Author Topic: RF Magic. Using return loss to measure a set of cavity filters VS Transmission L  (Read 1566 times)

0 Members and 1 Guest are viewing this topic.

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
Wrongly posted at Beginners but bear with me.( Don't know how to move topics ) :'(
Dear All ,

I am an RF enthusiast , I have a siglent RF spectrum analyser with the reflection measurements enabled and an external directional coupler from mini circuits.

I have a SET of VHF cavities , with a high Q factor for 0.6mhz channel spacing.

I have a following conundrum.

If I tune with the best of my abilities the filters with Transmission loss because of the nature of the measurement I am unable to reach the noise floor of the instrument ( even with TG power at max 0db in and normalised the cables between them ) if I use after that fine tunning return loss I am seeing something completely different and wrong ( high SWR on the cut frequency and close to 1.5swr on the pass frequency )
IF I use the return loss measurement to align the cavity filters. everything goes perfect at 1 SWR pass and close to infinity the fall frequency BUT if I do the Transmission loss measurement the graph looks like is no way to be correct.

I've being battling with this question for a long while without having a way to "trust" my instrument , more or less trust my testing OR my procedure or even my self and sanity.

I would like to hear your thoughts and hopefully some RF magician can point out the obvious that I am unable to grasp.

Sorry for the long detail.

Sorry if this was answered int he past.

I will try to get some screenshots from my spectrum analyser to describe the issue in a better manner.

Thank you in advance

For instance this is a different cavity filter :
2299725-0
And Picture 2 ABYSMAL Return Loss
2299729-1
 

Offline ftg

  • Regular Contributor
  • *
  • Posts: 186
  • Country: fi
    • ftg's RF hax paeg
What is the model of the Mini-Circuits directional coupler?
And what directivity and coupling do they claim in the datasheet for 430 - 440 MHz range?

You might be hitting the limits of the coupler before those of the spectrum analyzer.

But talking about the measurements, I would expect the notch or stopband frequency to have a really bad match looking in from the port in question.
How is the measurement set up, what port to what port are you measuring and what are you terminating the un-used port with?
 

Online ch_scr

  • Frequent Contributor
  • **
  • Posts: 881
  • Country: de
If the sweep speed is high, your filters might ring like a bell. That could mess up the measurement. The higher the Q, the slower you have to sweep.
Playing with the sweep speed, you should be able to see the curves change, if there is influence from that.
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
Dear Sir ,

The DC is model ZFDC-20-5-N+ 0.1-2000MHZ.

I find it very suspicious to not be proper for this Frequency range.

All unused ports are terminated with 50Ω loads. Also verified that they are ok SWR wise.

The measurements are measured as followed.
On TX loss ( with out the DC ) - TG is plugged on the antenna port and Signal in to the RX port.
On Insertion loss (with DC) - TG is plugged on the OUT port(DC) , Signal in is plugged on the coupled port(DC) and IN port(DC) is plugged on the port you want to measure ( in our case the RX port with other 2 ports with 50Ω loads.

DC= Directional coupler
TX Loss = Transmission loss measurement ( regular spectrum analyser measurement. )

Also all measurements are done with the TG output at 0bdm and the TG and Signal in are normalized to ZERO to cancel out any TX loss in the cable. so in essence all measurements start from the end of the patch cables from the spectrum analyser.

IN Insertion loss measurement ( Reflectometry ) all cables are coupled to the DC including the patch cable from the IN of the DC and Open end that will connect to the DUT and also calibrated with the spectrum analysers Cal open option that zeros out the graph.

Thank you in advance
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
If the sweep speed is high, your filters might ring like a bell. That could mess up the measurement. The higher the Q, the slower you have to sweep.
Playing with the sweep speed, you should be able to see the curves change, if there is influence from that.

The Given cavity filter is a small 8 cavity filter for UHF ( with spacing of RX and TX to 7.6mhz.) These are not the classical "jugs" that are used as a BrBr filter for VHF that the channel spacing is 0.6mhz.

But even at that it shouldn't be a so large difference between measurements.

