Connect with us

Very low volt osc.

Discussion in 'Electronic Design' started by amdx, Jun 2, 2004.

Scroll to continue with content
  1. amdx

    amdx Guest

    I need a 1Khz osc with a low recpricating voltage, appox. 30 mv.
    The 1Khz needs to be stable, I need to start with a crystal osc. and divide
    down. When I get to 1Khz the voltage can only swing 30 mv. I don't
    want any large voltages at 1Khz because of a very high gain amp in
    a later stage. ( this osc signal can't get into the later stage.)

    Any ideas how to develop a stable low voltage 1Khz signal?

  2. Activ8

    Activ8 Guest

  3. Tim Wescott

    Tim Wescott Guest

    "microcontroller averse" method:

    Use a mixer, like an NE/SE/whatever-612, a precision 4-quadrant
    multiplier, or some 405x switches. Use a crystal frequency that will
    divide down to two frequencies that are 1kHz apart (eg 4.032MHz/63 =
    64kHz, /63 it equals 64kHz). Attenuate one input to the mixer. For the
    example frequency you'll get 1kHz and 127kHz, so you can easily filter
    out the 127kHz.

    "microcontroller friendly" method:

    Use a multiplying DAC, and feed it a sine table from a microcontroller
    (or from a DDS, which will have to be controlled by a microcontroller).
    Make sure to use a reference frequency that isn't a multiple of 1kHz.
    Give it a _small_ reference voltage, and you can get whatever output
    signal you want.
  4. Why? This would simplify the table for the DAC.

  5. amdx

    amdx Guest

    Hi Activ8,
    I want to avoid the large signal in the first place, I'm afraid the large
    will couple into the very high gain amp in a later stage.

  6. Tim Wescott

    Tim Wescott Guest

    Because the OP has specified that he wants no high-level 1kHz signals
    anywhere in the system, due to a high-gain amplifier later in the chain.

    Had I thought of it last night I would have mentioned that you could get
    around that by adequately shielding an appropriate divider chip (perhaps
    going so far as an enclosure, although at 1kHz that may not be
    necessary) and attenuating it's output inside the shielding, so the only
    thing that escapes would be the desired low-level signal.
  7. Tim Wescott

    Tim Wescott Guest

  8. MikeM

    MikeM Guest

    How about synthsizing it directly using a rom and dac? The clock
    used to step through the lookup table memory can be high frequency
    relative to 1Khz.

  9. Fred Bloggs

    Fred Bloggs Guest

    It can't be done with a crystal- these have minimum drive requirements
    to develop reasonable Q- a very non-linear phenomenon, and 30mv doesn't
    do it.
  10. How about two oscillators using standard 31.5kHz and 32.56kHz Epson
    crystals- mix the signals then LPF giving a stable 1.06kHz?

    Best regards,
    Spehro Pefhany
  11. Fred Bloggs

    Fred Bloggs Guest

    The 32K is a 2^15 so a divide by 2^5=32 gets 1.024KHz- that's a 74HC4060
    type thing...
  12. ??? Any info to back that up ???
    Never expected Xtals to be non-linear devices. A hysteresis kind of thing?

    Arie de Muynck
  13. Tim Wescott

    Tim Wescott Guest

    And one that certainly runs counter to my experiences with crystal
    filters in radios -- those things efficently filter signals in the 10 -
    100uV range just fine.
  14. Yes, I thought the nonlinearity was introduced by the amplifier as it
    nears saturation- thus limiting the amplitude- and the oscillation
    actually starts from excitation of the crystal by the noise in the
    amplifier. And I've used crystals even down in the 60kHz range for
    signal filtering.

    Maybe a 1kHz crystal would be different.

    Best regards,
    Spehro Pefhany
  15. I think that high Q and nonlinear are contradictory concepts.
  16. Fred Bloggs

    Fred Bloggs Guest

    It is not an instantaneous non-linearity- it is hysteretic one with
    memory. Parzen has graphs of Xstal equivalent ESR variation as a
    function of RMS drive level for some types, it varies over a few orders
    of magnitude- it is the main reason for oscillator start-up problems
    where you have too much loss at low drive levels. The crystal L-C-R
    equivalent circuit is only valid over a range of drive levels and when
    the material is in steady state resonant vibration- outside of that it
    is dead lump of quartz with some shunt capacitance.
  17. I dont.

    Kevin Aylward
    SuperSpice, a very affordable Mixed-Mode
    Windows Simulator with Schematic Capture,
    Waveform Display, FFT's and Filter Design.
  18. Fred Bloggs

    Fred Bloggs Guest

    Ooookay- you asked for it- the increase in motional resistance at low
    drive level is known as "second level of drive effect."

    Bernstein,M. (1967) Increased Crystal Unit Resistance at Oscillator
    Noise Levels, Proc. 21st Ann. Freq. Control Symposium, US Army
    Electronics Command, Ft. Monmouth, N.J., 244-58


    Nonaka,S., Yuuki, T., and Hara, K. (1971). The Current Dependency of
    Crystal Unit Resistance at Low Drive Level, Proc 25th Ann. freq. Control
    Symposium, US Army Electronics Command, Ft. Monmouth, N.J., 139-47

    If you arrange the circuit to start up at high drive level and then back
    off to low drive- you end up with what is known as a "sleepy crystal"-
    and this introduces more complications with stability and
    intermodulation effects not present at conventional drive. An added
    complication is that the crystal behavior in these modes will be a
    strong function of various arcane characteristics of the manufacturing
    and quality control process.
  19. Or "drive level dependence".
    Thanks for the references.

    Okay, here's something on-line that mentions those effects:

    I see that some manufacturers indicate they test drive level
    dependence based on IEC444-6, and indicate that this kind of behavior
    is indicative of a *poorly made* crystal (bad surface finish, poorly
    adhering electrodes etc.). Worth looking into when surveying crystal
    suppliers that they have the equipment and perform those tests.

    But, unless the resistance increases a LOT, 30mV is a pretty healthy
    signal, no?

    Best regards,
    Spehro Pefhany
  20. Fred Bloggs

    Fred Bloggs Guest

    Well- the 30mV across a typical 50K motional resistance of a 32.768K
    would be 18nW- this seems rather small- that circuit almost certainly
    will not start- the only hope would be to start up at 3V and then try to
    back off.
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