Loading ...
Sorry, an error occurred while loading the content.
 

RE: [spectrumanalyzer] Re: Narrow Filter Module Results

Expand Messages
  • John Miles
    Just as another data point, I saw pretty good results from cascading a couple of those ECS10.7-15Bs. The trick was to use a couple of high-grade piston
    Message 1 of 11 , May 15, 2007
      Just as another data point, I saw pretty good results from cascading a
      couple of those ECS10.7-15Bs. The trick was to use a couple of high-grade
      piston trimmer caps as termination, rather than trying to guess the right
      termination capacitance. It also proved to be helpful to keep the filters
      well-isolated from each other, with the connection between them at DC
      ground.

      Small photo: http://www.thegleam.com/ke5fx/equinox/flt107.jpg
      Closeup: http://www.thegleam.com/ke5fx/equinox/large/flt107.jpg
      Measured response: http://www.thegleam.com/ke5fx/equinox/ecs.gif

      The input chokes are 6.8 uH, while the DC ground choke between the filters
      is 1 mH. The fixed silver-mica caps are 27 pF, and the piston trimmers are
      2-20 pF ceramic jobs that somebody on eBay was selling a long time ago.

      I was going for 50 ohms in, 50 ohms out; don't know if that's the case with
      the SSA/MSA project or not (this filter assembly was built for a different
      project).

      -- john, KE5FX


      >
      > The original Narrow Filter schematic shows using ECS 10.7-15B
      > filters however all of the parts are for ECS10.7-7B Monolithic
      > Crystal filters.
      >
      > The specs are as follows:
      >
      > ECS10.7-7B (Bandpass 3dB MIN +/- 3.75 Khz)(Terminating 1.8K/4.0pf)
      > ECS10.7-15B (Bandpass 3dB MIN +/- 7.50 Khz)(Terminating 3.0K/2.0pf)
      >
      > So if you are using the ECS 10.7-15B's change the terminating
      > impedance to 3.0K and the caps to 2pf. Also watch the DOT's for F1
      > and F2 they should be at C6 and for F3 and F4 at C8.
      >
      > On the IF2 again all the values are for ECS10.7-7B but in the BOM
      > ECS10.7-15B are called out, the schematic shows ECS10.7-7B in a two
      > pole configuration, caps should be 5.0pf for the single filter
      > configuration.
      > ECS10.7-7A (Bandpass 3dB MIN +/- 3.75 Khz)(Terminating 1.8K/5.0pf)
      >
      > Dyan... KV9R
      >
    • xjwmcx
      Hi Sam, Thanks for the info. I was a confused, because Cash has not had the Narrow Filter Module available for some time, now, and I wasn t sure (and still
      Message 2 of 11 , May 15, 2007

        Hi Sam,

        Thanks for the info.  I was a confused, because Cash has not had the Narrow Filter Module available for some time, now, and I wasn't sure (and still wonder if) he was planning to make it available again, because of the problems with the passband ripple.  I think it should work fine if the proper coupling capacitor value is used.  I hope he will duplicate what I did to be sure it is a good and repeatable solution to the problem.

        Regards

        Jim

            

        --- In spectrumanalyzer@yahoogroups.com, "Sam Wetterlin" <swetterlin@...> wrote:

