DIY Short, Open and LOAD for VNA calibration

[Jay_Diddy_B]

Hi group,

I would like to share a quick project with you. This project is more metal work than electronics.
If you have used a network analyzer you will quickly find that you need to have a short, open and load to perform calibrations.

I found myself needing a set of BNC female calibration standards. BNC is not really a good choice for VNA, but if you are working on a project with BNCs there is little choice. Normally the calibration standards are N, SMA, APC 7 or better connectors.

This set seems to work fine at a few hundred MHz.

The BNC connector that I used is Delta Electronics Mfg.Corp 1020-000-N911-1

Here is the drawing:



These will press fit into 0.312 (5/16) hole.

Schematic



It doesn't get much simpler than that.

Connector preparation




The short was constructed by making a brass disc with a small hole and soldering the disc to the back of the connector:






The open was constructed by filing the center pin.

The load was constructed by soldering two 0805 100 Ohm 1% resistors on the connector.

Housing

The housing started as 0.6" length of 0.875" hexagonal bar stock (aluminum 6061):



Three holes were drilled into the center of the block:




Assembly

The connectors were then pressed into the holes:



I gave this a quick test on my HP 3577A (5Hz to 200 MHz) VNA and they work fine.

I need to make a label.

Regards,

Jay_Diddy_B

[AF6LJ]

Nicely done and those should serve you well past 200MNZ.

[DaJMasta]

It's a nice little form factor, but are 1% resistors really sufficient (did you pick them for value beforehand?)

I'd be going with a little tighter tolerance and thin film (as opposed to thick) for a bit less current noise just as a go-to because it's a calibration reference point.  That may not be required, but I sort of would have expect something like.

[Jay_Diddy_B]

It's a nice little form factor, but are 1% resistors really sufficient (did you pick them for value beforehand?)

I'd be going with a little tighter tolerance and thin film (as opposed to thick) for a bit less current noise just as a go-to because it's a calibration reference point.  That may not be required, but I sort of would have expect something like.

The RF properties of the resistors are probably more important than the absolute value.

The load is an impedance standard, not a resistance standard, it should be 50 over a very wide frequency range.

If it was resistive, a 1% error in resistance causes a 46dB return loss, a VSWR of 1.01

This is probably better than the BNC connectors.


 will try and do some measurements later.

I did try and find the accuracy of VNA, but it wasn't easy to find.

Regards,

Jay_Diddy_B

[Mechatrommer]

The RF properties of the resistors are probably more important than the absolute value.
The load is an impedance standard, not a resistance standard, it should be 50 over a very wide frequency range.
If it was resistive, a 1% error in resistance causes a 46dB return loss, a VSWR of 1.01
This is probably better than the BNC connectors.
 will try and do some measurements later.
I did try and find the accuracy of VNA, but it wasn't easy to find.

Regards,
Jay_Diddy_B

i'm ordering a VNA that according to manufacturer is calibrated from factory using their standard calibration OSL kit... a diy OSL kit can be an interesting project, but verifying it we need some convincing reference/standard. with uncalibrated/unknown level VNA, this can be like shooting in the dark. with calibrated VNA, at least we can see the diy OSL respond based on / relative to the factory calibration standard... i think... i'm super newbie in this thing, expecting fun things in the future when i get my VNA...

[G0HZU]

I've got various VNA cal kits here and one of them is a home made BNC kit that looks quite similar to JDP's kit.

Mine was built quick and dirty style but I also use it with my own user cal file and can do full 2 port calibrations with it as well because I have a calibrated BNC thru 'barrel' connector as well. I rarely use the BNC cal kit but it's hard to avoid having one. I'm not sure how well mine actually performs because I only use it for casual stuff. But I used 2x 0805 100R resistors for the load as well. I think I selected a pair of resistors that gave a resistance close to 50 ohms.

I didn't spend much time optimising the 'user' cal file so I think mine could be improved quite a bit.

[DaJMasta]

It's a nice little form factor, but are 1% resistors really sufficient (did you pick them for value beforehand?)

I'd be going with a little tighter tolerance and thin film (as opposed to thick) for a bit less current noise just as a go-to because it's a calibration reference point.  That may not be required, but I sort of would have expect something like.

