VCO Design using Ceramic Resonator

[Fuzzbrain]

I am trying to implement a VC(X?)0 design at 503.5kHz using a ceramic resonator. The 503.5kHz clock will drive a PLD that develops TV timing pulses. The PLD works fine, but the oscillator design is the hang-up.

Ceramic resonators have been used in this application for years. Sync processing ICs using ceramic resonators can be found in TVs and VCRs. The typical application consists of 3 IC pins, one an output and two (differential?) inputs. The ceramic resonator is tied from the output to one of the inputs, and an RC (phase shift?) network spans the inputs. For reference, see data sheets at NJR or Mitsumi.

For various reasons, I cannot use one of these ICs, but would like to implement this type of VCO. Can anyone steer me towards a paper or article that describes the operation of this circuit? Or, if you know how it works and/or how to optimize it, please expound! Thanks!

 

[Lewish]

See if this helps:
For an oscillator to oscillate, you need amplification and feedback. Your amplification gain must be greater than 1. For some oscillators it must be considerably more than 1. For example, a WIEN-Bridge oscillator requires a minimum gain of 3 to work. Also, you must have 360 degrees of phase change from the output to the input, plus or minus about 15 degrees. In other words, you tie the output back to the input with no phase shift. In most ceramic resonators, you get 180 degrees of phase shift. Check to see what it is with the one you are using. It might not be 180 degrees. Therefore, when you tie the output back to the input thru the resonator, you must add another 180 degrees of phase shift to get your phase back to the proper relation.
If you meet all of these requirements, you will probably have your oscillator. But, life is usually not so kind, and you have to help the design along. I find oscillator design to be a real pain in the butt to get to work the way I think is acceptable.

Best of luck to you.

 

[Fuzzbrain]

Thanks for your response. Your points are well taken, and are correct insofar as they go.
The vital parameter of my design is "voltage controlled". This is the equivalent of a VCXO in the higher frequency crystal world. While there are several circuit topologies that can achieve this, most rely on varactor diodes. At this frequency, high-capacitance varactors are required to get a reasonable control range. The fosc/Cv function also tends to be quite non-linear.
Hence my interest in the mystery IC VCO design. As I said in my last post, this design has been commonly used for the past 20 years, and can still be found in current products. Ic manufacturers are notoriously secretive about the innards of their parts (lest I copy them?), and I haven't found a schematic of a VCO of this type using a ceramic resonator. However, an ancestor of this circuit can be found in an old RCA Linear part, CA3154. This part contains a 3-pin VCO that operates at 32fH (503.5kHz) using a series-LC tank (one of the resonant modes of the ceramic resonator). Moreover, the RCA Linear IC manual contains a complete schematic of the IC, but no theory of operation!
From the CA3154 internal schematic, I see that the oscillator is apparently a Gilbert Cell multiplier. One of the outputs drives the series LC tank through a RC phase shift network (~90 degrees?) to one of the inputs of the bottom transistor pair. Across the two bottom pair inputs is another phase shift network (~90 degrees?). The input to the upper transistor level is the control voltage, which comes from the phase detector through a LPF.
As an experiment, I did try to implement this circuit using an SA602, and it did work! How does it work? My guess is that the Gilbert Cell forms a voltage-controlled phase shifter. But how to optimize it? I have no real clue, and no time available to re-invent this wheel. Moreover, ceramic resonators for this type of service form a sub-group with the letter "F" and a number following it. Each F-number type is intended for use with a specific IC (for example, CSB503F10 is designed for use with Toshiba TA7777P). No electrical characterization of these parts is given, but they have Fr and Fa frequencies significantly higher than the standard resonator for the same frequency.
If anybody can shed any light on this best-kept secret in analog IC design, please do so. I appreciate your help.

 

[Lewish]

If it is a Gilbert cell, then Barrie Gilbert is the person to answer your questions. I have found him quite willing to describe circuits, although I quite often didn't follow all of his explanations. He also designed the NA/SA602 for Signetics.

You can contact him at barrie.gilbert@analog.com