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

Re: [softrock40] Re: SoftRock filters and inductor variability

Expand Messages
  • Jasmine Strong
    *sigh* Monolithic chip inductors are available in tight tolerances, have high Q, and we wouldn t even have to wind the f#$@# things. -J. ... *sigh* Monolithic
    Message 1 of 10 , Jan 6, 2013
    • 0 Attachment

      *sigh*

      Monolithic chip inductors are available in tight tolerances, have high Q, and we wouldn't even have to wind the f#$@# things.

      -J.

      On 6 Jan 2013, at 12:19, "gerhard972" <gerhard.vaneerden@...> wrote:

       

      Alan,

      Thanks for your response.
      I have uploaded the pictures to the PA1GVE album.

      73 Gerhard PA1GVE

      --- In softrock40@yahoogroups.com, "Alan" wrote:
      >
      >
      > ----- Original Message -----
      > Subject: [softrock40] SoftRock filters and inductor variability
      >
      >
      > Hello all,
      >
      > >first I was somewhat reluctant to accept that there could be a design flaw.
      >
      > Gerhard,
      >
      > I'm not sure there is a design flaw. It seems to me that some of the cores are just not of the nominal permeance resulting in higher
      > inductance than intended.
      > If the components actually conform to the design then there would be no trouble.
      >
      > >The inductance of L2 and L3 increased by 25% by squeezing the turns together: from 798 and 802nH to 990 and 1031nH.
      >
      > My cores produced 1uH when spread over 360°. I did think that 270° was recommended to avoid self-capacitance.
      >
      > >I wanted to upload the VNWA pictures to the files section, but got the message that my file (100k) was to big. We have obviously
      > >reached the size limit of 100MB.
      >
      > You may sens as attachments ot upload to the Photos section.
      >
      > >Based on the results above we should consider to clearly instruct builders to spread the turns over the full circumference (360°)
      > >of the core. Only a few mm should be left between the start and the end to avoid a short e.g. during soldering.
      >
      > I'm not sure what to recommend, basically some cores require fewer turns.
      > I have used Elsie to check the 30/20/17 build. The T37-6 cores I have (AL_Value 3.8) need one turn less than specified. With my
      > cores Elsie shows -2dB at 18MHz.
      > The 15/12/10m build is hardly affected by this problem.
      >
      > 73 Alan G4ZFQ
      >


    • Dave
      Jasmine: The Q of chip inductors is much less than that of toroid coils. I use chip inductors on by BP board. The common tolerance available is 5%.. The main
      Message 2 of 10 , Jan 6, 2013
      • 0 Attachment
        Jasmine:
          The Q of chip inductors is much less than that of toroid coils.  I use chip inductors on by BP board.  The common tolerance available is 5%.. The main affect of the lower Q is higher insertion loss.

        Dave - WB6DHW
        <http://wb6dhw.com>

        On 1/6/2013 12:20 PM, Jasmine Strong wrote:
         

        *sigh*

        Monolithic chip inductors are available in tight tolerances, have high Q, and we wouldn't even have to wind the f#$@# things.

        -J.

        On 6 Jan 2013, at 12:19, "gerhard972" <gerhard.vaneerden@...> wrote:

         

        Alan,

        Thanks for your response.
        I have uploaded the pictures to the PA1GVE album.

