Super Low Distortion Ultra Pure Audio Oscillators Revisited

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by mediatechnology »

I'm pretty happy with the oscillator but need to do some work on "Nacho," the notch filter, to make it capable of balanced measurements.

I'm going to see if I can use the existing single-ended filter differentially by following it with a +40 dB instrumention amp and common mode rejection stage.
My moving magnet phono preamp is ideally suited for experimentation since its an INA followed by a CM stage.

I may do a layout for ULDO and Nacho and out them on the same board with a "Generator Monitor" and "Measurement Input" relay to switch Nacho's input.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by mediatechnology »

I've been noodling with the Ultra Low Distortion Oscillator over the last month to fine tune it.

Most of the changes modify power supply distribution and bypass.
I switched to TL431 shunt regulators for the +/- 12V supplies.
I also found that the best method of bypassing the NE5532 was differential where a large value capacitor, C25, connects from Vcc to Vee.
Differential bypass was chosen because it offers the lowest THD with balanced loads at high levels where the output current flows from output to output and therefore rail to rail.
Fortunately differential bypass doesn't compromise single-ended THD.
Series resistors added to the 5532 supply pins isolate the 5532's current waveform from the rest of the oscillator.

Image
ULDO Ultra Low Distortion Oscillator Schematic

Large: https://proaudiodesignforum.com/images/ ... ropped.png

I'm ready to layout an oscillator PCB that includes "Nacho" but before I do I want to experiment with the notch filter.
A second "Nacho" notch filter section permits stereo single-ended or single channel balanced measurements.
User avatar
JR.
Posts: 3697
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by JR. »

Nice... I notice that several parts in there were not available back decades ago.

I'm guessing that there is quadrature (90') between the two integrators. I scribbled some after beer o'clock over the years trying to make a ripple-free rectifier, using the trig identity sin^2+cos^2=1.... I discussed this some with Brad (RIP) and he said it had been done (using OTAs for the multipliers). :lol: Ancients keep stealing my ideas. :lol:

The added effort to square the two stems is probably not worth the complexity with the distortion levels you are getting. I was interested in the ripple-free trick to maybe speed up settling time without the typical distortion increase penalty.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by mediatechnology »

Thank you JR.

Its actually a Wien topology: The outputs of the two stages are out-of-polarity and provide a balanced output.
IC3B is an inverter but the compensation caps make it look like a state variable topology.
Anti-polarity outputs open the door to simple FW rectification but I found I didn't need it.

I tried more complex rectification (e.g. full-wave) and found that it wasn't worth the effort since the oscillator is fixed-frequency.
I did add a third filter pole, C7, to wipe out the small amount of ripple that was poking out of the noise floor and creating THD-2.
The settling time for amplitude changes is pretty quick.

The dominate distortion mechanism now is supply bypass and 5532 output linearity.
I think I've done about the best I can in terms of bypass and ground routing on a Protoboard and what I learned will translate into the layout.

As far as 5532 output linearity goes I'm able to drive +20 dBu into 604Ω single-ended with distortion still in the noise floor.
Balanced outputs are not quite as good - I'm aiming for -140 dBc and am within spitting range at elevated levels.

I did try buffered outputs but decided that the very small improvement wasn't worth the complexity.
I tired a diamond buffer and it particularly sucked when loaded.
It's not often I drive 600Ω loads.

I don't have a dual op amp in my inventory that exceeds what I can get out of a 5532.
The OPA1612 and LME49720 are both slightly inferior which is somewhat surprising.
High current op amps, e.g.the NJM2114 and 4560 also don't cut it.

One thing I've overlooked throughout my career is the mimimum short-circuit current of a 5532.
SIgnetics and several manufacturers never specified a minimum limit - only 38 mA typical.
One would assume that the minimum Isc (shown as Ios on some datasheets) would be high enough to guarantee 600Ω drive capability to +20 but it simply isn't so.
TI and ON both show a 10 mA minimum which is +14.8 dBu/dBm into a 600Ω load.
Fairchild, Signetics and NJR just show 38 mA...

