Marcel wrote: ↑Thu Apr 06, 2023 10:37 am
Bird 43 Bridge return loss 80 to 120 MHz.
The old analyzer used is not very suitable for measuring very small losses (Yellow trace) has too much trace noise.
You certainly don't have to try this with a Siglent, Nano, Rigol VNA or an SA with TG as you will only see noise.
There are several things you can do to improve things. Unfortunately I can't see the middle section of the grey bar at the bottom of the screen in the picture well enough to be able to read what you've got your IFBW set to but first you can decrease it. I believe those E5062s go down to 1 Hz.
I haven't got such a nice modern network analyser, I've got a much older HP 8712 and can only select between a few predefined IFBWs.
I connected a short cable between the Reflection (S11) port and the Transmission (S21) port, pressed PRESET and then pressed CAL and did a S21 calibration. Without touching the short cable, I ran a single sweep. With default (preset) settings, the IFBW is set to 4 kHz. Trace as follows:
1.JPG
Not particularly good but probably not bad for a 90s instrument that hasn't been calibrated since the mid-2000s. Note the vertical scale is also set to 50 mdB/div like yours for S21.
Sorry for the crap pictures, the camera doesn't handle the brightness difference between the graticule and trace very well and I can't adjust the graticule brightness relative to the trace on this analyser. And I've lost my GPIB adaptor.
Leaving the short cable in place and reducing the IFBW to 15 Hz, gives the following once another calibration has been run (S21 cal only):
2.JPG
We can improve things further by increasing the number of points. So far, I've been using 201 points. Increasing to 1601 points, and re-running the cal:
3.JPG
You might argue that the 1601-point trace is significantly noisier than the 201-point trace and I'd agree, but you could go and low-pass filter the trace in post-processing.
You could also turn on averaging, which does improve things somewhat. I didn't bother taking a picture of the whole screen as you can't really tell the difference, but what I did to demonstrate that you can see an improvement, albeit approaching diminishing returns, is to save the previous trace to memory and set it to display the memory trace at half brightness behind the active trace. Here's a close-up:
4.JPG
It looks like the trace might be out of focus or have a slight shadow but that's not the case, look at the vertical scale which is quite crisp in comparison.
Of course, running at least 4 averages, plus a cal sweep, at 1601 points and 15 Hz IFBW takes a while!
I also learned something today - if you leave averaging on when you run the cal, it will average however many traces you set for averaging when doing the cal too. I let it run like this for 4 cal traces and at least 8 measurement sweeps - the whole thing took about 15 minutes to sweep, but gave quite an improvement! Comparison below:
5.JPG
There's some absolute drift which is probably thermal drift as the analyser hadn't been on for long. Note that I have expanded the scale to 10 mdB/div in the last image, shown above. I find this quite impressive performance especially since the span is the full analyser range (300 kHz to 1.3 GHz) and not just the FM band.
Another thing I think you can do on the new-fangled analysers (but haven't got one to check) is to do what a lot of spectrum analysers can do and that is peak hold, and some signal analysers can do min. hold as well. The min and max traces would give you error bars and you could say quite comfortably that the reality is somewhere in the middle with a standard distribution.
I'll leave the analyser on sweeping for a few hours and see what happens. With the brightness turned down to the minimum of course!
You do not have the required permissions to view the files attached to this post.
Not finished fully repairing my ELENOS ETG1000 yet... but...
I never had the need for it, but now my little Tugicom TX180 1W exciter board
BUT the Reflected is +/- spot on 