        73 Gerhard PA1GVE

        --- In softrock40@yahoogroups.com, "Alan" wrote:
        >
        >
        > ----- Original Message -----
        > Subject: [softrock40] SoftRock filters and inductor variability
        >
        >
        > Hello all,
        >
        > >first I was somewhat reluctant to accept that there could be a design flaw.
        >
        > Gerhard,
        >
        > I'm not sure there is a design flaw. It seems to me that some of the cores are just not of the nominal permeance resulting in higher
        > inductance than intended.
        > If the components actually conform to the design then there would be no trouble.
        >
        > >The inductance of L2 and L3 increased by 25% by squeezing the turns together: from 798 and 802nH to 990 and 1031nH.
        >
        > My cores produced 1uH when spread over 360°. I did think that 270° was recommended to avoid self-capacitance.
        >
        > >I wanted to upload the VNWA pictures to the files section, but got the message that my file (100k) was to big. We have obviously
        > >reached the size limit of 100MB.
        >
        > You may sens as attachments ot upload to the Photos section.
        >
        > >Based on the results above we should consider to clearly instruct builders to spread the turns over the full circumference (360°)
        > >of the core. Only a few mm should be left between the start and the end to avoid a short e.g. during soldering.
        >
        > I'm not sure what to recommend, basically some cores require fewer turns.
        > I have used Elsie to check the 30/20/17 build. The T37-6 cores I have (AL_Value 3.8) need one turn less than specified. With my
        > cores Elsie shows -2dB at 18MHz.
        > The 15/12/10m build is hardly affected by this problem.
        >
        > 73 Alan G4ZFQ
        >



      • Orin
        ... Perhaps when originally tested, that set of cores had lower A_L values than expected? ... Hmmm. See:
        Message 3 of 10 , Jan 6, 2013
        • 0 Attachment
          --- In softrock40@yahoogroups.com, "Alan" wrote:
          >
          >
          > ----- Original Message -----
          > Subject: [softrock40] SoftRock filters and inductor variability
          >
          >
          > Hello all,
          >
          > >first I was somewhat reluctant to accept that there could be a design flaw.
          >
          > Gerhard,
          >
          > I'm not sure there is a design flaw. It seems to me that some of the cores are just not of the nominal permeance resulting in higher
          > inductance than intended.
          > If the components actually conform to the design then there would be no trouble.


          Perhaps when originally tested, that set of cores had lower A_L values than expected?


          >
          > The 15/12/10m build is hardly affected by this problem.

          Hmmm. See:

          https://picasaweb.google.com/lh/photo/peigIX9YYsJXR5W2VHPV0YLS1FToz1TiCJzHAUEj7Cg?feat=directlink

          vs:

          https://picasaweb.google.com/lh/photo/Ujjps0q183ySKsaP29mDm4LS1FToz1TiCJzHAUEj7Cg?feat=directlink

          I have replaced the filter with a Cauer design, but have yet to scan it on the VNWA. My 15m 2nd harmonic is now at -40dB and I have (according the the spectrum analyzer at least), 1W output on 10m. I wouldn't recommend this change unless you have a way of accurately measuring small value capacitors and can set the resonant frequency of the parallel coils/capacitors by tweaking one or the other. Not too bad with the VNWA, I could put the toroid and capacitor in parallel in my test fixture.

          Values for the filter came straight from SVC Filter Designer from Tonne Software (same source as Elsie): 5th order, 30MHz bandwidth, C-input lowpass, Cauer, 0.044 dB Passband ripple, 40dB Stopband depth. This gives over 26dB 'attenuation' of the 15m second harmonic. 3dB to go...

          Orin.
        • Alan
          ... Subject: [softrock40] Re: SoftRock filters and inductor variability ... Orin, The T37-6 has a specified AL_Value of 3.0 +/-5% The 40/30/20m LPF was
          Message 4 of 10 , Jan 7, 2013
          • 0 Attachment
            ----- Original Message -----
            Subject: [softrock40] Re: SoftRock filters and inductor variability


            >> I'm not sure there is a design flaw. It seems to me that some of the cores are just not of the nominal permeance resulting in
            >> higher
            >> inductance than intended.
            >> If the components actually conform to the design then there would be no trouble.
            >
            >
            > Perhaps when originally tested, that set of cores had lower A_L values than expected?
            >

            Orin,

            The T37-6 has a specified AL_Value of 3.0 +/-5%
            The 40/30/20m LPF was designed to be 0.8uH = 16 turns.

            I tested two 37-6 cores with 16 turns and the lowest inductance I could achieve with the turns spread was 0.97uH and 0.98uH. I used
            two completely different ways of measurement. A PIC meter and the old fashioned way with a 1% capacitor and dip meter and a RX to
            confirm resonant frequency.

            Working backwards this gives an AL_Value of 3.8. Only 14 turns are needed to get the corectly designed value.


            >
            >>
            >> The 15/12/10m build is hardly affected by this problem.
            >
            > Hmmm. See:
            >
            > https://picasaweb.google.com/lh/photo/peigIX9YYsJXR5W2VHPV0YLS1FToz1TiCJzHAUEj7Cg?feat=directlink
            > vs:
            > https://picasaweb.google.com/lh/photo/Ujjps0q183ySKsaP29mDm4LS1FToz1TiCJzHAUEj7Cg?feat=directlink
            >

            Sorry, I've only just got to terms with Elsie. I'm not quite sure what I am looking at, it's the 30/20/15 as built and with 2 turns
            removed? Did you measure the inductors?

            All I have checked with Elsie is the loss at the highest frequency, at 20m for the 40/30/20m version. Using the original design
            values Elsie showed about -0.3dB but with my measured values about -5dB.

            My interest is to find out why, maybe just some, of the 40/30/20s did not produce full power on 20m.
            Paul pointed the way, I think it may be due to the cores.

            While those like you can play with different designs I was wondering what to tell people who have built a bad 40/30/20.
            Do we tell them to rip out the LPF, get new capacitors and rebuild?
            I think just removing a turn, or two at the most, from each coil could return the LPFs to near enough the design parameters.

            I repeated this for the 30/20/17m version, again T37-6 cores. -2dB at 18MHz with the bad core value. Not so noticeable but one turn
            removed would be better.
            Elsie showed the 15/12/10 version to be only slightly affected by the high AL_Value.

            73 Alan G4ZFQ


            > I have replaced the filter with a Cauer design, but have yet to scan it on the VNWA. My 15m 2nd harmonic is now at -40dB and I
            > have (according the the spectrum analyzer at least), 1W output on 10m. I wouldn't recommend this change unless you have a way of
            > accurately measuring small value capacitors and can set the resonant frequency of the parallel coils/capacitors by tweaking one or
            > the other. Not too bad with the VNWA, I could put the toroid and capacitor in parallel in my test fixture.
            >
            > Values for the filter came straight from SVC Filter Designer from Tonne Software (same source as Elsie): 5th order, 30MHz
            > bandwidth, C-input lowpass, Cauer, 0.044 dB Passband ripple, 40dB Stopband depth. This gives over 26dB 'attenuation' of the 15m
            > second harmonic. 3dB to go...
            >
          • Orin
            ... Right. I might not have been clear when I said originally tested , I meant that perhaps when Tony originally designed and tested the filters, his cores
            Message 5 of 10 , Jan 7, 2013
            • 0 Attachment
              --- In softrock40@yahoogroups.com, "Alan" wrote:
              >
              >
              > ----- Original Message -----
              > Subject: [softrock40] Re: SoftRock filters and inductor variability
              >
              >
              > >> I'm not sure there is a design flaw. It seems to me that some of the cores are just not of the nominal permeance resulting in
              > >> higher
              > >> inductance than intended.
              > >> If the components actually conform to the design then there would be no trouble.
              > >
              > >
              > > Perhaps when originally tested, that set of cores had lower A_L values than expected?
              > >
              >
              > Orin,
              >
              > The T37-6 has a specified AL_Value of 3.0 +/-5%
              > The 40/30/20m LPF was designed to be 0.8uH = 16 turns.
              >
              > I tested two 37-6 cores with 16 turns and the lowest inductance I could achieve with the turns spread was 0.97uH and 0.98uH. I used
              > two completely different ways of measurement. A PIC meter and the old fashioned way with a 1% capacitor and dip meter and a RX to
              > confirm resonant frequency.
              >
              > Working backwards this gives an AL_Value of 3.8. Only 14 turns are needed to get the corectly designed value.


              Right. I might not have been clear when I said 'originally tested', I meant that perhaps when Tony originally designed and tested the filters, his cores had low AL values.


              > >> The 15/12/10m build is hardly affected by this problem.
              > >
              > > Hmmm. See:
              > >
              > > https://picasaweb.google.com/lh/photo/peigIX9YYsJXR5W2VHPV0YLS1FToz1TiCJzHAUEj7Cg?feat=directlink
              > > vs:
              > > https://picasaweb.google.com/lh/photo/Ujjps0q183ySKsaP29mDm4LS1FToz1TiCJzHAUEj7Cg?feat=directlink
              > >
              >
              > Sorry, I've only just got to terms with Elsie. I'm not quite sure what I am looking at, it's the 30/20/15 as built and with 2 turns
              > removed? Did you measure the inductors?


              Yes, the first scan is before tuning and the second is with two turns removed from the inductors. The mem traces in the second scan were imported from Elsie. BTW, if you haven't found it yet, try setting Stepsize to 5% from the button when displaying the plot, then hit the Monte Carlo button. This will show the expected limits in the frequency response of the filter with up to 5% component variation.

              I measured one of the inductors after removing it this weekend and with two turns removed, after tuning, it was 330nH. I may have disturbed it somewhat since that scan was taken. I got reasonable results after tuning for 10m, but the filter was inadequate for 15m.

              I don't know if I posted my results for 16 turns; I wound a 16 turn inductor this weekend just to see what happened. I got 790nH for the turns evenly spread and 825nH for 75% coverage. That inductor is no more as the Cauer design needed 240 and 320 nH inductors.


              >
              > All I have checked with Elsie is the loss at the highest frequency, at 20m for the 40/30/20m version. Using the original design
              > values Elsie showed about -0.3dB but with my measured values about -5dB.
              >
              > My interest is to find out why, maybe just some, of the 40/30/20s did not produce full power on 20m.
              > Paul pointed the way, I think it may be due to the cores.


              I would agree with it being due to the cores.


              >
              > While those like you can play with different designs I was wondering what to tell people who have built a bad 40/30/20.
              > Do we tell them to rip out the LPF, get new capacitors and rebuild?
              > I think just removing a turn, or two at the most, from each coil could return the LPFs to near enough the design parameters.


              I agree - I would recommend winding evenly spaced and if power out at 20m is low, try removing one or two turns.

              Orin.
            • gerhard972
              Alan, I certainly do not deny that cores can have more than 5% deviation in their Al value. So removing one or two turns if the ouput on 20m is low can be a
              Message 6 of 10 , Jan 7, 2013
              • 0 Attachment
                Alan,

                I certainly do not deny that cores can have more than 5% deviation in their Al value. So removing one or two turns if the ouput on 20m is low can be a good option. But it is still my opinion that winding in the way for which the Al value is valid (360°), will result in less variation in inductance value. However it will not compensate for deviations in Al values.

                For the sake of completeness I measured the inductance and SRF for 180°, 270° and 360° (estimated angels!). The results are:
                180° 1023nH, 261MHz so Cw=0.36pF
                270° 854nH, 280MHz so Cw=0.38pF
                360° 799nH, 237MHz so Cw=0.57pF
                So indeed the 270° has the highest SRF, but also 7% more inductance than expected, based on AL.
                The effect of the variation in winding capacitance is 0.01dB on 14MHz, 0.4dB on 28MHz and 1.5dB on 100MHz. This effect is not negligible, but small compared to the effect of the inductance variations.
                Above 100MHz the theoretical response is below -93dB, so probably other effects like crosstalk dominate.
                The SRF results are uploaded to the PA1GVE album.

                73 Gerhard PA1GVE

                --- In softrock40@yahoogroups.com, "Orin" wrote:
                >
                >
                > --- In softrock40@yahoogroups.com, "Alan" wrote:
                > >
                > >
                > > ----- Original Message -----
                > > Subject: [softrock40] Re: SoftRock filters and inductor variability
                > >
                > >
                > > >> I'm not sure there is a design flaw. It seems to me that some of the cores are just not of the nominal permeance resulting in
                > > >> higher
                > > >> inductance than intended.
                > > >> If the components actually conform to the design then there would be no trouble.
                > > >
                > > >
                > > > Perhaps when originally tested, that set of cores had lower A_L values than expected?
                > > >
                > >
                > > Orin,
                > >
                > > The T37-6 has a specified AL_Value of 3.0 +/-5%
                > > The 40/30/20m LPF was designed to be 0.8uH = 16 turns.
                > >
                > > I tested two 37-6 cores with 16 turns and the lowest inductance I could achieve with the turns spread was 0.97uH and 0.98uH. I used
                > > two completely different ways of measurement. A PIC meter and the old fashioned way with a 1% capacitor and dip meter and a RX to
                > > confirm resonant frequency.
                > >
                > > Working backwards this gives an AL_Value of 3.8. Only 14 turns are needed to get the corectly designed value.
                >
                >
                > Right. I might not have been clear when I said 'originally tested', I meant that perhaps when Tony originally designed and tested the filters, his cores had low AL values.
                >
                >
                > > >> The 15/12/10m build is hardly affected by this problem.
                > > >
                > > > Hmmm. See:
                > > >
                > > > https://picasaweb.google.com/lh/photo/peigIX9YYsJXR5W2VHPV0YLS1FToz1TiCJzHAUEj7Cg?feat=directlink
                > > > vs:
                > > > https://picasaweb.google.com/lh/photo/Ujjps0q183ySKsaP29mDm4LS1FToz1TiCJzHAUEj7Cg?feat=directlink
                > > >
                > >
                > > Sorry, I've only just got to terms with Elsie. I'm not quite sure what I am looking at, it's the 30/20/15 as built and with 2 turns
                > > removed? Did you measure the inductors?
                >
                >
                > Yes, the first scan is before tuning and the second is with two turns removed from the inductors. The mem traces in the second scan were imported from Elsie. BTW, if you haven't found it yet, try setting Stepsize to 5% from the button when displaying the plot, then hit the Monte Carlo button. This will show the expected limits in the frequency response of the filter with up to 5% component variation.
                >
                > I measured one of the inductors after removing it this weekend and with two turns removed, after tuning, it was 330nH. I may have disturbed it somewhat since that scan was taken. I got reasonable results after tuning for 10m, but the filter was inadequate for 15m.
                >
                > I don't know if I posted my results for 16 turns; I wound a 16 turn inductor this weekend just to see what happened. I got 790nH for the turns evenly spread and 825nH for 75% coverage. That inductor is no more as the Cauer design needed 240 and 320 nH inductors.
                >
                >
                > >
                > > All I have checked with Elsie is the loss at the highest frequency, at 20m for the 40/30/20m version. Using the original design
                > > values Elsie showed about -0.3dB but with my measured values about -5dB.
                > >
                > > My interest is to find out why, maybe just some, of the 40/30/20s did not produce full power on 20m.
                > > Paul pointed the way, I think it may be due to the cores.
                >
                >
                > I would agree with it being due to the cores.
                >
                >
                > >
                > > While those like you can play with different designs I was wondering what to tell people who have built a bad 40/30/20.
                > > Do we tell them to rip out the LPF, get new capacitors and rebuild?
                > > I think just removing a turn, or two at the most, from each coil could return the LPFs to near enough the design parameters.
                >
                >
                > I agree - I would recommend winding evenly spaced and if power out at 20m is low, try removing one or two turns.
                >
                > Orin.
                >
              Your message has been successfully submitted and would be delivered to recipients shortly.