibly still today they have rebranded Maury Microwave kits. I believe the all in one kits like the 85521a are rebranded.

Do you happen to know who makes 85521A? If it's below 1/2 Keysight's price I will get one to replace mine with crooked connector, just so I can sleep well.

No idea - but all of the 3.5mm stuff they make is PSC (precision slotless). Those cal kits look pretty generic, there seem to be quite a few companies who make them at the moment.

[cloidnerux]

Do you happen to know who makes 85521A? If it's below 1/2 Keysight's price I will get one to replace mine with crooked connector, just so I can sleep well.

Some of those cal kits are made by "Spinner GmbH":
https://www.spinner-group.com/de/produkte/messmittel/vna-messmittel
I think they are rebranded by R&S, Keysight and a few others.

[rastro]

Rosenberger 3.5mm female connector do not appear to be PSC (precision slotless).

The 03K30R-MSOTS3 (female 26GHz) kit uses there PRC-3.50 connector and their datasheet shows a slot on the female interface.

The concern with buying used PSC hardware is the delicate internal fingers inside the female connector can be damaged.  This can be hard to detect without close inspection under magnification. 

I think the slotted 3.5mm connectors are a little more mechanically forgiving/robust when compared to the slot-less. 

[OwO]

I don't get all the effort to make an "ideal" connector in a cal kit because the best accuracy is achieved when the cal kit connector closely matches the DUT's connector in characteristics.

[Bud]

Honestly I can't recall seeing anywhere in VNA books a requirement to match the DUT connector characteristics. This would simply be unreslistic to cover all use cases. Connectors in calibration kits are made with tight tolerances because they need to be precisely characterized for calibration coefficients and have repeatability when made. Nobody wants to individually characterize every single kit produced. In some cases cal coefficients are derived from simulation for which it is needed to know precise geometry of the connector parts.

[sixtimesseven]

Do you happen to know who makes 85521A? If it's below 1/2 Keysight's price I will get one to replace mine with crooked connector, just so I can sleep well.

Some of those cal kits are made by "Spinner GmbH":
https://www.spinner-group.com/de/produkte/messmittel/vna-messmittel
I think they are rebranded by R&S, Keysight and a few others.

I agree, R&S cal kits look exactly the same.
Contacted them and got the attached product catalog. Still waiting on the quote for the DC-13 and DC-26 OSLT cal kit from them(product nr: 533828 / 533882).

[sixtimesseven]

I got the offer from Spinner (Keysight / R&S OEM):

Spinner 533828 -- 950USD + Tax etc. --Datasheet Attached
OSLT 3.5mm female (IEC 60169-23) connector, 0-13Ghz

Phase match open:
0 - 4 GHz / 1,5° max.
>4 - 8 GHz / 3,0° max.
>8 - 13 GHz / 4,5° max.

Phase match short:
0 - 4 GHz / 1,0° max.
>4 - 8 GHz / 2,0° max.
>8 - 13 GHz / 3,5° max

Load, return loss
0 - 4 GHz / 40 dB min.
>4 - 8 GHz / 34 dB min.
>8 - 13 GHz / 28 dB min.

Thru, return loss
0 - 4 GHz / 34 dB min.
>4 - 8 GHz / 28 dB min.
>8 - 13 GHz / 25 dB min.


Spinner 533882 -- 2100USD + tax (Out of my league) --Datasheet Attached
OSLT 3.5mm female (IEC 60169-23) connector, 0-26.5Ghz

Phase match open:
0 - 5 GHz / 1,5° max.
5 - 15 GHz / 3° max.
15 - 26,5 GHz / 4,5° max.

Phase match short:
0 - 5 GHz / 1° max.
5 - 15 GHz / 2,5° max.
15 - 26,5 GHz / 4° max.

Load, return loss
0 - 5 GHz / 42 dB min.
5 - 15 GHz / 36 dB min.
15 - 26,5 GHz / 32 dB min.

Thru, return loss
0 - 5 GHz / 34 dB min.
5 - 26,5 GHz / 30 dB min.


