Topics

setting up an interdigital filter 23cm


G0FVI
 
Edited

Morning all,
I have built a three pole interdigital filter as described on W6PQL's website (http://www.w6pql.com/23cm_filters.htm ). Unfortunately my VNA is only 1 port which means it's going to be a faff setting up.  Just wondering what sort of return loss I should be aiming for? A quick fiddle last night gave me around 50db at 1296.2 as shown in pic. 

Andy G0FVI

 


Andy G4JNT
 
Edited

A one port is all you NEED for tuning up a filter.  You tune up based on return loss   -  Dishal's Method

Connect NA to one end and terminate the other;   detune all stages
Starting at the stage nearest to the NA, tune for a dip in return loss
2nd resonator, watch as second dip appears and centralise the pair ;  a PEAK in return loss
3rd resonator, tune and watch three dips appear, place the middle one at the centre frequency
and so on...  Alternating a new dip with a peak
You only need a 2-port NA when you're finished and want to check the final response, measure loss etc.   After a bit of experience, you don't even need to do that - the positions of the poles will tell you most of what you need to know about the response shape, an floss can be measured in other ways/

That's my description of Dishal - others vary slightly but all use the same idea; placing dips, or poles, in the right place.

Another description

Google "Dishal's Method"   for others



On Tue, 29 Sep 2020 at 11:03, G0FVI via groups.io <Andrew.gilfillan=ntlworld.com@groups.io> wrote:
Morning all,
I have built a three pole interdigital filter as described on W6PQL's website (http://www.w6pql.com/23cm_filters.htm ). Unfortunately my VNA is only 1 port which means it's going to be a faff setting up.  Just wondering what sort of return loss I should be aiming for? A quick fiddle last night gave me around 50db at 1296.2 as shown in pic. 

Andy G0FVI

 


Andy G4JNT
 
Edited

one addendum :

You do need a two-port NA to set up elliptic filters since an important part of the process involves tuning the zeros (nulls in the stopband) to their right settings.    If you're setting up an interdigital, it won't be an elliptic

As an aside - who realises that to get an elliptic response you have to have coupling between non-adjacent resonators?     

A dual mode circular waveguide cavity filter is perfect for this, if you go through the filter in one direction with one polarisation then back with the other polarisation, using 45 degree coupling screws to set the coupling between the two modes .   Within each cavity you couple cavity 1 to cavity N,   cavity 2 to cavity N-1  etc.  Just what is needed for an elliptic response

I once got offered a job at Filtronics just for knowing that, when chatting casually to a Radio Am who just happened to work for that company and had just such a filter in his hand at the time.  He'd being interviewing for the job only an hour before and on asking the candidate what it was, he didn't know !  

Turned down my job offer, didn't want to have to move to Yorkshire.




On Tue, 29 Sep 2020 at 11:13, Andy Talbot <andy.g4jnt@...> wrote:
A one port is all you NEED for tuning up a filter.  You tune up based on return loss   -  Dishal's Method

Connect NA to one end and terminate the other;   detune all stages
Starting at the stage nearest to the NA, tune for a dip in return loss
2nd resonator, watch as second dip appears and centralise the pair ;  a PEAK in return loss
3rd resonator, tune and watch three dips appear, place the middle one at the centre frequency
and so on...  Alternating a new dip with a peak
You only need a 2-port NA when you're finished and want to check the final response, measure loss etc.   After a bit of experience, you don't even need to do that - the positions of the poles will tell you most of what you need to know about the response shape, an floss can be measured in other ways/

That's my description of Dishal - others vary slightly but all use the same idea; placing dips, or poles, in the right place.

Another description

Google "Dishal's Method"   for others



On Tue, 29 Sep 2020 at 11:03, G0FVI via groups.io <Andrew.gilfillan=ntlworld.com@groups.io> wrote:
Morning all,
I have built a three pole interdigital filter as described on W6PQL's website (http://www.w6pql.com/23cm_filters.htm ). Unfortunately my VNA is only 1 port which means it's going to be a faff setting up.  Just wondering what sort of return loss I should be aiming for? A quick fiddle last night gave me around 50db at 1296.2 as shown in pic. 

