0
\$\begingroup\$

I need to coarse control the frequency of oscillation on a parallel resonant tank. The frequency band is in the 500kHz-1Mhz range, so the small capacitance of varactors doesn't suit the task.

Another way of doing it is by switching a capacitor branch with a MOSFET

schematic

simulate this circuit – Schematic created using CircuitLab

The problem is that the tank is floating, thus the source is not fixed biased and that makes it difficult to control the conduction on the MOSFET.

I've not considered adding a PLL or another stage because it's the coil where I need the output to be and the current in the tank is already pretty high.

A second problem is that even if I can switch the capacitor on and off, the amplitude of oscillation will vary according to the impedance change, and I need to get as stable amplitude as possible.

¿Are there any solutions for this specific arrangement or is it better to use a grounded tank so I can switch the then properly biased MOSFET?

Thanks in advance!

  • Note the values in the schematic are somewhat random
\$\endgroup\$
4
  • \$\begingroup\$ Funny. You used to get varactors that covered the 30-300pf(500pf?) range, to tune the whole AM band, which would comfortably cover your range of interest. If this is a one-off, might be worth scouring the internets looking for obsolete components... A quick search for "AM varicap diode" yields www.nteinc.com/specs/600to699/pdf/nte618.pdf \$\endgroup\$
    – user16324
    Commented Sep 5, 2017 at 11:35
  • \$\begingroup\$ Seems to me current must pass both ways through your MOSfet: it isn't a bidirectional device. \$\endgroup\$
    – glen_geek
    Commented Sep 5, 2017 at 11:51
  • \$\begingroup\$ Even I had high capacitance varactors, there would still the problem with the biasing caused by the voltage swing at each side of the tank. I forgot to mention Vcc is only 5V, and even with face to face varactors I don't know how to reverse bias them if the other side is swinging that much. \$\endgroup\$
    – Angel
    Commented Sep 5, 2017 at 11:54
  • \$\begingroup\$ @glen_geek That's right. I've seen switched capacitor bank arrangements when one side of the bank is at either rail, but I don't know if it can be done with a floating tank. \$\endgroup\$
    – Angel
    Commented Sep 5, 2017 at 11:57

1 Answer 1

0
\$\begingroup\$

I see two problems with that circuit:

1) as you wrote: the biasing of the MOSFET

2) The bulk diode of the MOSFET, when on the tank the voltage on the top node is negative relative to the bottom node, the source-bulk diode in the NMOS will turn on. Oops !

I have a proposal to solve both issues:

schematic

simulate this circuit – Schematic created using CircuitLab

The 1 M resistors bias the NMOSFET's Drains to ground, that solves 1)

You cannot get rid of the source-body diode in discrete MOSFETs so I propose to use two in anti-series. Combined the two NMOS will behave as one NMOS without a body diode.

Instead of biasing the drains to ground you can bias them to any DC voltage as long as Vgs is changed accordingly. Such that there's a positive Vgs to turn the NMOS on and Vgs = 0 V to turn them off.

Note that this only allows you to switch on/off capacitance (C2 in series with C4) so the adjustment is coarse but you describe that in your question so you're aware of that.

If you want finer adjustment, place many of these sections with different values for C2 and C4 in parallel with the tank.

\$\endgroup\$
4
  • \$\begingroup\$ Nice solution. I've simulated it and it works, but instead of biasing the drains, I've biased the sources, which I think I'ts what you were meaning. Anyway, there's no much change if remove the biasing resistors. That leaves me with the issue of equalizing the amplitude at each frequency. \$\endgroup\$
    – Angel
    Commented Sep 5, 2017 at 12:37
  • \$\begingroup\$ Indeed, only biasing the node in between the two NMOS will also work. In your simulator you can remove the biasing resistors but that does not mean they/it should not be there !!! Charge-buildup is not simulated so in the real world the DC voltage will drift away. That is what the biasing resistor prevents. Try giving C4 a 10 % larger value and check if there is a difference with and without biasing resistors. In the simulator the tank's voltage is likely ideally balanced, in practice it will not be so. \$\endgroup\$ Commented Sep 5, 2017 at 12:40
  • \$\begingroup\$ You're right, I've biased it as you say, and it also works on the bench. Now, another problem has arised. The branch switching makes the temperature in the caps go back and forth (there's 40V and 2A p-p) and the frequency is drifting a lot. That makes me think maybe this is not the best circuit for such power in the tank, but that's matter for another question. Thank you! \$\endgroup\$
    – Angel
    Commented Sep 5, 2017 at 13:04
  • \$\begingroup\$ The switched caps are much larger than that 68 nF, the switched branch basically shorts the tank. Better would be 68 nF fixed and 10 nF switched. \$\endgroup\$ Commented Sep 5, 2017 at 13:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.