Rosenbergs 03K30R-MSOS3 850USD (quote)
26.6GHz 3.5mm female (IEC 61169-23) connector
Specs for comparison:

Open
Error from nominal phase
1 ≤ 1.0°, DC to 4 GHz
 ≤ 2.0°, 4 GHz to 8 GHz
≤ 3.0°, 8 GHz to 26.5 GHz

Short
Error from nominal phase
≤ 1.0°, DC to 4 GHz
≤ 2.0°, 4 GHz to 8 GHz
≤ 3.0°, 8 GHz to 26.5 GHz

Load
Return loss ≥ 40.0 dB, DC to 4 GHz
≥ 35.0 dB, 4 GHz to 8 GHz
≥ 30.0 dB, 8 GHz to 26.5 GHz



In conclusion:
The Rosenberger OSL costs slightly less then the Spinner 0-13GHz version. It lacks the throu standard (which seems likt the easiest-to-replace standard to me), goes to 26GHz and has better phase specs for the open and short but a significantly better return loss for the load. Only bummer is 12-13 weeks of indicated delivery time

So far I lean towards the Rosenberger. Any objections?


EDIT: Mixed up Short and Thru for Spinner kits. Fixed.

[rastro]

I got the offer from Spinner (Keysight / R&S OEM):
...
So far I lean towards the Rosenberger. Any objections?

Thanks for sharing that information.  Unless I missed something the Rosenberger 26.5GHz looks like it has specifications still very close to the Spinner 26.5GHz standard.  I suspect that any random production Rosenberger unit would meet the Spinner 26.5GHz specification.  The specification gap may be attributed to how much margin each manufacture is comfortable with.

Dam it, now you got me thinking about purchasing a Rosenberger 26.5GHz cal standard.  Gonna have to wait for a few months for finances.  I like the case and coefficient card that is included.

Some other thoughts:
1. I wonder how easy it is to remove the standards from the holder?  I didn't see anything in the data sheet.
2. Out of curiosity I wonder how the Coefficients compare between the Rosenberger 26.5GHz and Spinner 26.5GHz.  Probably meaningless...
3. Outer conductor is different Rosenberger = Stainless steel Passivated; Spinner = CuBe / gold-plated.  Don't know this affects performance/longevity?

rastro

[sixtimesseven]

Dam it, now you got me thinking about purchasing a Rosenberger 26.5GHz cal standard.  Gonna have to wait for a few months for finances.  I like the case and coefficient card that is included.

Rosenbergs delivery time seems to match your financing requirements pretty well.

1. I wonder how easy it is to remove the standards from the holder?  I didn't see anything in the data sheet

I would suspect it to be very easy. The OSL part looks identical. I suspect one screw on the back.

2. Out of curiosity I wonder how the Coefficients compare between the Rosenberger 26.5GHz and Spinner 26.5GHz.  Probably meaningless...

Coefficents are in the datasheets. See atachments. They are also freely available on the net.

[rastro]

Sorry just realized I had the information at hand.  Was editing prior message when you replied...

[Bud]

@sixtimesseven are you sure you read the specs for the Short correctly?   
Seems it is absorbing energy, not reflecting it. 

[sixtimesseven]

@sixtimesseven are you sure you read the specs for the Short correctly?   
Seems it is absorbing energy, not reflecting it. 

You are correct. I mixed up Short and Thru specs while translating from German. Sorry!

[TheUnnamedNewbie]

I don't get all the effort to make an "ideal" connector in a cal kit because the best accuracy is achieved when the cal kit connector closely matches the DUT's connector in characteristics.

Could you care to explain/clarify? The best accuracy should be achieved when the error on the calkit with respect to the stored values used in the VNA are lowest, shouldn't they? The advantage of an ideal connector in a calkit is because an ideal connector means you need less precise compensation values to cancel out the non-ideal connector. After all, that is why sliding-loads are a thing - they wouldn't work with near-perfect connectors.

OP, getting a through that can match the performance of that rosenberger kit might not be as straight-forwards as you think. 3.5mm air-dielectric through are not exactly cheap.

[Bud]

@sixtimesseven are you sure you read the specs for the Short correctly?   
Seems it is absorbing energy, not reflecting it. 

You are correct. I mixed up Short and Thru specs while translating from German. Sorry!

Ah, that RL figure was for the Thru, makes sence.

[OwO]

Could you care to explain/clarify? The best accuracy should be achieved when the error on the calkit with respect to the stored values used in the VNA are lowest, shouldn't they? The advantage of an ideal connector in a calkit is because an ideal connector means you need less precise compensation values to cancel out the non-ideal connector. After all, that is why sliding-loads are a thing - they wouldn't work with near-perfect connectors.

