drkirkby
Full Member level 6
I wanted to reply to this thread
https://www.edaboard.com/showthread.php?140944-VNA-accuracy-without-a-calibration-kit
but it is closed, so I can't. I don't really see the point in closing threads, since if someone can add something useful, even years later, why not let it be added?
Speaking of short/open/load/thru (SOLT) calibration, someone wrote
"The most limited in frequency from all these 4 elements is the Load element, because you do not have the special volume-type resistors (very low parasitic inductance) used by Agilent."
Having worked on the topic of VNA calibration kits for several years, and running a company that specializes in this area
https://www.kirkbymicrowave.co.uk/
I know this is incorrect. There are a number of relatively inexpensive commercial SMA loads, with reasonable return losses to a few GHz. To keep costs down, we buy in loads from Minicirucits, Rosenberger etc, and select those with the return loss we want (> 32 dB at 7 GHz). Not all make that, so we don't use all of them, but for non-critical work, these loads will be okay.
The tricky bits are the opens shorts and to a lesser extent the thrus. Their the design is far from trivial. There are tons of web pages on the subject of making your own VNA cal kits, almost all of which say how important it is to make the short standard as physically short as possible. One author calls such a short his 'dream short'. Yet despite this, a zero length short is never optimal. If you look at some of the older calibration kits by HP/Agilent, such as the obsolete 6 GHz 85032B type-N kit, you will see the female short has a delay of 93 fs (0.093 ps).
**broken link removed** (standard #6)
Yet on Keysight's most expensive type-N kit 18 GHz kit (85054B), the female short has a delay of 27.99 ps!
**broken link removed** (standard #10)
Clearly, if Keysight could produce a short with a delay of 0.093 ps a few decades ago, and this was so important, they would not be using a delay on the female short 300x larger in their most expensive N kit!!!
A more recent 9 GHz kit from Keysight, the 9 GHz 85032F (not B), the female short has an even longer delay at 45.955 ps.
Needless to say, there's a lot of rubbish written on the internet about making your own calibration kits. If anyone starts by saying the short calibration needs to have the shortest possible delay, then you should immediately realize they don't understand the topic.
https://www.edaboard.com/showthread.php?140944-VNA-accuracy-without-a-calibration-kit
but it is closed, so I can't. I don't really see the point in closing threads, since if someone can add something useful, even years later, why not let it be added?
Speaking of short/open/load/thru (SOLT) calibration, someone wrote
"The most limited in frequency from all these 4 elements is the Load element, because you do not have the special volume-type resistors (very low parasitic inductance) used by Agilent."
Having worked on the topic of VNA calibration kits for several years, and running a company that specializes in this area
https://www.kirkbymicrowave.co.uk/
I know this is incorrect. There are a number of relatively inexpensive commercial SMA loads, with reasonable return losses to a few GHz. To keep costs down, we buy in loads from Minicirucits, Rosenberger etc, and select those with the return loss we want (> 32 dB at 7 GHz). Not all make that, so we don't use all of them, but for non-critical work, these loads will be okay.
The tricky bits are the opens shorts and to a lesser extent the thrus. Their the design is far from trivial. There are tons of web pages on the subject of making your own VNA cal kits, almost all of which say how important it is to make the short standard as physically short as possible. One author calls such a short his 'dream short'. Yet despite this, a zero length short is never optimal. If you look at some of the older calibration kits by HP/Agilent, such as the obsolete 6 GHz 85032B type-N kit, you will see the female short has a delay of 93 fs (0.093 ps).
**broken link removed** (standard #6)
Yet on Keysight's most expensive type-N kit 18 GHz kit (85054B), the female short has a delay of 27.99 ps!
**broken link removed** (standard #10)
Clearly, if Keysight could produce a short with a delay of 0.093 ps a few decades ago, and this was so important, they would not be using a delay on the female short 300x larger in their most expensive N kit!!!
A more recent 9 GHz kit from Keysight, the 9 GHz 85032F (not B), the female short has an even longer delay at 45.955 ps.
Needless to say, there's a lot of rubbish written on the internet about making your own calibration kits. If anyone starts by saying the short calibration needs to have the shortest possible delay, then you should immediately realize they don't understand the topic.