I found myself in need of a very low-distortion 1,000 Hz sine wave
generator for an upcoming project.
A quick search on Ebay found several used pieces of test equipment that met or exceeded my requirements, but were priced at well over $1,000 USD.
I did also find some low-cost China sourced 1 kHz generator boards, but they had a distortion of 0.1% which was way above my 0.005% limit.
My current test gear includes a Keithley 2015 THD Multimeter with a low distortion sine wave generator, but it is only capable of 0.03% THD at 1 kHz.
It was time to see if I could build something for under $100 that would meet my needs.
After breaking out some of my old audio design books, I noticed a design for a Wien Bridge oscillator that used diodes to control the signal amplitude. After a bit more research I found a myriad of ways to control amplitude and reduce distortion.
A method described by Larned A. Meacham in 1937, using a filament lamp for automatic gain control was used by William Hewlett, and David Packard, and found its way into the HP200A audio signal generator, the first product of Hewlett-Packard.
So I decided to go down the Wien Bridge Oscillator rabbit hole.
My first breadboard prototype used a dual low-noise JFET-Input TL072 operational amplifier and a 12 volt automotive light bulb. The THD was around 0.75% and the breadboard layout was contributing to some of that. A printed circuit board for the TL072 along with some different passive components and a 14 volt filament bulb, got me down under 0.18% THD.
While researching low-noise op-amps, I stumbled across the Texas Instruments OPA16XX series of SoundPlus operational amplifiers. These op-amps had some very impressive specifications, including a low of 1.1‑nV/√Hz noise density with a distortion of 0.000015% at 1 kHz for the OPA1612.
I designed and ordered a set of boards that would accommodate some different sized passive components, along with an OPA1612A and a couple of OPA1662 op-amps.
After experimenting with several combinations of op-amps, lamps, and passive components, the board with a combination of the OPA1612A along with low-noise precision thin-film resistors, and film capacitors produced the lowest THD readings.
They were so low that they exceeded the capabilities of my Keithley 2015 THD Multimeter.
The circuit uses the series and parallel RC pairs along with two 14 volt filament light bulbs in L1 and L2 positions. I included trimmers in both the series and parallel circuits and another trimmer for the negative feedback adjustment.
The second section of the op-amp is set for a gain of 10, and the output amplitude is adjusted using VR4 a 10K ohm variable resistor that can be externally mounted.
The circuit uses an external low noise power supply of +15 and ‑15 volts, and includes additional filter and de-coupling capacitors on board.
The two 14 volt filament bulbs are connected in series to effectively make a 28 volt bulb with a large filament mass. I was initially worried that doubling the bulbs would create more microphonics from vibration of the filaments, but the effects of microphonics are almost non existent.