Super Low Distortion Ultra Pure Audio Oscillators Revisited

[mediatechnology]

A poster at diyaudio.com provided the new link for "flip-flop world"

The oscillator link is here: http://flip-flop.world.coocan.jp/audio/ ... index.html

[mediatechnology]

I just built a 10 kHz test version of ULDO-Nacho.

I used Mouser's 5% 1 nF polystyrene caps because I wanted to try them.
I didn't scale the tuning resistors: They remained 15K9 and 15K9/2.

The polystyrene caps I had are the radial version and the long leads allowed me to tack solder them.

The caps ran about 5% high in value so I had to pad the tuning resistors to move it from about 9500 Hz to 10K.
At 9500 Hz "Nacho," the notch filter, couldn't be pulled far enough to get a good null.
Once I got the oscillator moved to 10 kHz I could get a deep null.

I expected the distortion to be about 20 dB higher than the 1 kHz version owing to the fact that the op amp open loop gain falls off at 20 dB/decade.
That proved to be the case with distortion measuring, at +20 dBu output, around -123 dBc or just under 0.7 ppm with a 48 kHz measurement bandwidth. (H2-H4, Fs=96k)

With a 0 dBu output feeding the Focusrite 2i2 A/D, distortion of the A/D measures 0.0002%.
This is far, far better than the 2i2 measures in loopback.

I'm impressed with the 2i2 A/D's distortion at that frequency and pleased with ULDO-Nacho's 10 kHz performance.

[mediatechnology]

Here's' the FFT results of the 10 kHZ Uldo-Nacho and the Focusrite 2i2:



The sample rate is 96 kHz, 24 bit. FFT is a 1M pt 10 sample average.
The output is 0 dBu loaded only by the notch filter and A/D.
The red trace is the notch filter output. The measured values do not reflect the +40 dB gain of the post-notch amp.
(The 30 kHz spur is actually -125 dBc/dBu.)

The hash in the notch filter output is due to long capacitor leads, wrapped construction of the polystyrene caps and lack of sufficient shielding.
I paid no attention to "outer-foil-to-lowest-impedance" on the caps and the board is only shielded by a sandwich of cardboard-clad grounded aluminum foil.

The calculated H2...H4 HD of ULDO-Nacho is -119 dBc relative to 0 dBu or just over 1 ppm.
With the oscillator running at higher level and using a pad to obtain 0 dBu I expect it to measure slightly lower.

The blue trace shows the Focusrite 2i2 (Gen 3) A/D HD at 0.0002% or about -113 dBc.
Note that the H2 is introduced by the A/D.
Overall ULDO-Nacho is about 6 dB "better" than the A/D which says a lot for both of them particularly the 2i2.

I've got some 1 nF 1% stacked polypropylene Wima's in my Mouser cart for my next 10 kHz build.
I'm also going to try 100 Hz.

[SandyTodorov]

Lots of old ideas and nothing new under the sun.
This was done just to test the idea, now an improved version with new operational amplifiers is coming.

Here are the two generators with NE5534 and with ORA1642.
0dB=5Veff 3 more zeros must be added to the THD shown in the picture as a number.
The harmonic level in dB on the picture is accurate. The fundamental harmonic is suppressed by -60dB and the scale is corrected by -60dB.

[mediatechnology]

Great job and thank you for joining us Sandy!

If I've done the mental gymnastics right it looks like you've manged to achieve 28 ppb THD.

Gotta ask: What resistors are you using?

From the noise floor it looks like your notch filter insertion loss is being corrected.