DIY Test Equipment for Audio and Ham Radio Enthusiasts:
An Ultra Low Noise (=<1ppb) Programmable Audio Oscillator

The "Uber Oscillator" -- a design from Linear Technology's AN-67, by Dale Eagar "Using Super Op Amps to Push Technological Frontiers, an Ultrapure Oscillator".  The author reported that the distortion was immeasurable !!!

To see the article on Linear Technology's website, click on "Application Notes" and scroll down to AN-67

 
For the complete description you should refer to Linear's Application Note.  We have produced the schematic here, using our schematic capture program Multisim so that the design could be converted to a printed circuit board.  Speaking of circuit boards, this design uses a quad current-feedback operational amplifier, the LT1230,  in the "Super Gain Block" (first schematic) -- these very fast devices should be used with a printed circuit board.  The "OffPage" connectors are used to link the two schematics in Multisim.  

 

 

 
Construction:  Place and solder in the following order: resistors, integrated circuits, capacitors, trimmer resistors, power and output connectors.  The oscillator should be placed in a shielded chassis, preferably one with good shielding properties like a steel cookie tin.  To get the project started I used a large cookie tin and 10 D-Cells to provide +/- 7.5 volts.  
Testing:  You will need an oscilloscope to test the oscillator.  The first trick is to get it to oscillate -- place the Output Amplitude and 2nd Harmonic trimpots to their 50% positions. With your scope set as follows: 50mV/division vertical, 20uS/horizontal, DC coupling adjust the 2nd Harmonic Trimpot  until the output voltage offsets some tens of millivolts above 0 volts reference.  Then slowly adjust the Gain Trimpot until oscillation kicks in.  You should get anwhere from 700 mV to 7.5 V peak to peak reading, and it won't be pretty.  For a pure sine-wave the you will have to back off the Gain Trimpot and "twiddle" the Output Amplitude Trimpor for a reading of 5V peak to peak.   
Results: We ran a comparison of the Super Oscillator and the oscillator of the Boonton 1120 THD analyzer. Both devices were fed to a Tektronix DA4084 Distortion Analyzer and the notched output displayed on the HP 3577a. Spectrum Analyzer.  Note that the vertical sensitivity of the Super Oscillator graph is 1uV/division, but 2uV/division for the Boonton.  Both plots were done with a 20dB attenuator.  The average noise level of the Super Oscillator is a bit higher than the Boonton, but the harmonics are virtually non-existent.  In fairness to the Boonton, once it settles down for a few minutes the 2nd and 3rd harmonics become very small.