Thank you in advance
 

Offline Joel_Dunsmore

  • Regular Contributor
  • *
  • Posts: 181
  • Country: us
Usually cavity filters form band-pass responses but yours looks like a low-pass response.  What does it look like if you start at a lower frequency? Is a band pass response.
Further (and no, not just because I'm a VNA guy), one issue is you may not be providing a good load from the SA receiver. If you measure a thru connection (instead of the filter) what is you return loss? Ideally it should be -infinity into a good load, but -40 dB is good, -30 dB is also acceptable.

And, are these filters designed for 50 ohm input and output impedances? or are they designed to be terminated in a high impedance?

do you have a picture or schematic representation of the filter?
 

Offline Emo

  • Regular Contributor
  • *
  • Posts: 133
  • Country: nl
Your resolution bandwidth is very large compared to the range you want to measure. Decreasing it will make the sweep slower and the noise floor substantially lower.
As a second step you could also try to measure in time domain response so you can see what every cavity is tuned to. Not sure if the SA is accurate enough for dough.

Eric
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
Usually cavity filters form band-pass responses but yours looks like a low-pass response.  What does it look like if you start at a lower frequency? Is a band pass response.
Further (and no, not just because I'm a VNA guy), one issue is you may not be providing a good load from the SA receiver. If you measure a thru connection (instead of the filter) what is you return loss? Ideally it should be -infinity into a good load, but -40 dB is good, -30 dB is also acceptable.

And, are these filters designed for 50 ohm input and output impedances? or are they designed to be terminated in a high impedance?

do you have a picture or schematic representation of the filter?

Dear Sir ,

These were typical cavity filters allegedly used for 50Ω impedance radios. High impedance I highly doubt it.

https://renaironline.co.uk/?product=uhf-450-mhz-duplexer

Something like this a small 8 cavity filter. UHF.

while zeroing out the spectrum in reflection mode I get infinity with open port of the DC.
Before zeroing it out I never looked what is the actual grapth.

While testing my 50Ω load I never took the time to test it but is a reputable band 50Ω load ( one was a bird load and the other was racal or something like this )

Thank you in advance

I will test the bigger frequency space(span) to look indeed if its mistuned.
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
Your resolution bandwidth is very large compared to the range you want to measure. Decreasing it will make the sweep slower and the noise floor substantially lower.
As a second step you could also try to measure in time domain response so you can see what every cavity is tuned to. Not sure if the SA is accurate enough for dough.

Eric

Dear Sir ,

Yes I am playing around with the RBW and VBW to try and get lower values BUT my analyser doesn't seem to like it too much , ( an old Anritsu can do 1hz RBW and VBW and get down to -120dbm ( an other expert said so ) mine can't do that.  )
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
 

Offline ftg

  • Regular Contributor
  • *
  • Posts: 186
  • Country: fi
    • ftg's RF hax paeg
Your resolution bandwidth is very large compared to the range you want to measure. Decreasing it will make the sweep slower and the noise floor substantially lower.
As a second step you could also try to measure in time domain response so you can see what every cavity is tuned to. Not sure if the SA is accurate enough for dough.

Eric

Dear Sir ,

Yes I am playing around with the RBW and VBW to try and get lower values BUT my analyser doesn't seem to like it too much , ( an old Anritsu can do 1hz RBW and VBW and get down to -120dbm ( an other expert said so ) mine can't do that.  )

I have a Siglent SSA3032X at the office and I have not experienced this.
Do keep the VBW locked to RBW and keep Sweep Time on auto, so that the sweep time is automatically adjusted to a suitable one.
You could also try adding a max hold or min hold trace in addition to the realtime one.
It will help when you have really slow sweep times to get low noise floors.
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
Your resolution bandwidth is very large compared to the range you want to measure. Decreasing it will make the sweep slower and the noise floor substantially lower.
As a second step you could also try to measure in time domain response so you can see what every cavity is tuned to. Not sure if the SA is accurate enough for dough.

