Low-distortion Audio-range Oscillator

[davada]

Originally Posted by jackinnj

Here's an equation which describes the VTL LDR behavior --r2 is very high as the errors of a few percent are offsetting. I plugged the datasheet from the VTL5C9 into Eureqa and went to the deli for a sandwich:



The lines are quite compressed at the low R end, so the errors exceed 10% as the current approaches 40mA.

Edit == Yes, there are too many parens!

Thanks for this jackinnj,

I can't operate my calculator after working all night installing a fiber node hanging on the strand on a pole. I can barely type.

Is there a formula that includes the time constants? That would be needed for modelling.

CdS can be turned on rapidly but are rather passive and slower on the turn off which make them difficult to use in a full range oscillator at higher frequency.


David.

[DDB]

I've been to Mr. Moore's pages before and I can't find a schematic for his wide range oscillator. I can layout a PCB board myself.

He used Cordell's schematic with a few minor changes as listed in his article.
Cordell's schematic is on his site here. http://www.cordellaudio.com/papers/thd_analyzer.pdf

[dirkwright]

Originally Posted by DDB

He used Cordell's schematic with a few minor changes as listed in his article.
Cordell's schematic is on his site here. http://www.cordellaudio.com/papers/thd_analyzer.pdf

Hmmm, unless something has changed since the article was written, the best performance of the oscillator (which I guess is the residual on page 60 of Part 3) is 0.003% or thereabouts, depending on frequency, range and output level.

The LT Wein bridge circuit claims the same number.

So, are people seeing much better numbers out of the Cordell oscillator than the article presents?

[davada]

Originally Posted by dirkwright

Hmmm, unless something has changed since the article was written, the best performance of the oscillator (which I guess is the residual on page 60 of Part 3) is 0.003% or thereabouts, depending on frequency, range and output level.

The LT Wein bridge circuit claims the same number.

So, are people seeing much better numbers out of the Cordell oscillator than the article presents?

Dick has the distortion down a couple of magnitudes lower now. If you look around his site you will find the latest improvements. The noise and residual in earlier mods was about 0.0003%.

I'll see if I can find likes to this.

http://www.moorepage.net/IG-18-2.html

If you look at the bottom of the page the last spectrum display is 0.00063% THD+N and 0.00058% THD @ 10Vrms out to 10KHz.
This is very good. Of course the disto will go down at lower level but with the cost of lower SNR.


David.

[jackinnj]

Originally Posted by davada

Is there a formula that includes the time constants? That would be needed for modelling.

Beyond my pay grade...but one of my sons could do it.

[davada]

This is a very simple track and hold circuit I developed.
There are to variations shown here. The first is a single phase reset and the second
is a quadrature reset. The quadrature reset version having a longer hold time.
A1 generates the signals for the simulation and is replaced by an oscillator in a real world circuit.

The circuit offers some improvement over conventional track and hold
systems. There is an absents of glitch and ringing associated with fast switching circuits.
All of the signals are supplied from the input signal for the rectification and reset.
No pulse shaping circuits are required.

This TH could be used to replace the conventional peak detector circuit found in oscillator AGC's
The output power is far greater than that of a conventional peak detector and will decrease settling time of an oscillator.

The circuit illustrated uses a transistor as a very low leakage diode. The output buffer op amp should be a Jfet input type
because of their low input bias and leakage currents. The input buffer op amp is not critical provided it cover the desired bandwidth.
The arrangement with out feed back around the rectifier offers better performance at higher frequencies and doesn't suffer as much overshoot
in the absence of resistive loading. R1 or R3 set the reset current. Higher values of resistance reduce the reset pull down.
Like any conventional diode rectifier there is some peaking in the output from the peak of the sine.
The circuit tracks and holds on the positive of the sine and resets on the negative swing.
For a negative version simply replace the transistors with PNP type.

This could be used as a drop in replacement for the peak detector in Bob Cordell's oscillator circuit and others.


Cheers,

David.

[dirkwright]

Originally Posted by davada

Dick has the distortion down a couple of magnitudes lower now. If you look around his site you will find the latest improvements. The noise and residual in earlier mods was about 0.0003%.

I'll see if I can find likes to this.

IG-18 #2, the BIG-18

If you look at the bottom of the page the last spectrum display is 0.00063% THD+N and 0.00058% THD @ 10Vrms out to 10KHz.
This is very good. Of course the disto will go down at lower level but with the cost of lower SNR.


David.

I'm sorry but I miss-wrote earlier. All these darn zeros! OK, the LT Wein bridge oscillator is claimed to have 0.0003% THD. I know everyone seems to think it's not very good but if they really measured the thing then who's to say they are wrong?

Also, LT has another circuit with an RMS -> DC converter in the AGC:
http://cds.linear.com/docs/LT%20Jour...TC1968-Pei.pdf
What I find interesting is that he measured it at 0.0025% THD @ 100kHz. He doesn't mention anything about a lower frequency.

Maybe if I combine the features of this oscillator with the other bridge oscillator, I might have something good?

[dirkwright]

Originally Posted by Bob Cordell

Hi Dirkwright,

Thanks for bringing this link to my attention; I had been unaware if it. It looks like Glen has done a very nice job of updating my THD analyzer from 1981, using modern parts and using relays to replace those damned rotary switches I had in mine. I have often thought of updating mine by using relays for frequency selection as well. It does take a good number of them, epecially if you cover 11 spot frequencies per decade, but it is worth it, especially in terms of assembly labor. I don't think that Glen provided a link to my article. It can be found at CordellAudio.com - Home under the instrumentation tab.

Cheers,
Bob

Hi Bob, your website is a gold mine of good information. Thank you for providing it. I really enjoy your power amplifier book too.

If and when I do build something, I will take your suggestions into consideration. Thanks for your comments!

[dirkwright]

Originally Posted by davada

This is a very simple track and hold circuit I developed.
There are to variations shown here. The first is a single phase reset and the second
is a quadrature reset. The quadrature reset version having a longer hold time.
A1 generates the signals for the simulation and is replaced by an oscillator in a real world circuit.

The circuit offers some improvement over conventional track and hold
systems. There is an absents of glitch and ringing associated with fast switching circuits.
All of the signals are supplied from the input signal for the rectification and reset.
No pulse shaping circuits are required.

This TH could be used to replace the conventional peak detector circuit found in oscillator AGC's
The output power is far greater than that of a conventional peak detector and will decrease settling time of an oscillator.

The circuit illustrated uses a transistor as a very low leakage diode. The output buffer op amp should be a Jfet input type
because of their low input bias and leakage currents. The input buffer op amp is not critical provided it cover the desired bandwidth.
The arrangement with out feed back around the rectifier offers better performance at higher frequencies and doesn't suffer as much overshoot
in the absence of resistive loading. R1 or R3 set the reset current. Higher values of resistance reduce the reset pull down.
Like any conventional diode rectifier there is some peaking in the output from the peak of the sine.
The circuit tracks and holds on the positive of the sine and resets on the negative swing.
For a negative version simply replace the transistors with PNP type.

This could be used as a drop in replacement for the peak detector in Bob Cordell's oscillator circuit and others.


Cheers,

David.

Very interesting, thank you.

[coluke]

Originally Posted by dirkwright

I know everyone seems to think it's not very good but if they really measured the thing then who's to say they are wrong?

0.0003% isn't a bad figure at all, but an SVF ring can do better - even a bridged-t can be as simple as a wien bridge and still perform way better. Have a look a the HP 239A schematics - my 239A clone sits at about 0.00005% @ 20 Vpp out, 1kHz.

L.