Cavity Resonator for 29  MHz                                                                   18/5/13

With the advent in sunspot activity enhancing propagation in the higher frequencies  ie 10 m  I became interested in the possibility  of a local 10m repeater to make use of this phenomenon . There are of course many repeaters around on 10  with standard 100 KHz split . We  here in Rotorua  NZ  tend to hear the New York 10Fm  CTCSS repeater on 29.620  in the morning  and later on in the day that fades out and the Australian   29.620 MHz in Adelaide  and the Melbourne repeater on 29.640 appear, these are  open access . some times together  but strangely hear  one then the other .  These repeaters run receive and transmit sites geographically isolated by quite considerable distances  20-50 Kms and are linked by uhf point to point systems .  I notice that there are quite often times when one can hear the repeater well and cant get into it  and  vice  verse  I suspect the path between the distant transmitter and receiver  are not always  open to us  simultaneously  ... much  like the phenomenon of  diversity  reception .. thus when things are marginal this phenomenon causes the  communication problem  .  having the 10 m receiver and transmitter  on the same site or very close im sure would minimise this  problem  but because of the lack of  29 Mhz filtering  duplexers ,so  this co -siting  doesn't happen ...

I decided it would be interesting to build a cavity resonator  for 29 MHz . I had done many at 2m  and 70 cms so it shouldn't be too different . just  physically  larger  !. This was  to see what the performance would be   .Characteristics such as  pass-band.  insertion loss . notch depth  and the various configurations  and the notches they  produce  .. if this  project  showed some success !  well I may build  three  more !!. Im sure we could  duplex off two separate antennas  with reasonable horizontal separation .   I looked up the Net  and found the formula  for the coaxial dimensions to  obtain  the best  impedance for lowest loss ,   This was 77 ohms  and the ratio of coaxial diameters to produce this is about 3.6 to one .  I  used 60 mm O.D  Aluminium pipe as the centre resonator .  I was  given some cast off steam pipe , thin walled aluminium thermal jacket tubing . it was 216 mm in diameter . and in length 1.25 m,  normally filled with rock wool to thermally insulate  geothermal steam conduit ..

             

               Components  on the trailer         at my brothers workshop                            Overlapped and  screwed together  with self tapping screws

. I joined by overlap and self tapping metal screws , two of these  lengths to make one of 2.4 m long  .  I also  found a very useful formed Top end-cap structure of substantial rigidity to which I would weld my centre resonator and attach my BNC  RF connectors    in the form of an inexpensive nonstick aluminium frying pan of 216mm O.D !!  !    Check these frying pans out if your building cavity filters of reasonable diameters , as they are an excellent well formed piece of material,  cheap and readily obtainable in many different diameters  ... don't  forget to cut the handle off before you press them into service  !!

               

The cheap tiawanese  Aluminium       frypan ............... without  handle           Back home....slots in base to insert coupling loops and connectors

I very good friend of mine Brian Cowien in the  Rotorua Classic motorcycle club, to which I belong , was able to T.I.G weld the aluminium resonator the the  internal  base of the non stick  coated  pan for me (  coating duly removed  before welding )   , It was welded as a continuous bead around the entire circumference of the resonator to provide as low RF Impedance as possible  to minimise any losses in this high current  region ..

      

             RF Hot end held central with  2 inch thick polystyrene  foam  disk 

It  is bloody enormous!  and towers above me if I stand beside it !  and one will require a repeater site that has at least a ten foot stud  to accommodate the vertical  filter . Tuning of this beast  in situ , will have to be done using a step ladder !!  You will notice that this prototype  does not have the ability to tune its frequency response  except with judicious application of a hacksaw !  . I want to keep the prototype as simple as possible and see if its RF  performance as a coaxial resonator warrants proceeding further mechanically  and attaching a  trombone / telescoping  type tuning  mechanism  at a later date ..

                       

     RF coupling loops 170 mm  long and spaced 25.4 apart  copper strip 6mm wide 1 mm thick ( 2 of )                   resonator  Welding  and loops 

OK , finally made and placed the two coupling loops  scaled from the dimensions of a 6 m cavity  made by Gary Patterson   NZ5V.....  At 29 Mhz These where just under 17 cm long and spaced apart 25.4 mm   I have bent my coupling loops away from the centre resonator to lower the coupling thus  raise the loaded Q  and hopefully better  performance with respect to notch depth , position and insertion loss . These  loops provide too much coupling  , they need to be  "smaller"  but still a 50  match  .
. If I am going to build more of these filters I will use an RF  connector a bit more  robust than BNC  connectors ,  N type  ( overkill )  or  Flange mount  .. SO239 connectors .  But  one would have to use double screened  coax in application of the filters to maximise the system performance.
  .
   Turns out my 2.4 m  length of resonator  tunes at ~32 Mhz !! Any way that is no big deal as I can add a telescoping piece to the existing resonator to make it longer  and also I will have to lengthen the outside aluminium case .. .As it  is perfectly ok to experiment with .  below is a  loaded sweep of the filter as a simple bandpass showing the 3db points . .              