The point was that when you are measuring the DUT you are going through its non-ideal connectors, which for example may have a nominal characteristic impedance different from 50 ohms due to physical design. No matter how good your cal kit is, this error gets added to the measurement, and conversely if you construct a cal kit out of a connector that closely resembles the DUT's connectors, the error of the connector gets calibrated out.

[OwO]

In general I think people pay way too much for anything related to a network analyzer, and never realize the instrument or even cal kit is usually not the performance bottleneck, but rather measurement technique and the DUT itself.

[dcarr]

In the spirit of OwO’s comment— if you really want to characterize a PCB mounted device, you should be using on board TRL or similar fixtures.  These will calibrate out your connectors, PCB transitions, and your transmission line geometry.  Best of all, it’s cheap!  You don’t necessarily have to have a traditional cal kit to do it.

[sixtimesseven]

In the spirit of OwO’s comment— if you really want to characterize a PCB mounted device, you should be using on board TRL or similar fixtures.  These will calibrate out your connectors, PCB transitions, and your transmission line geometry.  Best of all, it’s cheap!  You don’t necessarily have to have a traditional cal kit to do it.

Well this is true if you use <10$ SMA edge launch connectors which will vary by quiet a bit but is pretty cheap. And arguably this precision is all you need for most application which use SMA edge launch connectors in the real product anyway. However, if you want to get rid of connector variations, they won't do. Have a look at the 3.5mm/2.92mm edge launch connector prices! And you will need 6+ for the TRL kit (depending on frequency range) and 2 for each DUT!

[rastro]

...
OP, getting a through that can match the performance of that rosenberger kit might not be as straight-forwards as you think. 3.5mm air-dielectric through are not exactly cheap.

I didn't see any indication on the 3.5mm(f) MSOT (03K30R-MSOTS3) data sheet that the "Through" is an airline. 
It looks like both use the same short/open/load on the data sheet.

Both mouser and digikey have prices between $1200USD and $1600USD for these kits.  The 850USD looks like good deal in comparison.

[dcarr]

Great point about the connector repeatability.  I can recommend Rosenberg 32K243-40ML5 as a high performance yet moderately priced option.  It clamps on to the edge of the board and if you bend down the center PCB contact slightly, it acts like one of the expensive Southwest Microwave connectors.  (ie: You clamp it on to the PCB in one place and then when finished you can unclamp it and move it to a new location.)

David

[TheUnnamedNewbie]

Could you care to explain/clarify? The best accuracy should be achieved when the error on the calkit with respect to the stored values used in the VNA are lowest, shouldn't they? The advantage of an ideal connector in a calkit is because an ideal connector means you need less precise compensation values to cancel out the non-ideal connector. After all, that is why sliding-loads are a thing - they wouldn't work with near-perfect connectors.

The point was that when you are measuring the DUT you are going through its non-ideal connectors, which for example may have a nominal characteristic impedance different from 50 ohms due to physical design. No matter how good your cal kit is, this error gets added to the measurement, and conversely if you construct a cal kit out of a connector that closely resembles the DUT's connectors, the error of the connector gets calibrated out.

If you want to measure only the DUT on a test-board, you need to calibrate on the test board with TRL (ideally even TRLL). I don't know about the math and I never do this, since I tend to either use GSG probes or waveguide stuff, but I have once been told that you can get better performance if you first calibrate up to the ports of your expensive VNA cables, then calibrate the TRL, instead of going straight for the TRL (this can also give you an idea about the repeatability of the TRL, since you can see the difference between connectors if you look at the time-domain).

In the spirit of OwO’s comment— if you really want to characterize a PCB mounted device, you should be using on board TRL or similar fixtures.  These will calibrate out your connectors, PCB transitions, and your transmission line geometry.  Best of all, it’s cheap!  You don’t necessarily have to have a traditional cal kit to do it.

Well this is true if you use <10$ SMA edge launch connectors which will vary by quiet a bit but is pretty cheap. And arguably this precision is all you need for most application which use SMA edge launch connectors in the real product anyway. However, if you want to get rid of connector variations, they won't do. Have a look at the 3.5mm/2.92mm edge launch connector prices! And you will need 6+ for the TRL kit (depending on frequency range) and 2 for each DUT!