        >
        > Jim,
        >
        > I designed the IF2 AGC modules with the noise filter, so I can respond
        > to that part of your question. The noise filter is intended to be
        > used as part of the AGC loop to reduce the bandwidth of noise
        > introduced by the amplifiers, which are acting after the Narrow Filter
        > has already established the signal bandwidth (RBW). The amplifiers
        > have some LC noise filtering built in, but the noise filter is
        > designed to allow further narrowing of the noise bandwidth. It can
        > provide a useful function in detecting very low signal levels, but is
        > not absolutely necessary to use.
        >
        > The noise filter does not need an extremely high level of rejection.
        > There actually is some amplification after the noise filter, which
        > will introduce a bit of new noise anyway. I originally intended that
        > the noise filter could also further refine the RBW, but it turns out
        > that the noise filter must have a broader bandwidth than the Narrow
        > Filter, to avoid the possibility of certain signals overloading the
        > first part of the AGC loop amplification. Therefore, all things
        > considered, a single MCF does the job fine. Also, because the noise
        > filter has an op amp following the MCF, which provides a high
        > impedance load and also makes up for insertion loss in the MCF, the
        > impedance matching is relatively straightforward without transformers
        > or fine tuning.
        >
        > Because the noise filter bandwidth is broader than the RBW established
        > by the Narrow Filter, it should not contribute significantly to the
        > overall RBW shape (i.e. the shape you see on the SA at 0Hz with no
        > signal applied, or the shape of any pure signal.) I do not have a
        > screen shot of the noise filter response, but as long as the top 20db
        > is reasonably smooth, and the ultimate rejection is 40db or better, it
        > will be fine. Any bump in the response will be irrelevant if it does
        > not occur near the top, because anything not near the top should be
        > outside the RBW. And there is no reason there should be any bumps
        > until perhaps near the base of the response.
        >
        > Finally, the noise filter is intended to be inserted between the VGA
        > (or the two VGAs, if you use two) and the IF2 AGC module, and so it is
        > only useful if you are using the AGC loop detector system. If you are
        > using the log amp, you might think you could use the noise filter
        > prior to the log amp. However, because so much amplification occurs
        > inside the log amp, the noise filter can have only a very modest
        > effect. For the noise filter to be effective, it needs to be inserted
        > where there is only about 40db or less of amplification to follow it.
        > (What you would really want to do is put the noise filter inside the
        > log amp in the middle of the amplification chain.)
        >
        > The log amp is a straightforward and simple amplifier, but I designed
        > the AGC loop to squeeze every last db of dynamic range out of the SA,
        > perhaps an extra 10-15db, and maybe 20db if you get very narrowband. I
        > wouldn't even say that most people need that extra range, but the AGC
        > loop is there for people who want it. (Note also that extra range is
        > not entirely good; you may end up seeing spurs and internal anomalies
        > which will complicate your life.)
        >
        > Sam Wetterlin
        >
        > --- In spectrumanalyzer@yahoogroups.com, "xjwmcx" jwmc@ wrote:
        > >
        > >
        > > Hi Cash, Sam, Scotty:
        > >
        > ....
        > > Cash and Sam, I think you collaborated on the design of the IF2
        > > amplifier? (Or was it Cash and Scotty?) The IF2 Noise Filter has no
        > > transformers, and it has 4 pF for the value of all of the capacitors.
        > > What does the passband response look like? Would you please post a
        > > screen shot of the response, so I can compare it to my results? What is
        > > the -3 dB bandwidth? Why does it use only one MCF? Does one MCF provide
        > > adequate filtering? Cash, do you plan to make the NFM available?
        > >
        > ....
        > >
        >

      • xjwmcx
        Hi Scotty, Thanks for your analysis, your technical tips, and your kind works of encouragement. They are sincerely appreciated. At times, I felt like I was
        Message 3 of 11 , May 15, 2007

          Hi Scotty,

          Thanks for your analysis, your technical tips, and your kind works of encouragement.  They are sincerely appreciated.

          At times, I felt like I was hacking my way through a thicket of techncal issues and challenging assembly work, but it has never been dull or uninteresting.  On the contrary, this has been an interesting and rewarding project.  And a good learning experience.  I hope to get started on the tracking generator project soon, but I will work on the "spur" problem first.  I want to check for spurs and/or unwanted signals that may be leaking through my system.  I plan to  check it out over a large frequecy range. 

          But first, I have to get my presentation ready for the Boulder Amateur Radio Club meeting on June 19!  I have not been able to locate the picture of yourself which you previously had posted on your website.  Do you have a picture file you would be willing to email to me.  I would like to use it in my presentation to the club.

          By the way, is there a software application you have that works better for you for opening my posted files?  I have all of the Office 2000 applications, but not much else.  And I don't have any way to create PDF files.  Can you deal better with Excell 2000?

