Connect with us

AM Radio Design Log

Discussion in 'Electronic Design' started by M. Hamed, Jul 16, 2013.

Scroll to continue with content
  1. M. Hamed

    M. Hamed Guest

    Gives me an idea of a new FM modulator element, the capacitor modulator. Changes in audio voltage produces equivalent changes in capacitance and frequency changes accordingly :)
  2. M. Hamed

    M. Hamed Guest

    Actually I did both. But measurement showed about 100 nF capacitance which is strange because that would have changed the frequency by much much more than the observed change. I also tried the calculation method which produced a capacitance of a few pico Farads.
    Yes, I did hence I reversed the coil so the gain isn't so much.
    I was trying to calculate the output impedance of the circuit but failed.
  3. Glenn

    Glenn Guest

    Is this mixer better than the Gilbert cell?

    The Tayloe Sampling Mixer - also called:
    Tayloe Detector
    IQ based tranceiver
    linear detector
    Quadrature Sampling Detector QSD

    Actually it is a "bang-bang"/digital mixing that ought to produce a lot
    of aliasing products.

    It seems to have something in common with the diode ring mixer:

    What you really need is an analog multiplier, but this is emulated by a
    Gilbert cell - and maybe a diode ring mixer?:

    But then how do you theoretically optimize the diode ring mixer for low
    IP3 and other "false" mixing products? "Of cause" you can not suppress
    the signal mirror unless IQ-mixing is used.

    Actually another techniques is used with e.g. a harmonic diode mixer -
    or you are just "tuning" for the "false" product :) :

    I happened to find the harmonic diode mixer techniques in the HP 141T's
    8555A input mixer - and it can be a "bad" techniques because of
    aliasing, that is why an external preselector is needed in front of the
    harmonic mixer.


    Gilbert cell:


    James Long, Ph.D., P.E. Analog and RF Consulting Engineer
    History of Mixer Technology
    (And Other Stories of Misplaced Credit)
    Things are seldom as they seem. Skimmed milk masquerades as cream.
    Quote: "...
    All of mixer technology in use today was mathematically known several
    hundred years ago and reduced to practice by 1935 when the DPDT vibrator
    modulator was used in low frequency amplifiers. By 1949 this function
    was performed by the 7360 vacuum tube in radio circuits and six vacuum
    tubes in rocket telemetry circuits.

  4. Glenn

    Glenn Guest

    That is cool!


    Somebody has used a superconductor cavity to trap and detect a single
    microwave photon - they say:

    Centre National De La Recherche Scientifique (2007, April 2). Life And
    Death Of A Photon 'Filmed' For The First Time. ScienceDaily. Retrieved
    March 22, 2008:
    Quote: "...
    A photon is an elementary particle of light. In general it can only be
    observed when it disappears
    The end result is that the atom changes to state 1 if the cavity
    contains a photon and remains at state 0 if it is empty, as in the
    standard method. However, this time the energy absorbed by the atom is
    taken from the auxiliary field and not from that of the cavity. As a
    result, the photon is still there after having been seen, and is ready
    to be measured again
    Suddenly the atoms appear in state 1, showing that a photon has been
    trapped between the mirrors. The photon comes from the residual thermal
    radiation which surrounds the cavity
    The moments at which the photons appear and disappear reveal the quantum
    jumps of light, which occur at random


    Maybe you can detect its OAM modulation? ;-) :

    Nov 1, 2012, Spooky action with twisted beams:
    Quote: "...
    Using this technique, Zeilinger and co-workers found they could obtain
    differences in quantum number as high as 600 (in other words l = +300 on
    one photon and l = –300 on the other). Lapkiewicz points out that there
    is, in theory, no upper limit to a photon's l value, which suggests that
    a photon – a quantum object – could acquire as much OAM as a macroscopic
    object, leading to what he calls a "tension between the quantum and
    classical worlds"

  5. Tauno Voipio

    Tauno Voipio Guest

    The Gilbert cell is a four-quadrant analog multiplier,
    which should be the best mixer, but due to the imperfections
    in the multiplication gives often worse results than a raw
    bang-bang switching mixer (diode bridges, rings, 7360 beam-
    deflection tube, CMOS analog switches).
    It does, but it it very linear for the signal channel.