The RF properties of the resistors are probably more important than the absolute value.

The load is an impedance standard, not a resistance standard, it should be 50 over a very wide frequency range.

If it was resistive, a 1% error in resistance causes a 46dB return loss, a VSWR of 1.01

This is probably better than the BNC connectors.


 will try and do some measurements later.

I did try and find the accuracy of VNA, but it wasn't easy to find.

Regards,

Jay_Diddy_B

Interesting, I sort of expected at <200MHz that maybe the resistance was a dominant enough component, but I don't have any measurements to back that up.  I imagine those RF characteristics of the resistors get to be real important in the multi-GHz region.

When it's all said and done, though, if they're performing well... that's all you need 

[AF6LJ]

It's a nice little form factor, but are 1% resistors really sufficient (did you pick them for value beforehand?)

I'd be going with a little tighter tolerance and thin film (as opposed to thick) for a bit less current noise just as a go-to because it's a calibration reference point.  That may not be required, but I sort of would have expect something like.

The RF properties of the resistors are probably more important than the absolute value.

The load is an impedance standard, not a resistance standard, it should be 50 over a very wide frequency range.

If it was resistive, a 1% error in resistance causes a 46dB return loss, a VSWR of 1.01

This is probably better than the BNC connectors.


 will try and do some measurements later.

I did try and find the accuracy of VNA, but it wasn't easy to find.

Regards,

Jay_Diddy_B

Interesting, I sort of expected at <200MHz that maybe the resistance was a dominant enough component, but I don't have any measurements to back that up.  I imagine those RF characteristics of the resistors get to be real important in the multi-GHz region.

When it's all said and done, though, if they're performing well... that's all you need 

You have to be fairly sloppy to make a load that will be reactive at 200MHZ, or use wire would resistors.
Seriously;
The quality of the resistor does make a difference, we are talking about 200MHZ, and a load with short direct leads. A commercially made 50 ohm termination good to a GHZ can be had for almost nothing, this one will work just fine for it's upper frequency of use. 

[Jay_Diddy_B]

I've got various VNA cal kits here and one of them is a home made BNC kit that looks quite similar to JDP's kit.

Mine was built quick and dirty style but I also use it with my own user cal file and can do full 2 port calibrations with it as well because I have a calibrated BNC thru 'barrel' connector as well. I rarely use the BNC cal kit but it's hard to avoid having one. I'm not sure how well mine actually performs because I only use it for casual stuff. But I used 2x 0805 100R resistors for the load as well. I think I selected a pair of resistors that gave a resistance close to 50 ohms.

I didn't spend much time optimising the 'user' cal file so I think mine could be improved quite a bit.

This is really the point. The BNC standards are 'working' standards. They are not really for precise work. If you doing precision work you would use a different connector.


Test results

It is very difficult to compare two shorts or two opens because the electrical distance from the reference plane changes the properties.

It is relatively easy to compare two loads, because in a impedance matched system the distance doesn't matter.

Equipment

HP3577A with HP35677A S parameter test set.

After doing a full calibration on port 1

This is the short measurement:



And the open measurement:



Here is the measurement of a different load, made by Meca (401 1F3):



Note the scale is 50E-3 units per division (zoom in 20x) The phase difference between the two loads is -0.345 degrees at 100 MHz.


Here is a picture of the return loss:



Regards,

Jay_Diddy_B

[Bud]

m ordering a VNA that according to manufacturer is calibrated from factory using their standard calibration OSL kit... a diy OSL kit can be an interesting project, but verifying it we need some convincing reference/standard. with uncalibrated/unknown level VNA, this can be like shooting in the dark. with calibrated VNA, at least we can see the diy OSL respond based on / relative to the factory calibration standard... i think... i'm super newbie in this thing, expecting fun things in the future when i get my VNA...

You still need your own calibration set. ? factory calibrarion would set the reference plane at the device connector, and it is often needed to move the calibration plane elsewhere which requires recalibration by the user.
Also,  based on that this device is a cheap chinese gimbo, the "factory" calibration is going to be out of wack soon due to ambient changes in temperature, aging of the components, etc. it still may be ok for antenna measurements , which typhcally do not need to be precise, but forget using it for moderately accurate generic measurements unless you get a good calibration set.