          Thanks again and best regards,

          Jim

           


          --- In spectrumanalyzer@yahoogroups.com, "Scotty" <wsprowls@...> wrote:

          >
          > Hi Jim,
          > Your newest plots are excellent. I have a minor problem, in that, I
          > must open your document in Lotus Word Pro. I have an old (96) version
          > of MS Word, which is usless.
          > Anyway, here are my observations.
          > First, your final filter 3 dB bandwidth looks to be about 10.8 KHz. I
          > will use this number for noise analysis.
          > The center frequency level is at -30 dB.
          > The noise level dip at approx 27 KHz away from CF is -79 dB. By the
          > way, if you slow the sweep by changing the "Wait" to 100 or more, both
          > dips should be equal. Your sweep may be faster than the settling time
          > of the Log circuit.
          > To determine Total Phase noise in a 10.8 KHz bandwidth: -79 dB - (-30
          > dB) = -49 dB. (S to N ratio)
          > Your final filter has an effective noise bandwidth of 10Log(10,800 Hz)
          > = 40.33 dB.
          > Therefore, to determine Total Phase noise in a 1 Hz bandwidth: -49 dB
          > - 40.33 dB = -89.33 dB
          > I expect a "standard" MSA to have from -90 to -92 dBc/Hz. So your's
          > is pretty darned close. I think the extra noise you have (1 to 2 dB)
          > is due to the speed of the sweep.
          > Now, for the funny spur that is appx 190 KHz below the carrier. Take
          > a REALLY close look at that signal. It is not just one signal. It is
          > at least 2 distinct signals, possibly 3, very close together. The
          > reason I can tell is that it has a 5 dB dip in the center. Your final
          > filter has a only a 1 dB ripple. Therefore, if the spur were a single
          > signal it's ripple would also be 1 dB. Prove this by sweeping many
          > times and verifying the signal pattern is consistant. If it changes,
          > it may be a spurious oscillation somewhere in the MSA.
          > It is possible that these signals are close-in DDS spurs. They are 67
          > dB below the carrier (-30 dB, -97 dB = delta 67 dB). This is 7 dB
          > below the DDS spec of -60 dBc. Try commanding your Working Window's
          > "Spur Test" button to ON, and see if the spur moves. This will
          > certainly prove the spur is not a real signal entering your MSA's
          > front end. It is quite possible it is not a DDS spur, so more
          > analysis is necessary.
          > I can see from your wide plot (.76 MHz) your MSA has a dynamic range
          > of 70 dB at 200 KHz. This is quite good. It means your MSA has at
          > least 70 dB dynamic range for an input signal of 200 KHz or greater.
          > You will be able to work with 455 KHz signals very nicely.
          > Well done, Jim.
          >
          >
          > --- In spectrumanalyzer@yahoogroups.com, "xjwmcx" jwmc@ wrote:
          > >
          > >
          > > Hi Cash, Sam, Scotty:
          > >
          > > I have done some work on the ugly passand ripple problem with my Narrow
          > > Filter module. My work leads to some interesting results and more
          > > questions.
          > >
          > > Cash did some work on the Narrow Filter Module (NFM) a while back (see
          > > the attached message from Cash, below) , and after pondering his
          > > results, my hunch was that the coupling capacitor between the two
          > > monolythic crystal filters (MCFs) could be the problem. (Note that the
          > > coupling capacitor is connected from the node between the two MCFs to
          > > ground; still, it is called a coupling capacitor.) Cash had tried
          > > removing the transformers and using 1800 ohm resistors to match the
          > > input and the output of the NFM. This resulted in no significant
          > > difference in in the passband ripple (i.e., still ugly). He also,
          > > changed all of the capacitors from 5 pF to 4 pF with no apparent change
          > > in the passband ripple.
          > >
          > > So, I decided to see what effects would be produced by changing the
          > > value of the coupling capacitor. First, I touched the blade of a small,
          > > insulated screwdriver to the node between the two MCFs. The passband
          > > ripple response changed instantly for the better, so that the passband
          > > ripple was only about 1.5 dB, and the bandwidth narrowed a bit. Then, I
          > > soldered a 3 pF smt capacitor in parallel with the 5 PF capacitor which
          > > was originally loaded in the circuit. As expected, this caused the
          > > passband ripple to decrease to about 1.5 dB. Next, I added another 1.5
          > > pF in parallel with the 8 pF, and the passband ripple decreased to less
          > > than 1 dB, and the passband ripple was almost perfectly symmetrical.
          > > (The 1.5 pf was actually two 3 pF caps in series.) And the -3 dB
          > > bandwidth if the NFM is 9.5 kHz. (It was about 12 kHz.) Now the
          > > effective value of the coupling capacitor is 9.5 pF.
          > >
          > > Cash and Sam, I think you collaborated on the design of the IF2
          > > amplifier? (Or was it Cash and Scotty?) The IF2 Noise Filter has no
          > > transformers, and it has 4 pF for the value of all of the capacitors.
          > > What does the passband response look like? Would you please post a
          > > screen shot of the response, so I can compare it to my results? What is
          > > the -3 dB bandwidth? Why does it use only one MCF? Does one MCF provide
          > > adequate filtering? Cash, do you plan to make the NFM available?
          > >
          > > The specs for the MCFs used in the NFM specify a minimum -3 dB bandwidth
          > > of 7.5 kHz, but the maximum -3 dB bandwidth is not specified. The
          > > attenuation in the stopband is speced at 40 dB at Fo plus/minus 14 kHz.
          > > So, a -3 dB bandwidth of 9.5kHz seems reasonable. I presume that the
          > > combination of the two MCFs in the NFM should produce a stopband
          > > attenuation of 80 dB at plus/minus 14 kHz, but I can't measure this
          > > because it is obscured by the phase noise in my system. (Or should the
          > > NFM remove most of the phase noise?) I am planning to shield the output
          > > of the NFM from the input, but I haven't done this yet.
          > >
          > > Finally, how can the results I got with the NFM be explained. (And the
          > > results Cash got, also?) It actuall makes sense, intuitively, because
          > > the specs imply that the output of the MCF with 4 pF connected to ground
          > > is 1800 ohms. If so, connecting two MCFs in series would require each
          > > one to have 4 pF connected to ground, for a total of 8 pF. But intuition
          > > is not always correct.
          > >
          > > I have posted a file with some screen shots obtained with my system. The
          > > file name is 20MHz_1.doc. I added a 1.0 uF capacitor at the input of my
          > > A/D converter, as Scotty suggested. (Scotty's home-brew 12-bit A/D
          > > converter/controller board.) The capacitor really cleaned up the input
          > > to the A/D converter.
          > >
          > > Scotty, I get a spur at Fo minus 190 kHz no matter what frequency Fo is.
          > > I looked at it with 4 or 5 different frequencies from 10 MHz to 200 MHz.
          > > Figure 6 is a close-up of the spur. Any suggestions about where this may
          > > be coming from? Any general (or specific) comments?
          > >
          > > Regards
          > >
          > > Jim McLucas
          >