    There are two things in the mixing which should be
    considered separately:

    1. How to get an effective analog multiplication and
    linearity, at least for the signal input;

    2. How to suppress unwanted mixing responses.

    In practice, point 1 is difficult to achieve, especially
    for a large input amplitude range. For many practical
    purposes it is sufficient to multiply with sign of the
    oscillator signal, which is much easier to implement
    linearly for the signal channel.

    The quadrature methods (phasing, IQ) attack question 2,
    by implementing a full complex number multiplication
    (called analytic signal in DSP theory).
  6. Robert Baer

    Robert Baer Guest

    * --------------------^ Stirling
  7. John S

    John S Guest

    1.8 MHz. I changed to C0G (NP0) and it cured the problem.
  8. M. Hamed

    M. Hamed Guest


    I realized how much fun oscillators are and how Sine waves are so beautifulso I decided I'm going to up it a notch and take a shot at a 100 MHz oscillator.

    I tried a Colpitts variation of the Hartley oscillator I posted earlier with the same MPF102 JFET, a diode, 1 MOhm Res, the tank, and its surrounding capacitors. Inductors and Caps had to get much smaller. So L is in the nH range and C is in the low pF.

    Couldn't get the thing to oscillate at all no matter how hard I tried. I tried too many things to keep track of. I was suspecting the inductance is solow and the inductor Q is inadequate. I tried different types of air-core wire turns. Eventually I moved to the Toroid version but no luck.

    Next, I decided a JFET isn't the thing for this and decided to move to BJT.I used a 2N3904. I played a lot with the number of turns and capacitor values until I reached a circuit similar to this:

    My L is 3-5 turns on a 68-6 toroid. The value on the circuit is just a guess based on the frequency I'm getting. I had the following observations:

    - I had a big 300p cap for C2 this killed the ability to oscillate (later Irealized it had to do with the amount of feedback. If you reverse C1 and C2 however, simulation shows the voltage swing is low. I am guessing it has to do with transistor gain. Increasing C1 should probably also mean decreasing R3. I am taking shots in the dark here, I will try to do some analysis later).
    - Even with C2=300p, increasing C3 to 66p restored oscillation.
    - After reducing C3 back to 33p, I was able to reduce C3 back to 33p and get oscillation.
    - I kept on trying to reduce inductance to increase frequency but that camewith a price of reducing the voltage swing at the tank.
    - I still do not have a unified theory of how the values of L and Cs affectfrequency and voltage swing but it seems I should strive for having higherL for better stability and voltage swing. This means I need to reduce the C's to as low as I can get.
    - In the end I was able to reach 67 MHz as the max I could squeeze out of my circuit but it's not very stable.
    - Even with everything laid out on a perf-board and with no power supply bypass you still can see 67 MHz of oscillation that is clean looking on a scope. Stability is another matter.
    - Soldering and de-soldering is a pain when you have try different things.
    - a plain vanilla 2N3904 can go long ways!
    - I don't like JFETs. I love BJTs.

    Next steps:

    - Add more turns to my toroid
    - Try a better a toroid core, may be a 2 or 6 Mix
    - Reduce caps to the minimum possible.
    - Hope for 100 MHz.
  9. Fred Abse

    Fred Abse Guest

    You've stumbled upon "resonant impedance". There was some discussion about
    that, here, a while ago.

    To a first approximation, the impedance seen across an LC circuit, with
    loss only in the inductor, at resonance is:

    L/CR, where R is the inductor's series loss resistance (higher than at DC
    because of skin effect).

    Simply put, the higher the L/C ratio, the higher the voltage across the
    oscillator tank.
  10. M. Hamed

    M. Hamed Guest

    I went back to the thread. So in essence it's the equivalent parallel resistance of the inductor and we want to keep that high I presume by increasing the unloaded Q of the inductor which can be achieved by increasing inductance.
  11. M. Hamed

    M. Hamed Guest

    At the gate of the JFET in parallel with a diode. Similar to the Hartley circuit I got from EMRFD and linked to earlier in the thread (dropbox).
    At the moment I have a few MPF102 JFETs. The datasheet labels it "VHF amplifier" so I assumed it's adequate.
    Interesting idea. I'll try running some simulations.
    I was suspecting that is the case. The T68-6 is recommended up to 40 MHz. Ihave a few toroids on order that are recommended for higher frequency and have lower AL (mix 10). I started with an air form but when it didn't oscillate I thought my inductance must be too low to be usable. After all the iterations I went through I am going to go back and reexamine this. Thanks for the hint.
    Will do!
    Here is the dilemma. For 100 pF @100 MHz, L is around 25 nH. This is very small. How is that more stable when it's susceptible to parasitic lead inductances and also am I not supposed to keep L/C ratio high?
    Oops. Yes I was. Not the counter but I was using a scope. I figured out a few hours ago. More details in my next post. PS: I do have this wonderful pieces of engineering!

