Building a JFET voltage-tuned Wien bridge oscillator
Since the derivation of the Wien network by Physicist Max Wien (prior work was the ballistic bridge by J. C. Maxwell) in 1891, this network has been used in many oscillators’ designs. One of the most famous was the Hewlett-Packard 200 oscillator. What made this sine wave oscillator design different from most of the designs from that time, was the use of an incandescent lamp in the negative feedback divider to stabilize the gain required for constant amplitude level and stable oscillator operation over a wide frequency range. This idea was Professor Fred Terman’s suggestion for Bill Hewlett’s Masters Thesis at Stanford. Bill Hewlett’s audio oscillator became rather famous and one of Bill & Dave’s first HP products. Prior to the introduction of Bill Hewlett’s Wien bridge oscillator most audio frequency sources were LC Oscillator based.
The General Radio (GR) Company had been producing their Audio Oscillator initially with the Model 413 since the late 1920’s which later evolved to the Model 13042. This GR Audio Oscillator consisted of two LC-tuned oscillators. One was fixed frequency, the other oscillator was tunable. The outputs from these two oscillators where mixed. The resulting frequencies that were not in the oscillators’ specified range, were filtered out with the selected frequency band of interest amplified along with the amplitude leveled and controlled to the oscillators’ output terminals. This was an expensive and complex approach to producing an audio signal source. There were difficulties with frequency stability as the LC tuned oscillators would drift. This problem of frequency stability at the lower audio frequencies became a never-ending problem for these LC tuned audio oscillators. Harmonic distortion was also another very significant difficulty that they exhibited.
The fixed or single-frequency Wien bridge oscillator is not difficult to design and build, as the RC components used in the Wien bridge network can be matched to achieve the desired frequency of oscillation.
The amplitude control stabilization can be achieved by using the light bulb thermal-based scheme introduced by Bill Hewlett. Wien bridge oscillators of this design can be made to achieve very low distortion with good amplitude and frequency stability.
Variable frequency Wien bridge oscillators are more difficult to design and build due to the requirement of a matched pair of components in either the resistive or capacitive arms of the Wien Bridge network. This can be achieved with a tracking dual variable capacitor or switched capacitor values, tracking dual variable resistors, switched resistors, logic selected resistors (Digipot) to alter the frequency.
Voltage-tuned Wien bridge oscillators are not common and they can be the most challenging to design and build. There have been varactor diode-based (voltage-variable capacitor) Wien bridge voltage-controlled oscillators designed and built (Jim Williams Figure# 33.114, Analog Circuit Design: A Tutorial Guide to Applications and Solutions5). Another method to voltage-tune a Wien network is to use a pair of matched JFETs.
Linear Systems introduced a dual matched P-channel JFET which inspired the idea of using this matched device as a tracking voltage-controlled resistor in the Wien network. This idea has been patented (# US3432774A)4 1967 by Oliver A. Fick with the Atomic Energy Commission. There are drawings and descriptions in this patent that describe the concept and idea, but as with most patents the finer details of how to make the idea actually function are nebulous and not detailed enough to produce a design that works well enough in the real world. During the many years of tinkering with test gear, gleaning much information from their service manuals and looking on the web for examples of a JFET-tuned Wien bridge oscillator, no viable designs had been found (does not mean none have been designed and produced). This idea became an intellectual design curiosity to design and build a functional JFET voltage-tuned Wien bridge oscillator.
The Wien bridge is essentially a series-shunt RC network that results in zero degree phase shift when the series and shunt RC networks reach equilibrium. At a zero degree phase shift, the network becomes essentially a resistive voltage divider that can be used to deliver positive feedback into an amplifier to produce oscillations at that specific frequency.
This is a Wien network analyzer plot example using a Hewlett Packard 3577B network analyzer.
The Wien network values are 1000pf for the capacitor and 301 ohms for the resistor. Zero phase occurs at a calculated frequency of 512 kHz; the Network Analyzer displays close to 513 kHz. Wien Network loss, about 9.5 dB or a factor of 3 reduction. This is the amount of energy loss the amplifier or similar gain device will need to make up in order to produce oscillation when the Wien network is inserted as positive feedback in an oscillator. There will be a tendency for the gain device to increase oscillation amplitude until gain device saturation occurs. Stable oscillator output amplitude and oscillation frequency is achieved by adding a variable gain, negative feedback loop controlled by a reference, to level and balance the amount of positive feedback via the Wien Network.
Reducing the resistance, or R value, of the Wien network’s series arm, results in a slight magnitude increase in the network analyzer display at equilibrium and a slight upward shift in frequency.
This slightly increases the amount of positive feedback from the Wien network when used as an oscillator. To maintain constant level output from the oscillator, the negative feedback loop will need to adjust to and compensate for these dynamic Wien network tuning component changes.
This is an example of one parameter that needs to be considered when a JFET is used as the voltage-controlled running element in a Wien Bridge Oscillator.
JFETs, when used as voltage-controlled resistors, will have essentially square law non-linear transfer characteristics similar to a Triode vacuum tube. This JFET personality is more pronounced as the signal excursion levels are increased upon the JFET’s drain and source, which results in channel length modulation (channel resistance modulation). These small variations in drain-to-source resistance appear as curvature on both edges of the triangle wave across the JFET under test in the image below.
One half of the dual LSJ689 is configured as one resistor in a two-resistor voltage divider, driven with a 10-volt peak-to-peak, 500kHz triangle wave trace on the bottom; Gate Source bias voltage is about 30% of the JFET’s drain-to-source near maximum source-drain resistance.
The square law characteristics of the JFET become visible as curvature of the once linear edges of the triangle wave trace at the top.
Most JFETs are often symmetrical between drain-to-source. By adding a voltage divider between the drain-to-source with the gate connected to the center of this voltage divider, the divider is effectively applying corrective feedback to the gate of the JFET. Some of the non-linearity due to channel length modulation can be corrected as illustrated below.
Adding feedback from a resistive divider between drain to source of the JFET to the gate helps linearize the FET.
The voltage divider drive increased to 15 volts peak-to-peak to illustrate the improvement in drain-to-source resistance linearity across the JFET under test.
JFET distortion due to gate modulation from the voltage drop between drain and source can be further reduced by lowering the signal level across drain-to-source. This effectively reduces the channel length modulation effect resulting in a more linear resistance between drain and source of the JFET.
Top trace, Output Triangle wave across the JFET under test
Bottom trace, Triangle wave input to the JFET test divider