        • Jim Miller
          google for a free pdf writer. there are several out on the web. jim ab3cv ... From: xjwmcx To: spectrumanalyzer@yahoogroups.com Sent: Wednesday, May 16, 2007
          Message 4 of 11 , May 16, 2007
            google for a free pdf writer. there are several out on the web.
             
            jim ab3cv
             
            ----- Original Message -----
            From: xjwmcx
            Sent: Wednesday, May 16, 2007 1:58 AM
            Subject: [spectrumanalyzer] Re: Narrow Filter Module Results

              And I don't have any way to create PDF files. 

          • S. Cash Olsen
            Good morning Jim and the group, Jim I think you were on the right track by adjusting the capacitance of the coupling capacitor. Way back in the archives Harold
            Message 5 of 11 , May 17, 2007
              Good morning Jim and the group,
               
              Jim I think you were on the right track by adjusting the capacitance of the coupling capacitor.
               
              Way back in the archives Harold Johnson pointed to a web page with lots of "tuning" information. I will see if I can dig that up. The long and short of it, if memory serves correctly, was that the capacitances that ECS specifies for their parts is the capacitance of the node with in the part and that the actual capacitance needed is in parallel with this capacitance. The required capacitance might be several times the stated load capacitance.
               
              I think John Miles indirectly showed that with his set up. In the adaptation to Scotty's original circuit that I made to the boards by adding the transformers was because I did not understand one type of impedance transformer for another, at the time. The MFC filters that Scotty had pulled out of his "junk box" were not the same as I was going to be able to purchase from Digikey or Mouser and so I added the transformers. I do not have pictures of the transformer version (presently) but eventually it was made to work very well.
               