Andy G0FVI

 


Andy
 


offered a job at Filtronics

I was taught by Prof. Rhodes at Leeds Uni some considerable time ago.  He had a huge success with Filtronic and his other companies and so I must be one of his few failures ! :-)

Andy
(the other one)


From: UKMicrowaves@groups.io <UKMicrowaves@groups.io> on behalf of Andy G4JNT <andy.g4jnt@...>
Sent: Tuesday, September 29, 2020 3:26 AM
To: UK Microwaves groups.io <UKMicrowaves@groups.io>
Subject: Re: [UKMicrowaves] setting up an interdigital filter 23cm
 
one addendum :

You do need a two-port NA to set up elliptic filters since an important part of the process involves tuning the zeros (nulls in the stopband) to their right settings.    If you're setting up an interdigital, it won't be an elliptic

As an aside - who realises that to get an elliptic response you have to have coupling between non-adjacent resonators?     

A dual mode circular waveguide cavity filter is perfect for this, if you go through the filter in one direction with one polarisation then back with the other polarisation, using 45 degree coupling screws to set the coupling between the two modes .   Within each cavity you couple cavity 1 to cavity N,   cavity 2 to cavity N-1  etc.  Just what is needed for an elliptic response

I once got offered a job at Filtronics just for knowing that, when chatting casually to a Radio Am who just happened to work for that company and had just such a filter in his hand at the time.  He'd being interviewing for the job only an hour before and on asking the candidate what it was, he didn't know !  

Turned down my job offer, didn't want to have to move to Yorkshire.




On Tue, 29 Sep 2020 at 11:13, Andy Talbot <andy.g4jnt@...> wrote:
A one port is all you NEED for tuning up a filter.  You tune up based on return loss   -  Dishal's Method

Connect NA to one end and terminate the other;   detune all stages
Starting at the stage nearest to the NA, tune for a dip in return loss
2nd resonator, watch as second dip appears and centralise the pair ;  a PEAK in return loss
3rd resonator, tune and watch three dips appear, place the middle one at the centre frequency
and so on...  Alternating a new dip with a peak
You only need a 2-port NA when you're finished and want to check the final response, measure loss etc.   After a bit of experience, you don't even need to do that - the positions of the poles will tell you most of what you need to know about the response shape, an floss can be measured in other ways/

That's my description of Dishal - others vary slightly but all use the same idea; placing dips, or poles, in the right place.

Another description

Google "Dishal's Method"   for others



On Tue, 29 Sep 2020 at 11:03, G0FVI via groups.io <Andrew.gilfillan=ntlworld.com@groups.io> wrote:
Morning all,
I have built a three pole interdigital filter as described on W6PQL's website (http://www.w6pql.com/23cm_filters.htm ). Unfortunately my VNA is only 1 port which means it's going to be a faff setting up.  Just wondering what sort of return loss I should be aiming for? A quick fiddle last night gave me around 50db at 1296.2 as shown in pic. 

Andy G0FVI

 


Robin Szemeti - G1YFG
 

Didn't want to move to Yorkshire?  Unless you have a Yorkshire passport or can pass the entrance exam (which includes being able to name the entire cricket team, including substitutes, for any given year) it may not have even been an option.

On Tue, 29 Sep 2020 at 11:27, Andy G4JNT <andy.g4jnt@...> wrote:
one addendum :

You do need a two-port NA to set up elliptic filters since an important part of the process involves tuning the zeros (nulls in the stopband) to their right settings.    If you're setting up an interdigital, it won't be an elliptic

As an aside - who realises that to get an elliptic response you have to have coupling between non-adjacent resonators?     

A dual mode circular waveguide cavity filter is perfect for this, if you go through the filter in one direction with one polarisation then back with the other polarisation, using 45 degree coupling screws to set the coupling between the two modes .   Within each cavity you couple cavity 1 to cavity N,   cavity 2 to cavity N-1  etc.  Just what is needed for an elliptic response

I once got offered a job at Filtronics just for knowing that, when chatting casually to a Radio Am who just happened to work for that company and had just such a filter in his hand at the time.  He'd being interviewing for the job only an hour before and on asking the candidate what it was, he didn't know !  

