The T-Check confusion
|
Hi All
There has been several contribution on how to create a T-Check for the NanoVNA and I have made a document about why these T-Check are not so promising. It is in fact not possible due to the lag of adequate error correct models as NanoVNA allows only error correction (at the best as the exact models are unknown) for one direction and not bidirectional. Have a look and see the experiment I have made to enlighten the interested reader ? Kind regards Kurt |
|
Hi Herman
First of all it is not the mini-VNA it is the DG8SAQ VNWA. Sorry these kind of measurements is only possible with a VNA having 10/12 term error correction facilities. NanoVNA is not able to do such measurements. Kind regards Kurt -----Oprindelig meddelelse----- Fra: [email protected] <[email protected]> P? vegne af Herman De Dauw Sendt: 9. december 2019 00:01 Til: [email protected] Emne: Re: [nanovna-users] The T-Check confusion Please, can you not make the PDF with plots of the NanoVNA, and not the Mini-VNA ? |
|
Hi Eric
A 10dB att in the Ch1 path will not cure the problem as you also have a (invisible) mismatch in the Ch0 sourse impdance as well. So two attenuator might give some further improvements. But it is a fundamental problem as the 10/12 term error correction "totally" remove the tx and rx port mismatch and establish pure 50 ohm source and load impedance in the two calibration planes at the end of the two test cables. Always remember the test cables are an integral part of the VNA instrument and influence removed by the calibration. For the VNWA two custom traces with the expression SS and SL shows the source impedance (SS) and load impedance (SL) after a calibration at the two mentioned calibration planes. However the NanoVNA or for that matter any of the NanoVNA software has such facilities to do a complete S11/S21/S12/S22 measurement controlling a test switch reverting the DUT direction or any other smart tricks for passive bi-direction DUT's Kind regards Kurt -----Oprindelig meddelelse----- Fra: [email protected] <[email protected]> P? vegne af erik@... Sendt: 9. december 2019 08:57 Til: [email protected] Emne: Re: [nanovna-users] The T-Check confusion Kurt, It appears (in my limited understanding) the poor matching to Z0 of CH1 (port 2) is the biggest reason why the T-check is not performing well on the nanoVNA. Would you agree that using a 10dB attenuator between DUT and CH1 (port 2) would solve most of this? What would be the impact of the remaining non-compensated delay? Should we add to the instructions that, for best nanoVNA 2 port DUT measurements one should seriously consider using a 10dB attenuator between DUT and CH1 (port 2) till new HW becomes available that solves the poor matching of CH1 (port2)? -- NanoVNA Wiki: /g/nanovna-users/wiki/home NanoVNA Files: /g/nanovna-users/files Erik, PD0EK |
|
Hi Gabriel
You are right it is a matter og software and for that matter firmware. As the NanoVNA is presently it does not have facilities embedded for neither calibration kit data and full 10/12 term error correction and none of the present software packages I know off has. That counts for NanoVNA-saver, NanoVNAsharp, NanoVNApartner or the TAPR adjusted program. Thee link you provide is not implemented in the NanoVNA-saver but of course of interest for Rune to consider. You say it is trivial to use in NanoVNA-saver but that requires a "software developer brain" my comment are based on what is on hand for everyone. The VNWA is actually a full two port devise as when fitted with a testset it swaps the direction so your comment is not entirely correct. It has besides a feature in the software (by pressing the F2 key) to do a full two port measurement, without having a testset attached and that is done by mechanically reverting the DUT between the forward and reverse sweep. So you study of the VNWA manual has not been complete. Kid regards Kurt -----Oprindelig meddelelse----- Fra: [email protected] <[email protected]> P? vegne af Gabriel Tenma White Sendt: 9. december 2019 09:06 Til: [email protected] Emne: Re: [nanovna-users] The T-Check confusion On Mon, Dec 9, 2019 at 12:54 AM, Kurt Poulsen wrote: This is wrong. I double checked the VNWA user guide and it is also a T/R VNA just like the Nano, not a full two port VNA. Therefore any calibration types used on the VNWA also apply to the Nano, it's just a matter of software. What you are describing is the Two Port One Path calibration, which is nothing new and there is an implementation of it in scikit-rf: Since scikit-rf is a Python library it is also trivial to use in nanovna-saver. |
|
Hi Erik
Just one more practical comment. Whatever you do the single forward and the single reverse measurement s1p files does not carry any information about the port mismatch to take into account for the T-Check calculation. And besides that I consider this discussion is only of pure interest how T-Check is done the right way, and my report was only to demonstrate some facts. I do not think 99% of the NanoVNA users are benefitted if such facilities was implemented in future and would not try to motivate such a initiative. This product is great as is and a calibration is tested quite well investing in a semirigid cable of some 25cm both for frequency range up to 500MHz and 900MHz. To 1500MHz is does not show useable trace along the Smithchart circumference. Kind regards Kurt -----Oprindelig meddelelse----- Fra: [email protected] <[email protected]> P? vegne af erik@... Sendt: 9. december 2019 08:57 Til: [email protected] Emne: Re: [nanovna-users] The T-Check confusion Kurt, It appears (in my limited understanding) the poor matching to Z0 of CH1 (port 2) is the biggest reason why the T-check is not performing well on the nanoVNA. Would you agree that using a 10dB attenuator between DUT and CH1 (port 2) would solve most of this? What would be the impact of the remaining non-compensated delay? Should we add to the instructions that, for best nanoVNA 2 port DUT measurements one should seriously consider using a 10dB attenuator between DUT and CH1 (port 2) till new HW becomes available that solves the poor matching of CH1 (port2)? -- NanoVNA Wiki: /g/nanovna-users/wiki/home NanoVNA Files: /g/nanovna-users/files Erik, PD0EK |