              But, it is not necessary to use the transformers, in fact, they are expensive if purchased and some what of a hassal if wound by hand. An inductive impedance  transform can be used with almost any combination of MFC and the boards. In that circuit a simple inductor (or capacitor) can be used and then a resonating capacitor (or inductor) can be used. This scheme actually has several advantages. There are no resistive losses and because the transformation is tuned to the frequency desired you get additional selectivity in the filter. You will probably need a similar circuit on the output side depending on the exact application.
               
              I will plan to offer the filter boards again, possibly as early as July. I want to clean up a couple of problems and will re-release the filter board and the 120dB Log detector board. The 1 an 3 amplifier boards need a similar tweek and I will make them available again.
               
              Cash
               
              P.S. I have been distracted by another project that has diverted most of my attention but this is temporary and nearly complete.
              ----- Original Message -----
              From: xjwmcx
              Sent: Tuesday, May 15, 2007 11:35 PM
              Subject: [spectrumanalyzer] Re: Narrow Filter Module Results

              Hi Sam,

              Thanks for the info.  I was a confused, because Cash has not had the Narrow Filter Module available for some time, now, and I wasn't sure (and still wonder if) he was planning to make it available again, because of the problems with the passband ripple.  I think it should work fine if the proper coupling capacitor value is used.  I hope he will duplicate what I did to be sure it is a good and repeatable solution to the problem.

              Regards

              Jim

                  

              --- In spectrumanalyzer@yahoogroups.com, "Sam Wetterlin" <swetterlin@...> wrote:
              >
              > Jim,
              >
              > I designed the IF2 AGC modules with the noise filter, so I can respond
              > to that part of your question. The noise filter is intended to be
              > used as part of the AGC loop to reduce the bandwidth of noise
              > introduced by the amplifiers, which are acting after the Narrow Filter
              > has already established the signal bandwidth (RBW). The amplifiers
              > have some LC noise filtering built in, but the noise filter is
              > designed to allow further narrowing of the noise bandwidth. It can
              > provide a useful function in detecting very low signal levels, but is
              > not absolutely necessary to use.
              >
              > The noise filter does not need an extremely high level of rejection.
              > There actually is some amplification after the noise filter, which
              > will introduce a bit of new noise anyway. I originally intended that
              > the noise filter could also further refine the RBW, but it turns out
              > that the noise filter must have a broader bandwidth than the Narrow
              > Filter, to avoid the possibility of certain signals overloading the
              > first part of the AGC loop amplification. Therefore, all things
              > considered, a single MCF does the job fine. Also, because the noise
              > filter has an op amp following the MCF, which provides a high
              > impedance load and also makes up for insertion loss in the MCF, the
              > impedance matching is relatively straightforward without transformers
              > or fine tuning.
              >
              > Because the noise filter bandwidth is broader than the RBW established
              > by the Narrow Filter, it should not contribute significantly to the
              > overall RBW shape (i.e. the shape you see on the SA at 0Hz with no
              > signal applied, or the shape of any pure signal.) I do not have a
              > screen shot of the noise filter response, but as long as the top 20db
              > is reasonably smooth, and the ultimate rejection is 40db or better, it
              > will be fine. Any bump in the response will be irrelevant if it does
              > not occur near the top, because anything not near the top should be
              > outside the RBW. And there is no reason there should be any bumps
              > until perhaps near the base of the response.
              >
              > Finally, the noise filter is intended to be inserted between the VGA
              > (or the two VGAs, if you use two) and the IF2 AGC module, and so it is
              > only useful if you are using the AGC loop detector system. If you are
              > using the log amp, you might think you could use the noise filter
              > prior to the log amp. However, because so much amplification occurs
              > inside the log amp, the noise filter can have only a very modest
              > effect. For the noise filter to be effective, it needs to be inserted
              > where there is only about 40db or less of amplification to follow it.
              > (What you would really want to do is put the noise filter inside the
              > log amp in the middle of the amplification chain.)
              >
              > The log amp is a straightforward and simple amplifier, but I designed
              > the AGC loop to squeeze every last db of dynamic range out of the SA,
              > perhaps an extra 10-15db, and maybe 20db if you get very narrowband. I
              > wouldn't even say that most people need that extra range, but the AGC
              > loop is there for people who want it. (Note also that extra range is
              > not entirely good; you may end up seeing spurs and internal anomalies
              > which will complicate your life.)
              >
              > Sam Wetterlin
              >
              > --- In spectrumanalyzer@yahoogroups.com, "xjwmcx" jwmc@ wrote:
              > >
              > >
              > > Hi Cash, Sam, Scotty:
              > >
              > ....
              > > Cash and Sam, I think you collaborated on the design of the IF2
              > > amplifier? (Or was it Cash and Scotty?) The IF2 Noise Filter has no
              > > transformers, and it has 4 pF for the value of all of the capacitors.
              > > What does the passband response look like? Would you please post a
              > > screen shot of the response, so I can compare it to my results? What is
              > > the -3 dB bandwidth? Why does it use only one MCF? Does one MCF provide
              > > adequate filtering? Cash, do you plan to make the NFM available?
              > >
              > ....
              > >
              >