Turned down my job offer, didn't want to have to move to Yorkshire.




On Tue, 29 Sep 2020 at 11:13, Andy Talbot <andy.g4jnt@...> wrote:
A one port is all you NEED for tuning up a filter.  You tune up based on return loss   -  Dishal's Method

Connect NA to one end and terminate the other;   detune all stages
Starting at the stage nearest to the NA, tune for a dip in return loss
2nd resonator, watch as second dip appears and centralise the pair ;  a PEAK in return loss
3rd resonator, tune and watch three dips appear, place the middle one at the centre frequency
and so on...  Alternating a new dip with a peak
You only need a 2-port NA when you're finished and want to check the final response, measure loss etc.   After a bit of experience, you don't even need to do that - the positions of the poles will tell you most of what you need to know about the response shape, an floss can be measured in other ways/

That's my description of Dishal - others vary slightly but all use the same idea; placing dips, or poles, in the right place.

Another description

Google "Dishal's Method"   for others



On Tue, 29 Sep 2020 at 11:03, G0FVI via groups.io <Andrew.gilfillan=ntlworld.com@groups.io> wrote:
Morning all,
I have built a three pole interdigital filter as described on W6PQL's website (http://www.w6pql.com/23cm_filters.htm ). Unfortunately my VNA is only 1 port which means it's going to be a faff setting up.  Just wondering what sort of return loss I should be aiming for? A quick fiddle last night gave me around 50db at 1296.2 as shown in pic. 

Andy G0FVI

 


--
Robin Szemeti - G1YFG


 

Yes, as Andy beat me to it, set it up looking at the return loss.
I would say
1) You don't need to concern yourself with anything greater than 30 dB return loss - it is meaningless in terms of any practical application when connected to an amateur antenna, amplifier, even adapter or cable.
2) But, this type of filter is normally intended to have an "equal ripple" or Chebyshev response in which case you should be seeing three dips in the return loss, like this.


3) Having said that, you have a good match at the intended frequency, so, except for internal losses, the power must be coming out the other end. Just might be a rather "peaky" response rather than having a flat top.


Mark GM4ISM
 

 Harder without  through response but  actually common practice. For a 3 pole filters you should see  a wider response with  3 peaks evenly spaced and of approx same amplitude over the expected bandwidth of the filter. In polar  this would look like a cloverleaf pattern  surrounding the centre of the chart

 You should be able to get  about 20dB worst case return loss on the 3 peaks within the passband, (good 50 Ohm term),  that would usually equate to less than 1db of insertion loss and a nice symmetrical response. I do 5 pole filters this way often.

We have  loads of 6 pole filters that are cross coupled  cavity 2 to 5 if i recall  also some 8 pole versions, they are a  pain to tune especially as you usually have to make 2 identical, (amplitude and phase responses) for use in a Constant impedance filter / combiner

Mark GM4ISM

On 29/09/2020 11:03, G0FVI via groups.io wrote:
Morning all,
I have built a three pole interdigital filter as described on W6PQL's website (http://www.w6pql.com/23cm_filters.htm ). Unfortunately my VNA is only 1 port which means it's going to be a faff setting up.  Just wondering what sort of return loss I should be aiming for? A quick fiddle last night gave me around 50db at 1296.2 as shown in pic. 

Andy G0FVI

 

Virus-free. www.avg.com


Mark GM4ISM
 
Edited

Ken beat me to it and got it right its N dips and N -1 peaks of course  fro an N Pole filter

On 29/09/2020 11:50, Ken G3YKI wrote:
Yes, as Andy beat me to it, set it up looking at the return loss.
I would say
1) You don't need to concern yourself with anything greater than 30 dB return loss - it is meaningless in terms of any practical application when connected to an amateur antenna, amplifier, even adapter or cable.
2) But, this type of filter is normally intended to have an "equal ripple" or Chebyshev response in which case you should be seeing three dips in the return loss, like this.


3) Having said that, you have a good match at the intended frequency, so, except for internal losses, the power must be coming out the other end. Just might be a rather "peaky" response rather than having a flat top.