    Thanks for the time you took responding!!
  12. M. Hamed

    M. Hamed Guest


    Went ahead with my plans of increasing L and reducing Cs. Started with a relatively larger inductance, and reduced the tank cap to 6pF and the other caps are 6, 10, 10 pF. I got around 40MHz then I progressively reduced the number of turns on the inductor. At some point slightly above 66 MHz, oscillations seized.

    Then it occurred to me what I should have known better. All this time I've been probing the tank with my oscope probe. Effectively I'm adding a cap inparallel that would shift frequency and hurt the L/C ratio, especially when my tank caps are 6 pF.

    After this simple realization, I added a JFET source follower and connectedthe scope to the JFET source, and voila! My scope showed around 80 MHz. Itwas a matter of shaving off one extra turn from the inductor and there I got 108 MHz! Goal achieved.

    The JFET datasheet lists input capacitance as 7pF @ 1 MHz and 15VDC. So my guess is that the tank resonant frequency is probably around 200 MHz without the JFET load!

    At some point I got concerned that I could be generating interference and my little oscillator could be acting as a transmitter. Sometimes I cranked up my power supply voltage to 18V and with the supply showing less than 1 mAdrawn, that's 18mW which I thought should be nothing.

    To test the theory I tuned my FM radio to 108 and I got some ugly sounding noise. That got me scared a bit that I could be jamming something somewhereso I immediately turned off my supply. With such low power and at 100 MHz I don't expect my interference would travel very far. But at least now I understand the importance of shielding.
  13. M. Hamed

    M. Hamed Guest

    I will probably invest in one of these little fellows:
  14. Fred Abse

    Fred Abse Guest


    With high Q inductors, L/C dominates the resonant impedance.

    R is equivalent *series*, not parallel resistance, BTW, ie. R=wL/Q

    If you've found the previous thread, see my second posted Spice simulation
    to that thread (the current-fed, constant Q, one).
    That demonstrates dependence of Zr on L/C.
  15. Fred Abse

    Fred Abse Guest

    That's misleading. The Q of each circuit, as drawn, is not the same; the
    difference in bandwidth is due to Q, not L/C ratio.

    The effective Q of the 220pF circuit is 38.
    The effective Q of the 22pF circuit is 3.8.

    That's a 10:1 bandwidth ratio, right there. Nothing to do with L/C ratio.

    Assuming the same wire gage, a 1.84uH inductor will have sqrt(1000) times
    the resistance of a 184nH coil. That's about thirty times. You used the
    same resistance in both cases.

    Your simulation doesn't prove your case, regarding "bigger C", which is
    wrong, anyway.

    Now go figure how to simulate L/C ratios properly...
  16. josephkk

    josephkk Guest

    Hooray. You are actually doing very well. Good enough to make jealous
    of all the fun you are having. Keep going.

  17. YD

    YD Guest

    Congratulations, you've just discovered the buffer amplifier, to
    isolate the oscillator from the load. From there you can take the
    signal and do something with it without disturbing the frequency.
    And now you know to take parasitics into account. Good work!
    At that power it's probably detectable at some tens of metres at most.

    - YD.
  18. YD

    YD Guest

    Late at night, by candle light, "M. Hamed" <>
    penned this immortal opus:

    So recalculate for some more reasonable L, like 60 nH. You probably
    already know, but when you have two of f, L or C you can calculate the
    third by

    f = 1/(SQRT(2 * pi * L * C))
    L = 1/(4 * pi^2 * f^2 * C)
    C = 1/(4 * pi^2 * f^2 * L)

    For winding the coil, do a search on "inductor calculator", there's a
    lot of them on-line.

    - YD.
  19. M. Hamed

    M. Hamed Guest

  20. M. Hamed

    M. Hamed Guest

    It turned out connecting my cheap frequency counter is the source of this noise. It probably totally messes up the biasing. When I don't have that connected, I get exactly the effect that you mentioned. It's pretty cool when the frequency on my radio matches the frequency readout on my scope!
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day