I probably need to clarify the plot - x axis is a sweep from 3-30 MHz. Y axis is s21 after I divided by 2 cores - loss per core.
Ariel NY4G
toggle quoted message
Show quoted text
On Mar 17, 2021, at 22:00, Ariel NY4G <ny4g@...> wrote:
?What am I doing wrong? The chart I plotted from the nanoVNA into Excel just does not look right. This is the back to back method. Two cores, secondaries together, grounds together, and measure loss between ports 1 and port 2 (S21) - photo also attached. My calibration looks OK.
Thanks
Ariel NY4G
?On 3/17/21, 3:58 PM, "[email protected] on behalf of jmr via groups.io" <[email protected] on behalf of jmrhzu@...> wrote:
Hi WB2AUQ, that all sounds good now.
I had another play with the s2p file of the (big) stacked FT240 49:1 transformer. This includes the 110pF capacitor at the input. I put the s2p model of the transformer and the input cap back to back and did an insertion loss comparison. I added a x2 factor to the trace for the back to back measurement and the plot is shown below. This shows that in this case the back to back method would be OK across most of the band as the traces agree very closely. However, it is a risky method especially if the network has a lot of reactance across the passband. You can see that this is what happens at the band edges where the traces diverge quite a bit.
Note that this s2p file was not taken with a nanovna. It was taken some time ago using an Agilent VNA and a full 2 port cal using an Aglent N4431-60006B Ecal. It also uses the 'fixture simulator' feature in the Agilent analyser that has a very accurate s2p model of my 2 port text fixture. This corrected fixture system works very well up to many GHz and it can accurately shift the reference plane right to the pins of the device under test so I think the results below can be trusted with high confidence.
<Screen Shot 2021-03-17 at 9.51.45 PM.png>
<Vjxgfv3oTf6FZPv8wmN6Qg.jpg>
ariel.jacala