Virus-free. www.avg.com


Andy G4JNT
 


On Tue, 29 Sep 2020 at 11:45, Robin Szemeti - G1YFG <robin@...> wrote:
Didn't want to move to Yorkshire?  Unless you have a Yorkshire passport or can pass the entrance exam (which includes being able to name the entire cricket team, including substitutes, for any given year) it may not have even been an option.
 
I think my Birminham passport may still be valid, unless they expire before 44 years.  (Even still have a faintly detectable bit of the accent, so I'm told, occasionally).   
That's half-way up there there,  does it allow  visa-free entry to Yorkshire?

"Cricket",  what's that.   Thought it was an insect that makes chirping noises on quiet summer evenings.

 


 

Hi Andy, 
Having looked at your original link, I suspect that the design is not a three pole filter, but is a two pole filter. The middle screw is not a resonator, but will affect the coupling between the two resonators. You will never get the 3 dips I showed, only two.
Another comment would be that a diecast box does not generally have very consistent contact between the lid and box all the way round. If you filter is unduly affected by squeezing it, it may be leaking and would benefit from a conductive gasket or adhesive metal tape over the join.
Hope that helps.


Chris Bartram G4DGU
 

Just moving this discussion on a little, filter theory continues to progress, particularly with the availability of computer based electromagnetic filter synthesis. People are now using this to make filters with much faster passband - stopband transitions and to place arbitary transmission zeroes and poles. Even the spurious responses of resonators in microwave filters are being employed to improve performance. Out of a continuing interest going back to the days of Zverev,  MYJ etc., I've been trying (and often failing!) to keep up with this and it's quite tough going. In the last couple of decades, filter design has moved from something most of us doing RF/microwave design were comfortable with using design tables, to a specialisation of its own.

One area which is rapidly creeping-up is the application of photonic filtering at microwaves.

Progress isn't only measured in ones and zeroes!

73

Chris

G4DGU


Dominique Dehays
 

Hi

What is not a progres is measuring 50db rl or more like in last dubus and believe its true
73
Dom


Robin Szemeti - G1YFG
 

The return loss is almost certainly as described, modern instruments have good dynamic range ... what is probably questionable is whether the load on the end is truly 50 ohms, it might be a -50dB match, but possibly not against a real 50R. That said, when in use, it is unlikely to be fed from or into a true 50R either.

On Tue, 29 Sep 2020 at 12:33, Dominique Dehays <f6dro@...> wrote:
Hi

What is not a progres is measuring 50db rl or more like in last dubus and believe its true
73
Dom


--
Robin Szemeti - G1YFG


Dominique Dehays
 

Hi

Dynamic range is ok , but directivity , even with a good calibration kit can not be high enough for such a rl
73
Dom

-------- Message d'origine --------
De : Robin Szemeti - G1YFG <robin@...>
Date : 29/09/2020 13:41 (GMT+01:00)
À : UKMicrowaves@groups.io
Objet : Re: [UKMicrowaves] setting up an interdigital filter 23cm

The return loss is almost certainly as described, modern instruments have good dynamic range ... what is probably questionable is whether the load on the end is truly 50 ohms, it might be a -50dB match, but possibly not against a real 50R. That said, when in use, it is unlikely to be fed from or into a true 50R either.

On Tue, 29 Sep 2020 at 12:33, Dominique Dehays <f6dro@...> wrote:
Hi

What is not a progres is measuring 50db rl or more like in last dubus and believe its true
73
Dom


--
Robin Szemeti - G1YFG


Andy G4JNT
 

In terms of filter hardware, I wonder where High Temperature Superconducting, HTS, filters went.
Just as I was leaving Marconi Space (as it was then) in 1992 we had started a contract with {I forget who} to look at microwave filters made from HTS   running at the liquid nitrogen temperature of 77K.  These were aimed at reducing the weight of spacecraft payloads to remove those very same waveguide coaxial cavity filters just mentioned.   
We had a few breadboards that worked,-ish, but the real killer was direct radiation from the substrate, as it used microstrip technology, that led to loss and uncontrolled cross-resonator coupling.

IIRC they were talking about moving to enclosed stripline to minimise radiation, but the complications of zero tweakability of that medium caused the project to slow.  
Then I left for a change of direction of at MoD.   Anyone here know where it went, or ever heard of it?



