LDO and I dont mean regulators

[mikep]

where could I find a REALLY low distortion oscillator? (analog). I believe the three letter acronym used to be "LDO", but that means something else now. for my purposes it could be fixed frequency, or coarsely selectable. Id rather not build it myself, but if it were simple enough.... any ideas?

mike

 

[NewYorkDave]

Jim Williams from Linear Technology published a design for a super-low distortion sine oscillator. I know it's out on the web somewhere. From what i remember of it, it wouldn't be too much trouble to build on a piece of perfboard.

 

[mikep]

I just ordered a copy of williams' "Art, Science and Personalities" book. aparently it is in there. thanks

 

[JohnRoberts]

There's multiple ways to skin that cat. The hard part, in my experience is the AGC. Assuming a low distortion gain cell, and the gain cell need only pass enough signal to make a circuit oscillate, but the next gotcha is the rectifier. One approach that I think is clever as hell is using the trig identity Sin^2+Cos^2=1 , Or in simpler terms using quadrature outputs from a state variable oscillator gives you the Sin and Cos terms. I was going to square them using log relationship of transistor b-e junctions but Brad clued me to a recent design using some OTAs to do the math.

Depending upon your needs there may be a few relatively simple techniques to consider, much easier than tweaking out a SOTA AGC loop. First, begin with an only decent oscillator and remove the harmonic distortion components with filtering.

If it's too hard to lower the water (distortion) consider raising the bridge (fundamental). Running the sine wave through a parametric EQ boost stage, narrow enough to not also boost the distortion components, will improve the signal purity by the amount of the boost. A couple of 15-20dB boost stages in series will make a square wave start looking good...

Of course at some point you will be reading the distortion of your EQ stage electronics so use decent opamps and perhaps a passive filter pole or two on the very output might be useful if going for the new record, but it will also be pretty difficult to measure a really silly low distortion circuit. You might want to measure it before the boost stage and calculate the estimated output distortion level.

Have fun...

JR

 

[NewYorkDave]

Mike, you won't regret it. It's a great book.

 

[Larrchild]

After you read it, spend $50 on Ebay and get a Tek Tm 500 module that's small and calibrated.
.0002%thd

Ebay Search

 

[SonsOfThunder]

I'd recommend Leader LAG-120B. You should be able to pick one up on evilbay for less than $75 if you are patient.

HTH!
Charlie

 

[audiox]

How about good old Wien bridge oscillator with filament lamp stabilation? Only few components needed and easily goes down to 0.001% and even 0.0001% if high performance opamp and carefully selected lamp is used.

 

[clintrubber]

[quote author="JohnRoberts"] but the next gotcha is the rectifier. One approach that I think is clever as hell is using the trig identity Sin^2+Cos^2=1 , Or in simpler terms using quadrature outputs from a state variable oscillator gives you the Sin and Cos terms. I was going to square them using log relationship of transistor b-e junctions but Brad clued me to a recent design using some OTAs to do the math.[/quote]
Interesting! On-line somewhere ?

Thanks,

Peter

 

[clintrubber]

[quote author="mikep"]where could I find a REALLY low distortion oscillator? (analog). I believe the three letter acronym used to be "LDO", but that means something else now. for my purposes it could be fixed frequency, or coarsely selectable. Id rather not build it myself, but if it were simple enough.... any ideas?

mike[/quote]
If it needs to be simple and your computer-soundcard is decent, how about generating some sinewaves in a wav-editor ? Add some additional filtering as mentioned here before and you're done.

The parametric is nice indeed. Before I got a better setup, for THD-measuring I used a not so good oscillator followed by a param EQ. Did the trick.

 

[Samuel Groner]

Cordell has a good paper on the topic: thd_analyzer.pdf

But the next gotcha is the rectifier. One approach that I think is clever as hell is using the trig identity Sin^2+Cos^2=1 , Or in simpler terms using quadrature outputs from a state variable oscillator gives you the Sin and Cos terms. I was going to square them using log relationship of transistor b-e junctions but Brad clued me to a recent design using some OTAs to do the math.

Interesting! On-line somewhere?

I just recently started to think about LDO design as well and found this a good solution. I'd implement it with a pair of RMS-to-DC converters. Take the RMS of the two according outputs of the state variable oscillator and sum the outputs and ideally you'll get a perfect DC output.

Samuel

 

[JohnRoberts]

[quote author="Samuel Groner"]Cordell has a good paper on the topic: thd_analyzer.pdf

I just recently started to think about LDO design as well and found this a good solution. I'd implement it with a pair of RMS-to-DC converters. Take the RMS of the two according outputs of the state variable oscillator and sum the outputs and ideally you'll get a perfect DC output.

Samuel[/quote]

Yup, the typical RMS circuit is the SQRT of integral of X^2, so two of them should work. The design I scratched out, but never finished used two discrete transistor Vb-es in series to generate the ^2 term (fed from FWR quadrature outputs), and then those two added before performing SQRT with two more junctions. The circuit Brad shared (sorry no web link AFAIK) used 4 quadrant multipliers so didn't require rectification, simpler than mine.

Attractive thing about these approaches is you should have ripple free level detection without time delay of typical smoothing filters.

JR

PS: and yes, this too is academic when you can generate sinewaves digitally of high purity. In hindsight a decent digital sample with even modest analog enhancement of the fundamental should exceed your digital measurement floor. Trying to measure this analog (check out Cordell design) is IMO too much like work.

 

[mikep]

[quote author="clintrubber"] If it needs to be simple and your computer-soundcard is decent, how about generating some sinewaves in a wav-editor ? [/quote]

the thing is, I want to get a bleeding edge ADC (probably the PCM4222_EVM, when it becomes available) for AF spectrum analysis, and I want an oscillator that has no harmonics above about -130dB (I can dream, right?). I have a benchmark DAC-1 and an apogee PSX-100 that I use in my mastering rig, I dont think either will cut it... but with some analog filtering maybe this is the way to go. I was already thinking about filtering the output of my sound tech 1710A. it is pretty darn low distortion already. at least -100dB acording to the not so good sound card I tested it with. its own measurement maxes out at .002% (or -94dB), at optimal settings. but my second oscillator, an analog sweeper, is like 1% at best, so im lookin for something new to join the party. thanks for the ideas. anyone have a Radford LDO? that looks like a cool piece.

mike

 

[clintrubber]

I see, your requirements are not that modest. That's perfect of course, but as I understood it initially you were more after a quick + easy route.

Bye,

Peter

 

[Samuel Groner]

Finally found the Linear Technology design I had in my mind--check application note 67, pages 62 to 65.

Some measurements: www.amplifier.cd/Technische_Berichte/Amplifier_reiner_Sinus/amplifier_clean_sine.htm

Samuel