|
Hi Erik
Using the transfer switch is just a lazy method to swap the direction of the DUT. Actually dependent of the transfer switch quality you bought you make get poorer isolation between port 1 and port2. Often the isolation in the region of 60dB for such transfer switches. Although you have isolation calibration e.g. in forward direction it will not be the same in reverse direction. With the NanoVNA + softwares you cannot do independent calibration in the forward and reverse direction so it is not helping you, on the contrary. Nothing is better than swapping the DUT direction when measuring forward and reverse. You will have different calibration result via the transfer switch for forward and reverse. I did in the past a test set with 4pc. 18GHz SMA microwave relays and used rigid interconnection cable of same phase length. I had isolation up to 140dB with this setup, which was made for my N2PH VNA going to 60MHz. The N2PK VNA was controlled by the VNWA software, which both has separate calibration in forward and reverse direction and did automatic controlled the test set. This way it work also with a transfer relay which I also have works the same way. So bottom line.. You may develop an interface for you homemade VNA controlling the transfer relay and implement the needed error correction in software. There are no free lunches in this game. I have no idea if the TAPR VNA software have such facilities. To the general question about the hobby level then by adding a 10dB attenuator (included in the calibration) between DUT and RX port linearize the measurements quite a lot, so that would be enough in daily use for the accuracy you are after. Kind regards Kurt -----Oprindelig meddelelse----- Fra: [email protected] <[email protected]> P? vegne af erik@... Sendt: 9. december 2019 12:47 Til: [email protected] Emne: Re: [nanovna-users] The T-Check confusion I've found and bought an affordable SMA transfer switch (2x2 switch) on eBay and I was hoping it would allow me to extend nanoVNA or one of my home build VNA's to full two port capability by extending the calibration routines in the TAPR VNA software Do I understand correctly that, as you are not able to measure all errors using a transfer switch connected to port 1 and port 2, you can only do the TwoPortOnePath calibration? What would be the impact of this in practice???? Given I can measure and ensure both ports are rather close to a real 50 ohm using a second VNA: would the inability to correct some errors have a major impact on practical measurements of filters that are sensitive to port mismatch? Remember for me this is all hobby use and being able to measure with an error less then some/one dB is great -- NanoVNA Wiki: /g/nanovna-users/wiki/home NanoVNA Files: /g/nanovna-users/files Erik, PD0EK |
|
Hi Erik
Fantastic ... I did not know you was maintaining TAPR VNA (or I had forgotten it) Looking forward to follow your project. My transfer switch is also Sivers Lab but only PM7551 ? with a small control PCB with 5V to 28DC converter. It went bust recently so now supplied by good old fashion AC/DC converter. I did measure some WiFi channel filter with the VNWA successfully using a filter for the entire WiFi band to clean up the TX signal. That worked great. One og these days I will try it with the NanoVNA Kind regards Kurt -----Oprindelig meddelelse----- Fra: [email protected] <[email protected]> P? vegne af erik@... Sendt: 9. december 2019 14:30 Til: [email protected] Emne: Re: [nanovna-users] The T-Check confusion Hi Kurt. Thanks. This is the switch I bought: SIVERS LAB PHILIPS COAXIAL SWITCH 28V DC-18Ghz PM7553 Isolation below 1GHz is better than 80dB so I guess below 300MHz good enough. As I am currently the maintainer for the TAPR-VNA software so I can add whatever I want :-) Using the scikit-rf implementation as example I should be able to add (limited) 10/12 term calibration functionally to the TAPR-VNA SW, including controlling the transfer relay. Lately I have been trying to optimize and tune various filters such as a 2GHz cavity filter and as I have to tune 5 resonators it would be nice if I did not need to reverse the DUT all the time as for best tuning you need to see S11,S21,S22 and S12 together. So it is more a matter of being kind to SMA cables and connectors and making tuning faster and easier. For my home build VNA's I can also buy/build a second bridge, add a 4th receiver and put the transfer switch between the LO and the bridges. This will give me a fairly symmetrical setup allowing measurement of the errors in the bridge impedance. Disadvantage is the transfer switch is locked inside one VNA So many options to choose from, sounds like a hobby -- NanoVNA Wiki: </g/nanovna-users/wiki/home> /g/nanovna-users/wiki/home NanoVNA Files: </g/nanovna-users/files> /g/nanovna-users/files Erik, PD0EK |
|
On Mon, 9 Dec 2019 at 07:56, <erik@...> wrote:
Kurt,I would not say the matching is poor on port 2. I measured it using my HP VNA. At least on the NanoVNA sample I have, the return loss on port 2 is better than the specification on my HP 8720D VNA. I did post some data on the NanoVNA Facebook group. I was unable to measure the source match at port 1 well due to the power coming out of port 1. I had set my HP VNA source power to -20 dBm in an effort to not cause any non-linearities. Testing at +10 dBm might have been possible, but I don't know if the results could be trusted. Certainly switching the NanoVNA off showed excellent match on port 1, and on port 2 is was quite reasonable, and largely independent of whether the unit was powered on or not. Erik, PD0EK Dave |
Kurt Poulsen
Dr. David Kirkby, Kirkby Microwave Ltd