On Tue, 29 Sep 2020 at 12:26, Chris Bartram G4DGU <chris@...> wrote:
Just moving this discussion on a little, filter theory continues to
progress, particularly with the availability of computer based
electromagnetic filter synthesis. People are now using this to make
filters with much faster passband - stopband transitions and to place
arbitary transmission zeroes and poles. Even the spurious responses of
resonators in microwave filters are being employed to improve
performance. Out of a continuing interest going back to the days of
Zverev,  MYJ etc., I've been trying (and often failing!) to keep up with
this and it's quite tough going. In the last couple of decades, filter
design has moved from something most of us doing RF/microwave design
were comfortable with using design tables, to a specialisation of its own.

One area which is rapidly creeping-up is the application of photonic
filtering at microwaves.

Progress isn't only measured in ones and zeroes!

73

Chris

G4DGU







F6DZK Michel
 

There are many sources of uncertainty, (bridge or coupler directivity and many more errors).
It's quite hard to measure -25dB RL with good accuracy, so -40, -50 etc makes no sense.
Dynamic range plays a more important role for transmission measurement.
73 Michel F6DZK


Le mar. 29 sept. 2020 à 13:41, Robin Szemeti - G1YFG <robin@...> a écrit :
The return loss is almost certainly as described, modern instruments have good dynamic range ... what is probably questionable is whether the load on the end is truly 50 ohms, it might be a -50dB match, but possibly not against a real 50R. That said, when in use, it is unlikely to be fed from or into a true 50R either.

On Tue, 29 Sep 2020 at 12:33, Dominique Dehays <f6dro@...> wrote:
Hi

What is not a progres is measuring 50db rl or more like in last dubus and believe its true
73
Dom


--
Robin Szemeti - G1YFG


Chris Bartram G4DGU
 

I agree, Dom.

Even with a good network analyzer, and more importantly, a set of suitably accurate calibration standards, you'd have to be very careful to make a real measurement of  much better than about 35dB!

It's too easy to become mesmerised by apparently good numbers when making measurements. Scepticism is always required.

73

Chris

G4DGU

On 29/09/2020 12:43, Dominique Dehays wrote:
Hi

Dynamic range is ok , but directivity , even with a good calibration kit can not be high enough for such a rl
73
Dom

-------- Message d'origine --------
De : Robin Szemeti - G1YFG <robin@...>
Date : 29/09/2020 13:41 (GMT+01:00)
Objet : Re: [UKMicrowaves] setting up an interdigital filter 23cm

The return loss is almost certainly as described, modern instruments have good dynamic range ... what is probably questionable is whether the load on the end is truly 50 ohms, it might be a -50dB match, but possibly not against a real 50R. That said, when in use, it is unlikely to be fed from or into a true 50R either.

On Tue, 29 Sep 2020 at 12:33, Dominique Dehays <f6dro@...> wrote:
Hi

What is not a progres is measuring 50db rl or more like in last dubus and believe its true
73
Dom

--
Robin Szemeti - G1YFG


Robin Szemeti - G1YFG
 

On antennas, I go with anything better than -20dB is good.  100 watts up, 1 watt back (if you subscribe to the "reflected power" notion) sounds good enough for most porpoises.


On Tue, 29 Sep 2020 at 12:58, Chris Bartram G4DGU <chris@...> wrote:

I agree, Dom.

Even with a good network analyzer, and more importantly, a set of suitably accurate calibration standards, you'd have to be very careful to make a real measurement of  much better than about 35dB!

It's too easy to become mesmerised by apparently good numbers when making measurements. Scepticism is always required.

73

Chris

G4DGU

On 29/09/2020 12:43, Dominique Dehays wrote:
Hi

Dynamic range is ok , but directivity , even with a good calibration kit can not be high enough for such a rl
73
Dom

-------- Message d'origine --------
De : Robin Szemeti - G1YFG <robin@...>
Date : 29/09/2020 13:41 (GMT+01:00)
Objet : Re: [UKMicrowaves] setting up an interdigital filter 23cm

The return loss is almost certainly as described, modern instruments have good dynamic range ... what is probably questionable is whether the load on the end is truly 50 ohms, it might be a -50dB match, but possibly not against a real 50R. That said, when in use, it is unlikely to be fed from or into a true 50R either.