One thing I like about this topology is that the level adjustment is DC-controlled over a pretty large range.
I can use a multi-turn pot and set output level with pretty high precision.
I only need an attenuator for phono and mic-level tests.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by terkio »

I am puzzled about a notch filter you plan to add on the oscillator board.
How will these be used ?

With the ULDO I understand you can go at measuring the THD of some audio device.
With the ULDO at the input of the DUT, the Focusrite at the output, you get the FFT then calculate the THD+N.
This should give an extremely accurate measurement at 1KHz.

Then I do not see what is the point to get the THD using the notch filter.

How do you match the notch frequency and the oscillator frequency, to perfectly reject this frequency.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by mediatechnology »

With the ULDO at the input of the DUT, the Focusrite at the output, you get the FFT then calculate the THD+N.
This should give an extremely accurate measurement at 1KHz.
Only to a point.
The Focusrite's A/D has a residual THD of around 0.0002% (https://proaudiodesignforum.com/forum/p ... =40#p15753)
and that's if you chose a level in its' sweet spot.
That's a residual level of about -114 dBc.

ULDO's THD is typically an order of magnitude times 2 better at 0.00001% (0.1 ppm -140 dBc) or less if its buried in the noise floor and can't be seen.
So there's no way I can get "an extremely accurate measure of THD" for DUTs that have THD in the range of 0.00001% to 0.0002%.
Then I do not see what is the point to get the THD using the notch filter.
Gain without A/D overload from the 1 kHz fundamental.

The notch filter removes the 1 kHz fundamental and then adds 40 dB of gain to expand the analyzer A/D range about two orders of magnitude.
By taking an FFT of the notched and amplified output, reading the levels of the individual distortion products and correcting for the filter loss I can then calculate the THD to levels impossible using the Focusrite A/D alone.
How do you match the notch frequency and the oscillator frequency, to perfectly reject this frequency.
Trim pots.
The 1 kHz doesn't have to be rejected perfectly - only enough to attenuate it significantly to prevent A/D overload when 40 dB of gain is applied.
I'm not reading the 1 kHz level - only 2 kHz and greater up to the 10th harmonic.

You may have missed some example test measurements I made on the DCAO-II headphone/line amp using ULDO and Nacho. https://proaudiodesignforum.com/forum/p ... =60#p15814
Prior to ULDO and Nacho I really couldn't quantify the DUT's distortion because it was better than any converter I had to measure it.

At 500 mW into 30Ω the DCAO measures about 0.00023% using Nacho.
When I measure it with the Focusrite A/D it's internal distortion level introduces enough error to double the THD measurement to 0.0005%.
At lower power levels, where the DCAO is in full class-A I wouldn't be able to see the distortion of the DUT at all without Nacho magnifying it.

This is a very good reference and the source of the notch filter I use: http://www.janascard.cz/PDF/An%20ultra% ... 0%20dB.pdf
User avatar
JR.
Posts: 3697
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by JR. »

mediatechnology wrote: Sun Jan 10, 2021 7:55 pm
With the ULDO at the input of the DUT, the Focusrite at the output, you get the FFT then calculate the THD+N.
This should give an extremely accurate measurement at 1KHz.
Only to a point.
The Focusrite's A/D has a residual THD of around 0.0002% (https://proaudiodesignforum.com/forum/p ... =40#p15753)
and that's if you chose a level in its' sweet spot.
That's a residual level of about -114 dBc.