The thing is that you want your connectors to be as ideal as possible. After all, any variability in the connectors (both between different connectors and the repeatability of the connector itself) will translate into larger errors and uncertainty on the device performance. Just because it will be used with cheap connectors does not mean it is okay to measure it with cheap connectors - you want as precise measurement as possible, since errors tend to compound in simulations and when you start cascading things. And while you might not get big amplitude variations, phase variations can be very important - after all, you need to 'un-rotate' the transmission lines you use.

...
OP, getting a through that can match the performance of that rosenberger kit might not be as straight-forwards as you think. 3.5mm air-dielectric through are not exactly cheap.

I didn't see any indication on the 3.5mm(f) MSOT (03K30R-MSOTS3) data sheet that the "Through" is an airline. 
It looks like both use the same short/open/load on the data sheet.

To be a APC3.5 mm connector that adheres to the standard, it has to be a airline connector. There is no such thing as a non-airline APC3.5 mm connector.

[sixtimesseven]

Just got my Rosenberg SOLT Kit - Much earlier than expected 

Went for the throu version. And it looks like it is indeed an airline. Looks really nice.
The throu is also not screwed on, but is attached via an internally fixed bolt and can be rotated by 90 degree. Which is great!


All in all very happy so far. Almost a bit scared to attach my Succoflex SMA cables to this 

[rf-messkopf]

Congratulations to your Rosenberger cal kit.

If somebody else is also pondering over the purchase of a professional cal kit: Don't overlook the possibility of obtaining a cheap cal kit and having it measured on an accurately calibrated VNA, and using the measured S-parameters directly for the correction of the kit.

In this way you can transfer almost all of the accuracy of a professional cal kit to a much cheaper one, including the return loss of the loads. A parametric error model of the cal standards will do that only to some extent, and even then most of the time the loads will have to be assumed ideal. Of course, you are left with errors due to the stability of the cal standards and the repeatability of connectors.

A number of VNAs support the use of Touchstone files for cal kit error correction. And even if your VNA doesn't, you still could do the vector error correction of your measured results offline. That would of course be an inconvenience, but then again you will not always need the full accuracy of your cal kit.

[TheSteve]

Looks great!

Now you can save up for the male version 

[TheUnnamedNewbie]

Congratulations to your Rosenberger cal kit.

If somebody else is also pondering over the purchase of a professional cal kit: Don't overlook the possibility of obtaining a cheap cal kit and having it measured on an accurately calibrated VNA, and using the measured S-parameters directly for the correction of the kit.

In this way you can transfer almost all of the accuracy of a professional cal kit to a much cheaper one, including the return loss of the loads. A parametric error model of the cal standards will do that only to some extent, and even then most of the time the loads will have to be assumed ideal. Of course, you are left with errors due to the stability of the cal standards and the repeatability of connectors.

A number of VNAs support the use of Touchstone files for cal kit error correction. And even if your VNA doesn't, you still could do the vector error correction of your measured results offline. That would of course be an inconvenience, but then again you will not always need the full accuracy of your cal kit.

The repeatability question got me wondered - how hard would it be (and would it be usefull) to home-brew a kind of sliding-match to do sliding-match calibration?

[rf-messkopf]

The repeatability question got me wondered - how hard would it be (and would it be usefull) to home-brew a kind of sliding-match to do sliding-match calibration?

A sliding load relies on the precision of the airline along which the load slides. Just like the airlines that are used for TRL and other purposes, these are manufactured to very tight tolerances. I doubt that it will be feasible to DIY them, unless you have a very sophisticated machine shop and the necessary skills available. And to manufacture a good load element that slides along the center conductor and makes seamless contact to the outer conductor will be all but trivial.

I don't think that using connectorized pieces of rigid coax of various lengths will be good enough to replace a sliding load, mainly due to the connector repeatability issue, and also due to the precision of the rigid coax. I do not remember which variability to expect due to connector repeatability, but I remember some publications in IEEE journals where this was investigated experimentally. I can have  look for references in my notes, but I will have virtually no time the next 10 days or so.

That said, sliding loads are more or less obsolete today since one can correct even a mediocre but stable load with the help of S-parameter correction data. This is the way that Ecal units work. I remember having read that their performance can hold up to that of a sliding load, and may only be very slightly worse. That will depend on the frequency range as well, of course.

[sixtimesseven]

My SMA-T Connector arrived and I was able to do the T-Check.

I loaded the cal. coef. manually. I tried to use the "VNA Cal Kit Manager 2" to share the cmk file but ran into problems :\
https://www.eevblog.com/forum/rf-microwave/vna-cal-kit-manager/

[John_ITIC]

Hello, as this is the only thread related to this Rosenberger cal kit, i'm reviving it.