On Tue, 29 Sep 2020 at 12:33, Dominique Dehays <f6dro@...> wrote:
Hi

What is not a progres is measuring 50db rl or more like in last dubus and believe its true
73
Dom

--
Robin Szemeti - G1YFG


--
Robin Szemeti - G1YFG


F6DZK Michel
 

Agree, -20dB is what is easily measurable.
You already need -35dB directivity for that !
-25dB is probably the limit.
73 Michel F6DZK


Le mar. 29 sept. 2020 à 14:00, Robin Szemeti - G1YFG <robin@...> a écrit :
On antennas, I go with anything better than -20dB is good.  100 watts up, 1 watt back (if you subscribe to the "reflected power" notion) sounds good enough for most porpoises.


On Tue, 29 Sep 2020 at 12:58, Chris Bartram G4DGU <chris@...> wrote:

I agree, Dom.

Even with a good network analyzer, and more importantly, a set of suitably accurate calibration standards, you'd have to be very careful to make a real measurement of  much better than about 35dB!

It's too easy to become mesmerised by apparently good numbers when making measurements. Scepticism is always required.

73

Chris

G4DGU

On 29/09/2020 12:43, Dominique Dehays wrote:
Hi

Dynamic range is ok , but directivity , even with a good calibration kit can not be high enough for such a rl
73
Dom

-------- Message d'origine --------
De : Robin Szemeti - G1YFG <robin@...>
Date : 29/09/2020 13:41 (GMT+01:00)
Objet : Re: [UKMicrowaves] setting up an interdigital filter 23cm

The return loss is almost certainly as described, modern instruments have good dynamic range ... what is probably questionable is whether the load on the end is truly 50 ohms, it might be a -50dB match, but possibly not against a real 50R. That said, when in use, it is unlikely to be fed from or into a true 50R either.

On Tue, 29 Sep 2020 at 12:33, Dominique Dehays <f6dro@...> wrote:
Hi

What is not a progres is measuring 50db rl or more like in last dubus and believe its true
73
Dom

--
Robin Szemeti - G1YFG


--
Robin Szemeti - G1YFG


John Lemay
 

An interesting discussion !

 

Just want to clarify my understanding …… If a termination has a measured RL of 35dB, and then the filter is placed between the analyser and that termination, and the analyser then shows 50dB RL …… the filter behaves as a matching device. Is that correct ?

 

John G4ZTR

 

From: UKMicrowaves@groups.io [mailto:UKMicrowaves@groups.io] On Behalf Of Chris Bartram G4DGU
Sent: 29 September 2020 12:58
To: UKMicrowaves@groups.io
Subject: Re: [UKMicrowaves] setting up an interdigital filter 23cm

 

I agree, Dom.

Even with a good network analyzer, and more importantly, a set of suitably accurate calibration standards, you'd have to be very careful to make a real measurement of  much better than about 35dB!

It's too easy to become mesmerised by apparently good numbers when making measurements. Scepticism is always required.

73

Chris

G4DGU

On 29/09/2020 12:43, Dominique Dehays wrote:

Hi

 

Dynamic range is ok , but directivity , even with a good calibration kit can not be high enough for such a rl

73

Dom

 

-------- Message d'origine --------

De : Robin Szemeti - G1YFG <robin@...>

Date : 29/09/2020 13:41 (GMT+01:00)

Objet : Re: [UKMicrowaves] setting up an interdigital filter 23cm

 

The return loss is almost certainly as described, modern instruments have good dynamic range ... what is probably questionable is whether the load on the end is truly 50 ohms, it might be a -50dB match, but possibly not against a real 50R. That said, when in use, it is unlikely to be fed from or into a true 50R either.

 

On Tue, 29 Sep 2020 at 12:33, Dominique Dehays <f6dro@...> wrote:

Hi

 

What is not a progres is measuring 50db rl or more like in last dubus and believe its true

73

Dom


--
Robin Szemeti - G1YFG


Virus-free. www.avg.com