ULDO's THD is typically an order of magnitude times 2 better at 0.00001% (0.1 ppm -140 dBc) or less if its buried in the noise floor and can't be seen.
So there's no way I can get "an extremely accurate measure of THD" for DUTs that have THD in the range of 0.00001% to 0.0002%.
Then I do not see what is the point to get the THD using the notch filter.
Gain without A/D overload from the 1 kHz fundamental.
while not apples and oranges, back in the day I took the product output from my heath-kit distortion analyzer (effectively a notch filter with gain) and fed that into my old 2nd hand (singer) spectrum analyzer. I could see the distortion from the Heathkit audio path from full signal so ran levels -10dB from nominal 0dB... Not bad measurement resolution for 40+ years ago. Cost me around $1k back then, today you can do it for a lot less.

JR
The notch filter removes the 1 kHz fundamental and then adds 40 dB of gain to expand the analyzer A/D range about two orders of magnitude.
By taking an FFT of the notched and amplified output, reading the levels of the individual distortion products and correcting for the filter loss I can then calculate the THD to levels impossible using the Focusrite A/D alone.
How do you match the notch frequency and the oscillator frequency, to perfectly reject this frequency.
Trim pots.
The 1 kHz doesn't have to be rejected perfectly - only enough to attenuate it significantly to prevent A/D overload when 40 dB of gain is applied.
I'm not reading the 1 kHz level - only 2 kHz and greater up to the 10th harmonic.

You may have missed some example test measurements I made on the DCAO-II headphone/line amp using ULDO and Nacho. https://proaudiodesignforum.com/forum/p ... =60#p15814
Prior to ULDO and Nacho I really couldn't quantify the DUT's distortion because it was better than any converter I had to measure it.

At 500 mW into 30Ω the DCAO measures about 0.00023% using Nacho.
When I measure it with the Focusrite A/D it's internal distortion level introduces enough error to double the THD measurement to 0.0005%.
At lower power levels, where the DCAO is in full class-A I wouldn't be able to see the distortion of the DUT at all without Nacho magnifying it.

This is a very good reference and the source of the notch filter I use: http://www.janascard.cz/PDF/An%20ultra% ... 0%20dB.pdf
Cancel the "cancel culture", do not support mob hatred.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by terkio »

mediatechnology wrote: Sun Jan 10, 2021 7:55 pm Gain without A/D overload from the 1 kHz fundamental.
Thanks, this is the key point, I was missing.
I had overlooked the importance of the x100 gain at the output of the double T filter.
The notch filter removes the 1 kHz fundamental and then adds 40 dB of gain to expand the analyzer A/D range about two orders of magnitude.
By taking an FFT of the notched and amplified output, reading the levels of the individual distortion products and correcting for the filter loss I can then calculate the THD to levels impossible using the Focusrite A/D alone.
This makes it perfectly clear to me.
One wants to see the harmonics without the fundamental that would dazzle the Focusrite.
What is actually expected from the Nacho filter is a brick wall filter.
An alternative to the double T filter, could be a high order Tchebychiev or elliptic filter. They are known to be good at cutting frequency bands, but may be no better in this application.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by mediatechnology »

The filter almost has to be passive in order be near-distortionless and it also needs to be made with ultra-low distortion passive components.

When I make a second Nacho filter, for either the right channel or for a quasi-balanced filter, I'll have a filter to check the filter.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Super Low Distortion Ultra Pure Audio Oscillators Revisited

Post by terkio »

Twin T calculator: http://sim.okawa-denshi.jp/en/TwinTCRkeisan.htm
It gives the equation of the transfert function.
It calculates everything, one wants to know about filters, when you plug in the C and R values.

A point I do not understand.
It gives two equations for f0, the "Center rejection frequency".
When it calculates from given component values, the two calculated f0s have the same value in the case of ideal C, C, 2C, R, R, R/2. So far so good.
However, if you skew the given values to see what happens with component tolerance or trimming, it gives two different f0 values.

Component tolerances and trimming is a tough issue ( at least to me ).
There are two trimmers, and they are adjusted to obtain the lowest output level.
I understand the trimming corrects the component deviations, to some extend.
I am curious to understand why it does work.
Last edited by terkio on Tue Jan 12, 2021 9:31 am, edited 1 time in total.
Post Reply