Eric

Dear Sir ,

Yes I am playing around with the RBW and VBW to try and get lower values BUT my analyser doesn't seem to like it too much , ( an old Anritsu can do 1hz RBW and VBW and get down to -120dbm ( an other expert said so ) mine can't do that.  )

I have a Siglent SSA3032X at the office and I have not experienced this.
Do keep the VBW locked to RBW and keep Sweep Time on auto, so that the sweep time is automatically adjusted to a suitable one.
You could also try adding a max hold or min hold trace in addition to the realtime one.
It will help when you have really slow sweep times to get low noise floors.

Dear Sir,

I need to test it , keep the VBW on auto , make RBW 1Hz and keep eveything else as I usually do ( 0db TG -10db attenuator  and normalized cables )

Thanks

Alex
 

Offline Harry_22

  • Frequent Contributor
  • **
  • Posts: 586
  • Country: 00
Hi!
It would be very interesting to know what set of cavity filters you configure? What is your task to decouple the receiver and transmission or organize the operation of several transmitters on one antenna? Can you attach the scheme and photos of the filters themselves?
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
Your resolution bandwidth is very large compared to the range you want to measure. Decreasing it will make the sweep slower and the noise floor substantially lower.
As a second step you could also try to measure in time domain response so you can see what every cavity is tuned to. Not sure if the SA is accurate enough for dough.

Eric

Dear Sir ,

I tried again today. With the pre-set settings.
TG output at 0dbm I was unable to set RBW to 1HZ the erro said timescale out of range ( I guess having to sweep the 1mhz I set as a span was too much for it )
BUT even at zero span same story.
Thanks
Dear Sir ,

Yes I am playing around with the RBW and VBW to try and get lower values BUT my analyser doesn't seem to like it too much , ( an old Anritsu can do 1hz RBW and VBW and get down to -120dbm ( an other expert said so ) mine can't do that.  )

I have a Siglent SSA3032X at the office and I have not experienced this.
Do keep the VBW locked to RBW and keep Sweep Time on auto, so that the sweep time is automatically adjusted to a suitable one.
You could also try adding a max hold or min hold trace in addition to the realtime one.
It will help when you have really slow sweep times to get low noise floors.
 

Offline The_Spectrum.A_idiotTopic starter

  • Regular Contributor
  • *
  • Posts: 97
  • Country: cz
Hi!
It would be very interesting to know what set of cavity filters you configure? What is your task to decouple the receiver and transmission or organize the operation of several transmitters on one antenna? Can you attach the scheme and photos of the filters themselves?

Dear Sir ,

Well they are BrBr duplexer. Indeed separating TX and RX. (exactly as you said)
One in VHF and one in UHF.

Photos at the moment I can't provide. The Filters are already installed.
Thanks
 

Offline Harry_22

  • Frequent Contributor
  • **
  • Posts: 586
  • Country: 00
Yes this is VHF/UHF mobile duplexer. It is based on spiral resonator that is equal to series oscillatory circuit.
To get the low RX/TX space we need high Q value. High Q we get with large L and low C. That is why the L is long thick wire spiral but C is a space between screw end. The series oscillatory circuit is for reject perposes.

The pass ways are formed by additional L and C. It is additional L and C that determine RX/TX space.
All resonators are connected via clitical length cables that calculated for the required bandwith.

Suggested adjustment method is following:
Connect VNA between High and Low ports of duplexer. Antenna input may be terminated by 50 Ohm.
Adjust screws at High section to get minimum at lower frequency. The same for the Low section. We adjust minimum at higher frequency.

At the final adjustment we connect VNA as a transmitter and try to get lowest VSWR with accepted rejection.
This rejection is needed to supress transmitter noise at Rx frequency that pass to receiver and reduce sencitivity. It should be more than 60 dB.

The same for receiver path. We connect VNA from Antenna to RX port and try to minimize VSWR with accepted rejection level. This is to prevent receiver from high strength transmitter signal. This level depends on receiver input filter.  Usually it should be more than 70 dB.

You can also find the holes near the cable entry. Via this hole we bend L in low pass section for final adjustment (Optionally).

Please see my drawings of the mobile duplexer in action.
The picture is the best explanation.


 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf