Announcement

Collapse
No announcement yet.

Multifrequency VLF IB Detector using Dual-Resonance Foster Network (6 18 kHz, 1st 3rd Harmonic)

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Multifrequency VLF IB Detector using Dual-Resonance Foster Network (6 18 kHz, 1st 3rd Harmonic)

    20260923 #2 Multifrequency VLF IB Detector using Dual-Resonance Foster Network (6 &18 kHz, 1st & 3rd Harmonic)

    The hardware base is the Quasar ARM: https://fandy.ucoz.org/QuasarARM/Quasar_ARM_SCH.pdf

    Multifrequency firmware is written, but the TX driver output is a 0/5 V square wave, and the boost to roughly 35 Vp for a single frequency is obtained from the TX series resonance with the quality factor controlled to about 6.
    With this hardware, non-resonant TX drive yields insufficient coil current. If two frequencies are acceptable and sufficient to be called as multifrequency - a dual-resonance Foster network should work well and enable multifrequency operation.
    Below is the LTspice simulation of the dual-resonance Foster network at 6 / 18 kHz (1st / 3rd harmonics): Q(6k / 18k) = 5.42 / 17.30 — both within the 5…20 band; Itx(6k / 18k) = 134.9 / 134.8 mA.
    5-th harmonic doesn't make sense doe to high about 41 quality factor, the stability of 5-th harmonic will be poor especially on wet conductive soils. So decision to use 6 and18kHz seams reasonable.
    Tx stage: Lp=0.68mH, Cp=200nF, Cs=353nF for my Ltx=1.153mH (Rdc=2.5R), Rdump=9.1R.
    Rx stage: Lrx=13.19mH, Rdc=24.7R, Cp=18nF.
    Click image for larger version

Name:	2026-09-23 214512  - Dual-Resonant Foster network 3-rd Harmonic 18kHz and 6kHz.png.png
Views:	822
Size:	271.9 KB
ID:	452999
    The first plot is the binary 0/5 V driver signal, which contains mostly the 1st and 3rd harmonics. The second plot shows the TX coil voltage (not sinusoidal) together with the coil current, which is a clean sum of the 6 kHz and 18 kHz components. The third plot is the RX coil voltage.

    The RX front end of the Quasar ARM requires only a minimal change for multifrequency operation:
    Click image for larger version

Name:	Quasar_ARM_SCH - Minimalistic Update.png
Views:	720
Size:	99.8 KB
ID:	453000
    Alternatively, a more thorough modification gives approximately three times lower noise and an expected detection depth gain of +23…31 %.
    ​ Click image for larger version

Name:	Quasar_ARM_SCH - Updated (+23..31 % depth).png
Views:	718
Size:	136.8 KB
ID:	453001
    Real HW shows similar to simulation waveforms: YELLOW: TX DRIVE, GREEN: TX COIL current, INDIGO: RX Amp output to ADC.
    Click image for larger version

Name:	20260923_232944.jpg
Views:	711
Size:	667.7 KB
ID:	453002

    Previous post 20260922 #1: DD Coil Electronic Balancing with Compensation Coil and Trimpot
    A balance-tuning coil for the RX circuit can be placed within the TX coil area. It consists of two series-connected compensation windings, each of 2–4 turns of wire of the same gauge as the RX coil. The center tap is connected to the cold end of the RX coil. The free ends of the two compensation windings are connected to the

  • #2
    Today I tested a 5" AliExpress coil with the Quasar ARM hardware and the firmware I’m developing.

    The coil has no resonance components inside. It’s sold for the Chinese TX-850 metal detector.

    The resonant networks are tuned for a TX frequency of about 10 kHz. I tuned the RX resonance lower than the TX frequency for better VDI stability:
    - TX, series resonance: Rs = 18 Ω, Cs = 330 nF (polypropylene)
    - RX, parallel resonance: Cp = 47 nF (polyester)

    Air test results at 9.615 kHz:
    Click image for larger version  Name:	Reborn 5-inch coil 9615Hz VDI.png Views:	0 Size:	59.6 KB ID:	453962

    The RX coil's residual imbalance is 30 mV peak-to-peak, so I had to reduce the compensation network resistor R5 from 91k to 22k.
    Click image for larger version  Name:	20261024_191252.jpg Views:	0 Size:	530.0 KB ID:	453963
    Click image for larger version  Name:	20261024_191723.jpg Views:	0 Size:	580.3 KB ID:	453964

    Here’s the adapter with the resonant networks:
    Click image for larger version  Name:	dongle_2.jpg Views:	0 Size:	203.4 KB ID:	453965 Click image for larger version  Name:	dongle_1.jpg Views:	0 Size:	318.6 KB ID:	453966

    With the adapter, any coil with any pinout can be tuned for any single or dual-frequency operation.

    Comment


    • #3
      I used another method in quasar arm, for tx I used custom half bridge with RCD driver adopted for high voltage (like whites spectra v3i) , powering made by dcdc step up mt3608 converter (that has high working frequency more than 1mhz and low inteference noise on quasar arm) , tx coil connected series with big cap (some uf , 470uf for example) , so in tx coil I have high p-p amplitude on any frequencys.
      For rx I connected rx coil in 1/2vcc virtual ground and positive input of opamp without any caps that made resonance with rx coil, I corrected parts around opamp that made bandpass filter and I cut out output antialiasing filter at all.
      so rx has working at any frequensys too.

      I replace quartz to 16mhz so quasar arm now working from 8 to 41khz!

      and as a result I used any frequency for working , tested sensors is garrett ace dd for ace euro/ace 350 1.74mh/2.6mh, karma 13 for fisher f75 6.8mh/0.68mh that hasnt caps so it has only rx tx coils like sensors for tx850.

      Multifrequency firmware for quasar arm is very interesting for testing.

      Comment


      • #4
        Notes on the Quasar ARM hardware

        1. No real-time TX current sensing. The board measures average TX current only, which is enough for resonance search. What it lacks is high-resolution, real-time TX current sensing on a second ADC channel sampled synchronously with the RX voltage, at the same sample rate. That would give the I/Q demodulators a true phase reference locked to the primary magnetic field of the TX coil, and would make the ferrite-calibration workarounds unnecessary.

        2. RX preamp. The current op-amp is adequate but not optimal. An AD8656 should give lower RX noise. The 100 kΩ feedback resistor is also a significant noise contributor, because resistor thermal noise voltage scales as √(4kTRB), so it rises with resistance and temperature.

        3. TX driver. The discrete 0/5 V TX driver is overly complex. An integrated half-bridge gate driver or a full-bridge DC motor driver IC would simplify the schematic considerably.

        4. Power supply. An LDO-based supply is inefficient for a lithium-battery system.

        Multifrequency-specific

        5. The 2×16 character LCD can show only two frequency signagraphs at once. The display options are therefore F1 (4–11 kHz), F1+F3 (6+18 kHz), and F1+F5 (6+30 kHz).

        6. Free choice of operating frequency (as opposed to optimizing sensitivity) requires non-resonant TX drive. At 0/5 V levels, though, non-resonant TX current would be very low. As a workaround I used a dual-resonant Foster network. The firmware has F1, F3 and F5 I/Q demodulators and processes all three harmonics in real time, but on this hardware only two frequencies are usable. A redesign would use a half- or full-bridge TX stage powered from a boost DC-DC converter (12–48 V). That would enable non-resonant drive and simultaneous operation on more than two frequencies.

        7. A differential RX preamp would reject common-mode crosstalk coupled from the TX conductors to the RX conductors in the coil cable.

        8. The TX drive could also be differential. The open question is TX/RX compatibility with the range of existing coils.

        9. I would replace the MCU with an STM32G4 series part.

        For now I accept the hardware as it is and focus on firmware only.​

        Comment


        • #5
          Firmware test targets: 1×1×1 cm cubes of iron, lead, tin, brass, zinc, aluminium and copper; US, Chinese and Korean coins; bottle caps, pull tabs and ferrite.
          Click image for larger version  Name:	20261101_155338.jpg Views:	0 Size:	614.2 KB ID:	454173
          The menu three:
          Click image for larger version  Name:	Screenshot 2026-11-01 185201.png Views:	0 Size:	100.9 KB ID:	454174


          If you own Quasar ARM hardware and would like to test Reborn-DF FW beta versions, please contact me by private message.​
          Last edited by Byko; 11-01-2026, 11:47 AM.

          Comment


          • #6
            5-inch Gold Finder coil adapter 3d model:



            Click image for larger version

Name:	MD_Adapter_5-inch_Coil.png
Views:	229
Size:	73.6 KB
ID:	454264 Click image for larger version

Name:	20261105_212727.jpg
Views:	208
Size:	504.8 KB
ID:	454265



            ​

            Comment


            • #7
              Hello! I flashed the unit without any issues. I connected the coil, went to the CALIBRATION menu, and started the process. Scanning began, but then an error appeared: "Tune at edge." I did not modify the input stage. My coil is a 14.4 kHz resonant type. Coil parameters in the archive. There is a capacitor in series with the TX [coil], and another capacitor in parallel with the RX [coil]. Both capacitors are located inside the coil assembly. The capacitance values ​​have not been preserved.
              Attached Files
              Last edited by ArchibaldSTM; Yesterday, 01:35 PM.

              Comment


              • #8
                I’m glad you flashed your hardware successfully.

                TX frequency range in the current release: 2,000–11,029 Hz (F3: 6–33 kHz, F5: 10–55 kHz).
                • Where the limit comes from: the ADC tops out at 176 kHz. The firmware takes 16 samples per TX period, and the smallest timer divider is 68. That gives 24 MHz / (2 × 68 × 16) = 11,029 Hz.
                • Above about 5.9 kHz: the processor can’t handle one 8-period block per DSP pass (740 blocks/s at most). The firmware then sums two blocks into one, so at the DSP level 11 kHz runs like 5.5 kHz. The harmonics are always multiples of F1: at 11 kHz, F3 is about 33 kHz and F5 about 55 kHz.

                I looked at two ways to support your coil:
                • A. Faster ADC sampling. Cut the ADC sample time from 7.5 to 1.5 cycles, so each sample takes 14 cycles instead of 20. The ADC ceiling rises to about 250 kHz, and the top TX frequency to about 15.6 kHz. Only the driver changes.
                • B. 8 samples per period. The top frequency rises to about 22 kHz, but with only 8 samples per period F5 folds onto F3, so it works for F1 + F3 only.


                Experimental firmware for your coil

                I’ve built an experimental firmware with the TX frequency raised to 15 kHz:
                • It’s the latest v0.4.63 with one change: the highest TX frequency goes from 11,029 Hz to 15,000 Hz. v0.4.63 measures the battery and TX current in a gap it makes in software after each block of RX samples. That gap costs time, which limits the sampling rate and therefore the TX frequency. In the experimental build a hardware timer starts that measurement between two RX samples, so sampling can run fast enough for 15 kHz.
                • Nothing else changed. Detection, VDI, menus, calibrations, stored settings and the GUI link are the same as in v0.4.63.
                • The screen shows the version as 4.63+. The “+” marks a build outside the release line. GUI2 v0.4.63 connects to it normally and shows “fw 0.4.63+” in its top bar.
                • It’s not a release. The next release will include this change only if it proves itself on the board and in the field.

                Once I’ve tested it successfully, I’ll share it with you.

                About this release

                This release has many improvements; the full list is in the release notes.

                The default motion high-pass filter has changed from 4 (similar to Quasar ARM) to 2, which gives better sensitivity. If the ground isn’t suppressed well after auto threshold (long press UP), please try raising the filter setting before you raise the threshold.

                Regards,
                ​

                Comment


                • #9
                  Okay, I’ll wait. Thank you for the quick response!

                  Comment


                  • #10
                    I’m still working on a bug in the ferrite calibration. The release was due now, but this bug has pushed it back. A quick heads-up: the GUI will ship with the upcoming release:
                    Click image for larger version

Name:	Screenshot_20261119-085512.png
Views:	0
Size:	93.4 KB
ID:	454692

                    Comment

                    Working...
                    X