I just got my 8720C VNA operational after various repair work and i'm now looking at cal kits. I got a quote for the 3.5mm SOLT kits from Rosenberg. Both the female and male versions are slightly under $1,200 (each) and i'm considering buying both versions, for future flexibility.

I have checked and my 8720C has features for adding user-defined cal standards. The open can be defined via C1-C4 constants and offset delay/loss. So far so good.
I however noticed that the Rosenberger cal kit spec card also specifies the L1-L4 for the short standard. My 8720C however does not have any menu option for entering this information. Only offset delay/loss. I'm suspecting the short inductance can only be entered on newer VNAs.

I'm wondering how much inaccuracy will result when I can't enter the short standard L1-L4?

See the attached PDF for the cal constants. Zoom in on the cal card. Note that they are showing an incorrect cal card as it only goes to 12 Ghz. OPs card displayed above goes to 26.5 GHz. So, they goofed up when taking the picture. Or just photoshopped an old 12 GHz cal kit picture for the 26.5 GHz kit and forgot to update the card.

Thanks,
/John.

[joeqsmith]

I however noticed that the Rosenberger cal kit spec card also specifies the L1-L4 for the short standard. My 8720C however does not have any menu option for entering this information.

I would have guessed it supported this.  I assume you looked if there was any upgrades available.

[John_ITIC]

Quote
I however noticed that the Rosenberger cal kit spec card also specifies the L1-L4 for the short standard. My 8720C however does not have any menu option for entering this information.

I would have guessed it supported this.  I assume you looked if there was any upgrades available.

No, I already have the latest ROMs (1.05).

Would it be correct to assume that the calibration standards have higher precision as they go to 26.5 GHz and the L1-L4 would not matter for the 20 GHz 8720C?
Or, would it be possible to correct for the difference in post-processing on the PC-side, in case more precision is ever needed?

Thanks,
/John.

[rf-messkopf]

I'm wondering how much inaccuracy will result when I can't enter the short standard L1-L4?

I wrote a small Octave script that calculates the S11 response of open and short cal standards for a given set of parameters (in Keysight format): https://www.mariohellmich.de/lostfound/files/ck-model. You have to modify the script (enter your parameters, frequency range, number of points) and run it. The script will generate two Touchstone (.s1p) files for the open and short. Then you can plot the S11 response of the shot, both with the specified parameters, and with the offset inductances set to zero. You can use an external Touchstone plotting software, or modify the script to use Octave's own plotting functions. Check the phase difference of both cases. I'd bet they will differ by at least a couple of degrees at 26.5 GHz when the short is not fully corrected.

Another question is what the impact on the measured S-matrix of a DUT would be, but I think it is safe to assume that phase accuracy would be uncertain by at least be the same amount.

Or, would it be possible to correct for the difference in post-processing on the PC-side, in case more precision is ever needed?.

Not exactly by post-processing, but you would have to do the calibration offline. That is, measure the cal standards with your test setup (open, short, load at both ports, and a thru for a standard SOLT), and then the DUT. Then export the data to a PC (that makes eight Touchstone files), and do the math as well as the plotting. Of course, that could be automated via GPIB.

[rf-messkopf]

Check the phase difference of both cases. I'd bet they will differ by at least a couple of degrees at 26.5 GHz when the short is not fully corrected.

Okay, I became curious, and I'm too tired to do something useful. See the attachment. At 20 GHz, the short will differ by about 8.5 degrees when the offset inductances are set to zero. That is already quite appreciable, and much more than the accuracy specification of <3 degrees up to 26.5 GHz.

[joeqsmith]

I  don't know why you couldn't just use the VNA to collect the raw data and do everything on the PC side.   A lot of work but certainly possible.

[John_ITIC]

Thank you both. I suspect the deviation due to not taking the short L1-L4 into account is acceptable for my needs. I mainly want to characterize differential serial communications channels and i suspect I will not need absolute accuracy. I have just started learning how to operate my VNA so unfortunately not yet at all competent in the field.

Edit: Apparently, there is a way to include the short L1-L4 compensation automatically when using the external METAS VNA Tools application over GPIB. I will look into this closer.
https://www.eevblog.com/forum/testgear/kirkby-calibration-kit-alternatives/msg1436372/#msg1436372

Thanks,
/John.