[Mechatrommer]

m ordering a VNA that according to manufacturer is calibrated from factory using their standard calibration OSL kit... a diy OSL kit can be an interesting project, but verifying it we need some convincing reference/standard. with uncalibrated/unknown level VNA, this can be like shooting in the dark. with calibrated VNA, at least we can see the diy OSL respond based on / relative to the factory calibration standard... i think... i'm super newbie in this thing, expecting fun things in the future when i get my VNA...

You still need your own calibration set. ? factory calibration would set the reference plane at the device connector

i know. what i meant is, when we build our own OSLT calibration set, we just put it near the "calibrated" device plane and see how its respond relative to factory calibration data, for example if i build a 50ohm load (or through) and snap into the device input, it should indicate (close to) flat line on the reflected power graph (or whatever it is), if not, then our 50ohm load model (or whatever "model" it is) is not good up to the tested frequency. that was my concern when deciding buying used or new VNA. if buying used VNA, we still need an expensive calibration set to ensure the VNA giving reliable result. if buy new, we have some assurance, the VNA will give a flat result out of factory if our diy OSL is good enough. without needing to access to higher calibration standard. anyway, for more assurance, i bought a complete set of the cheapy KC901V incl their standard so at least i know what i will be refering to... http://www.deepace.net/shop/kc901v-ultimate-safety-case/ as i said i'm new in this arena i may talk nonsense... feel free to educate me

[Wolfgang]

Hi,

I built an OSL the same way as you did (I used a Hammond box), and its is usable to at least 500MHz according to my VNA (alle values better than 20dB).
Why not ? You can also make an SMA version, then you can reach approximately 2GHz.

[TimNJ]

Necro post?   

I made my own set today. Just wanted to share another possible way to do it. I like JDB's tri-block the best, but in lieu of doing any real metalwork, I decided to do it as follows:

1. Use Amphenol 112536 female, through-hole, BNC connector
2. For open, short, and load, follow exact same steps as JDB. (I just used 0.2mm copper instead of brass for the short.)
3. Seal it all up with a 3/8" copper tube cap, purchased from Lowe's home improvement in the US. Solder the copper end cap to the base of the BNC connector. I used a good amount of flux, set the iron to about 400C, and flowed just enough solder to make the joint all around. (It tends to flow in the same way as when you use a butane torch for plumbing.)

I'm pretty happy with the end result. My VNA (NanoVNA) does not have BNC connectors, so the factory calibrated reference plane is extended slightly with an adapter. But all looks okay.

[Jay_Diddy_B]

TimNJ,

That is an interesting minimal metalwork solution.

It has been 3 years since I built the OSL from a piece of hexagonal bar. Over the years I have learned to really like having all three standards in my hand at the same time. It makes for very efficient calibrations.

Regards,
Jay_Diddy_B

[tautech]

A couple of good vids on this topic:

[Mechatrommer]

the guy will still need a good and known CAL kit in order to find the correct delay for his DIY kit. and he doesnt have to go to trouble to file the CAL kit to match the delay between Open and Short, they are what they are and the VNA can still give good result if correct delays are keyed in to the VNA, even if they are not matched. and about the Load, it is apparent from the video that whatever delay you put in the Load character, VNA will still shows a good 50 ohm load, this is due to very high return loss and VNA doesnt have enough returned signal to show any differences. this proved our conversation with Mr Kirkby that according to him, the idea of delay on 50 ohm Load is silly. a reasonably good 50 ohm Load will almost always shows good result on VNA. this is also proven with the 50 ohm Load that came with my NanoVNA, that it equally match my 50 ohm Load from Kirkby, but not so with Open and Short. the key to a tracable DIY kit is a good/known/characterized SOL CAL kit at hand in the first place to do comparison and to find the correct delay for DIY Open and Short. ymmv.

[Darko8]

Thank you,

Those videos are really good. I was playing with nanoVNA since it was released and now I'm in a process of buying proper VNA and I need to make better calibration standards.
I decided not to buy N standards, because I'm mostly working with SMA equipped devices, so results would be way off after I attach N to SMA adapter any way. Probably more than they will be off with good homemade ones.

Regards, Darko