            • xjwmcx
              Hi Cash, I think the situation with the MCF specs is that they simply got the coupling capacitor value wrong. The other cap values seem to be ok. I ll be
              Message 6 of 11 , May 17, 2007

                Hi Cash,

                I think the situation with the MCF specs is that they simply got the coupling capacitor value wrong.  The other cap values seem to be ok.  I'll be interested to know if you can duplicate my results.

                I know the alternative of using a simple LC matching circuit has some advantages, as Scotty pointed out a couple of times on this website.  However, for me, the most convenient solution was to use the transformers with the correct value of coupling capacitor, because I already had the parts on hand.  (Also, it was a challenge to see if I could make it work well.)

                Another consideration is that an LC matching circuit may have to be tuned in order to get it to work optimally.  The transformers don't require tuning.

                BTW, I would like to use a picture of yourself at my ham club presentation, if you would like to send me one.

                Regards,

                Jim


                --- In spectrumanalyzer@yahoogroups.com, "S. Cash Olsen" <KD5SSJ@...> wrote:

                >
                > Good morning Jim and the group,
                >
                > Jim I think you were on the right track by adjusting the capacitance of the coupling capacitor.
                >
                > Way back in the archives Harold Johnson pointed to a web page with lots of "tuning" information. I will see if I can dig that up. The long and short of it, if memory serves correctly, was that the capacitances that ECS specifies for their parts is the capacitance of the node with in the part and that the actual capacitance needed is in parallel with this capacitance. The required capacitance might be several times the stated load capacitance.
                >
                > I think John Miles indirectly showed that with his set up. In the adaptation to Scotty's original circuit that I made to the boards by adding the transformers was because I did not understand one type of impedance transformer for another, at the time. The MFC filters that Scotty had pulled out of his "junk box" were not the same as I was going to be able to purchase from Digikey or Mouser and so I added the transformers. I do not have pictures of the transformer version (presently) but eventually it was made to work very well.
                >
                > But, it is not necessary to use the transformers, in fact, they are expensive if purchased and some what of a hassal if wound by hand. An inductive impedance transform can be used with almost any combination of MFC and the boards. In that circuit a simple inductor (or capacitor) can be used and then a resonating capacitor (or inductor) can be used. This scheme actually has several advantages. There are no resistive losses and because the transformation is tuned to the frequency desired you get additional selectivity in the filter. You will probably need a similar circuit on the output side depending on the exact application.
                >
                > I will plan to offer the filter boards again, possibly as early as July. I want to clean up a couple of problems and will re-release the filter board and the 120dB Log detector board. The 1 an 3 amplifier boards need a similar tweek and I will make them available again.
                >
                > Cash
                >
                > P.S. I have been distracted by another project that has diverted most of my attention but this is temporary and nearly complete.
                > ----- Original Message -----
                > From: xjwmcx
                > To: spectrumanalyzer@yahoogroups.com
                > Sent: Tuesday, May 15, 2007 11:35 PM
                > Subject: [spectrumanalyzer] Re: Narrow Filter Module Results
                >

              Your message has been successfully submitted and would be delivered to recipients shortly.