                                          

                                            Sweep  of resonator  loaded  with 50 ohm terminations        a  "QLoaded"   configuration                               

  I measured the base level signal going into the filter its from the  tracking generator -- Spectrum Analyser as   0.34dB     I have  bent the coupling loops to achieve around ~0.7 db  coupling loss  and hope the Q stays up enough to get a reasonable rejection  notch  , For experimental purposes, the notch was achieved with a length of 50 ohm coax  about 0.8 M long  between the two coupling loops  using T connectors  to enable this . I have not optimised this  length  at all,  it was a random length    , I will modify the  coupling configuration  and place a air-spaced  variable  phasing capacitor inside  the top plate connected between the two coupling loop hot  ends .  this is easier to adjust the rejection notch and loop coupling to  enable  the optimum result   ,                                                                                                                                                                                                                                                                            

                                       

                                            Random length of coax  "shorting "  input and out put         connection

At the moment I have achieved a  ~30 dB deep  notch 110Khz    away from the pass frequency  so the system shows promise  so far .  100 Khz split is a very difficult  one to achieve , 600 Khz at 144 Mhz is a 0.41 % shift ?..439 MHz and a 5 MHz split is a 1.13% shift  ?  29.6 Mhz and 100 Khz  split is only 0.33 %   ?  so its a very tall order  to achieve the separation and the low insertion loss.  The  next thing to try is the phasing capacitor  that connects the input and output together . inside the  top plate  and then play the phasing capacitance against the amount of loop coupling and see what we can achieve .I would  like to see if I can deepen this notch if possible  . I may have to look at increasing the overall "Q" of the system  to achieve this ..  I will drag out the E5062A  VNA  from storage  and measure the loaded and unloaded Q of the filter  and see if we can improve the Q  by judicious use of metal crews to pull the overlapped aluminium sheet  closer  for better  contact .  VNA's  are not the most accurate devices  for deriving the absolute  Q  of the  circuit  but they do the calculations on the fly and all I want to see is if I am increasing or decreasing the "Q" when I make modifications..   What I want to do is examine the return loss of the coupling loops  and to see if we can squeeze  more  performance of the device by optimising them  . I saw in  an article by K. Custer  ..W8JG ........  about some commercial loops which had their series inductance increased  by the judicious use of added inductance .strategically placed  (at right-angles) so as not to interfere with main loop coupling    if you are going to have a go at building cavity filters and duplexers of various configuration you must look through the  Repeater Builder  website  http://www.repeater-builder.com/rbtip/    Particularly  here    http://www.repeater-builder.com/antenna/ant-sys-index.html        there is a wealth of information to be  gained .. I constantly refer to this excellent  site ..

 Unloaded  "Q"  I suspect I should be able to achieve better  performance than I am at the moment  but I need to evaluate the unloaded "Q" of the resonator   to get some idea if the filter is good enough    . An excellent article in The repeater builder website by Jacques Audet VE2AZX      http://www.repeater-builder.com/antenna/pdf/ve2azx-duplexerinfo.pdf       tells how to derive  this      You need to couple to the cavity with small sized loops that give a more than  -20  db insertion   loss    so as not to load the tuned circuit .              An approximation of the  formula  is the pass-band centre frequency  in Hertz  divided by the actual 3dB bandwidth in hertz  ..  when I do the approx  maths my "Qu"  is  only close to ~2800 . that is not  very good  for a cavity filter   , This is  probably why I can only achieve slightly more than a 30 dB notch at  100 Khz from the passband . 
         I now  have to look at ways I can improve the "Qu" of this  filter  .. First thing I will try is increasing the peripheral contact between the sides of the "frying pan"  and the  outer case  .heaps  more screws should help  here   ????   ....    have added  more  screws around the periphery  and got my 3dB points 1 Khz closer together !!   from 12.47  KHz to 11.47 KHz  ?  The law of diminishing returns .... I would love to try a rolled  and welded section of thick walled aluminium here  but its a bit difficult to arrange  and im trying to build these inexpensively  as not every one may have access to lots of  mechanical options ..

To check that I was measuring the "Qu" correctly I performed a measurement  on the "200" Mhz  ( civil aviation DME  frequency in New Zealand )  it was tuned to 185 Mhz approx . the 3db points with very light coupling , greater  than -20 dB , were 19-20 KHz   this yielded a staggering "Qu"  of approx 9200 !!   and its an aluminium cavity with silver plated copper  centre  components  inside  , I fellow Amateur  suggested I remove all the non stick coating from the interior of my 32 Mhz cavity top plate and see if that  makes a difference ...I shall do this and recheck the "Qu" But  I am pleased to see that the less expensive Aluminium is a suitable material for the project as shown by the sinclair  filter  performance . 

                                      

                                      Unloaded  sweep  >-20 dB loop coupling  ( very light coupling) so as not to load the resonant circuit

                                      

                             loops  length 70mm  by  35 mm    inductance to push them near 50 Ohm    5Turn     guess  

Some of you are looking at the coupling loops  side ways  !!  so  take a look here  http://www.repeater-builder.com/projects/tx-rx-loops/couplingloopresearch.html any trick to screw the rejection notch  deeper !

 

                            

            sweep of  variable cap  airspaced 100 pf  across  BNC  centre pins       0.7 dB loss                        32 Mhz  ,  200 Mhz     146  Mhz

Suffice to say considering the ease and the cost to build this filter  I guess that performance figure is not bad  and three of these in each leg of a repeater configuration would allow  use of a single antenna  


I suspect two of these filters in each leg , (2 Tx  and 2 Rx  .).. good horizontal  or preferable vertical  antenna separation and one could use these in a co sited  repeater  situation

Well I think they are worth  making  given that the cost of construction and the resulting  performance  !

  The  real killer  is the  Size  ..........    73's   Mike

 

 

 

 

                                        

 

 

                                                                                           

 

 

 

 

 

                                                                                                                     back to home page