A Low Distortion Sinewave Oscillator

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

dcollins

Active member
Joined
Mar 27, 2006
Messages
37
A very low distortion oscillator is described, with performance in the >-140dB range. 

http://www.tmworld.com/article/521682-Low_distortion_oscillator_tests_measurement_circuits.php


DC
 
I notice RAP and JW listed in the credits... Sam Groner has done his share of work with practical low distortion generators.

We quickly get into a chicken and egg scenario trying to measure circuitry this good.

JR
 
nice but it only drives 600 ohms,

HP will drive 10 ohms but not with those thd specs,

so goo for ADC's but you would need a stronger output amp to make it good for testing transformers down to 5 hz, jus sayin,
 
A Wein bridge oscillator with AGC.  This time with an opto instead of a tail-light bulb. 

75 years later, still a good idea.


DC

 
I rebuilt my old Heathkit oscillator with a Valley People VCA for AGC about 20 years ago, still works pretty good.
LT1010 output amp does 50 ohms OK too.
Not this good, but better than the generator in the 8903B
 
I always had trouble getting fast settling time "and" low distortion at the same time.. the 8903 may have made a trade off toward quick settling time for better usability on the bench.

JR
 
JohnRoberts said:
I always had trouble getting fast settling time "and" low distortion at the same time..
IM very HE, only 2 analogue oscillator types designs have low distortion and fast settling. 

One is the old Beat Frequency Oscillator designs from the valve days like the B&K 2010.

The other is Great Guru Baxandall's.  I had the dubious honour of getting a number of his designs into production .  GGB always had new and useful insights into a problem but his stuff wasn't always (?!) easy to make consistently.

http://www.aes.org/e-lib/browse.cfm?elib=3941  the original.  The 2 companies mentioned are KEF & Calrec.
http://www.aes.org/e-lib/browse.cfm?elib=14135  John Vanderkooy's take.

Are there 21st century digital solutions which do the above and still allow manual control with a (simulated?) knob?  Please post if you know of any free ones.  There is a useful Klark Teknik application but that is stepped frequency.
 
Thanks I guess... Perhaps I should have mentioned my experience was in the context of low cost test equipment I designed and sold some 30 years ago...  (Loftech TS-1).  That sine wave oscillator was voltage (actually current) controlled with a roughly 2,000:1 adjustment range..on one knob.  Using about a $2 parts cost, back in those '80s dollars.  Settling time wasn't my only problem, but I won't bore you with a veer into cheap design.

There was a good thread here looking over Sam's shoulder a while back as he dialed in a good bench oscillator, and I recall some good discussions back years ago when Brad Carso was active here. I recall some discussions with him off list about an approach I was interested in based on an old Trig identity.  Who didn't love Sin^2+Cos^2=1 ? .. In theory grabbing two quadrature outputs from a sinewave generator and squaring them should deliver a ripple free control side chain.  IIRC Brad even came up with a schematic somebody published doing that.. or close enough.

These days I'm inclined to just get a 24b dac and push bits.

JR
 
JohnRoberts said:
I always had trouble getting fast settling time "and" low distortion at the same time.. the 8903 may have made a trade off toward quick settling time for better usability on the bench.

They are pretty much mutually exclusive, aren't they?  The AP has a switch, so did my old Sound Technology for "Fast/High Accuracy."

Istr, that all-pass oscillators have faster settling time, or at least something to that extent.  I don't recall the actual mechanism.


DC
 
JohnRoberts said:
Who didn't love Sin^2+Cos^2=1 ? .. In theory grabbing two quadrature outputs from a sinewave generator and squaring them should deliver a ripple free control side chain.  IIRC Brad even came up with a schematic somebody published doing that.. or close enough.
This is sorta like the Baxandall method.  I'm not sure if John Vanderkooy does exactly that cos I kunt reed en rite.

They are pretty much mutually exclusive, aren't they?
They are with Wein & Twin T type stuff.  The BFO types simply clip and since they are heterodyned, all harmonics can be filtered out at RF.  No bounce.  Baxandall uses a state variable filter so has cos & sin available.  The SV can also be arranged to have little or no gain change for a wide range near resonance.  It's the gain changes which give rise to bounce.  Both these methods are fairly (?) easy to get to 0.1% with few high order products.  A bit more care can give you 0.01% and better.

But ultimately, Twin T & Wein bridges give better distortion.  Prof. Janasek says he got better results with Wein but isn't quite sure why.

These days I'm inclined to just get a 24b dac and push bits.
I miss my B&K 2010.  Very clean and single knob (variable cap) over 1000x range.

Having a big smooth knob is VERY useful sometimes.  Anyone know of cheap or free digital apps which simulate this with low distortion and no bounce?  I'm OK for spot frequencies or fancy sweeps.  But I want a knob.  Perfectly happy with proper 16b performance.
 
We quickly get into a chicken and egg scenario trying to measure circuitry this good.

Indeed. The measurement results presented in this article are plain naive, and possibly very misleading. It is very difficult to ensure that a passive notch filter does not contribute any distortion at the -140 dB level, which could potentially cancel the distortion of the oscillator. And the presented notch filter is not even a particularly smart one, as the rheostat-connected trimmers are a substantial invitation for distortion (the ill-behaved wiper contact resistance is directly in series with the signal).

I always had trouble getting fast settling time "and" low distortion at the same time...
They are pretty much mutually exclusive, aren't they?

Not at all. Most oscillators (including the HP 8903) from the 80ies and later used sample-and-hold based level dectectors, which theoretically give settling times of just a few cycles with zero ripple. See e.g. www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1980-08.pdf (page 10ff) and http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1985-12.pdf (page 31ff).

The AP has a switch, so did my old Sound Technology for "Fast/High Accuracy".

The AP "Fast/High Accuracy" switch does not affect distortion and basic oscillator settling time; it affects frequency accuracy. In the "High Accuracy" mode the actual oscillator frequency is measured and then adjusted. This of course requires some time, but the settling behaviour of the oscillator leveling loop has nothing to do with this.

Samuel
 
Samuel Groner said:
The AP "Fast/High Accuracy" switch does not affect distortion and basic oscillator settling time; it affects frequency accuracy. In the "High Accuracy" mode the actual oscillator frequency is measured and then adjusted. This of course requires some time, but the settling behaviour of the oscillator leveling loop has nothing to do with this.

Thanks for the clarifications, Samuel.

What actual performance would you expect to see from the posted circuit?


DC
 
What actual performance would you expect to see from the posted circuit?

As we dive below -120 dB, secondary effects such as layout (in particular mutual inductance and capacitance with the supply rail) and the quality of the passives become very important. Thus the performance is very dependent on a specific implementation and may even vary from specimen to specimen, and necessarily hard to predict from just a schematic.

I have little doubt that a decent implementation of this oscillator gives consistent performance below -130 dB. But interpreting distortion figures at the -140 dB and even -150 dB level as if they were the difference between a 1 VDC and 2 VDC measurement is just: :eek:

Samuel
 
Perhaps we should take a moment of silence to respect both RAP and JW  (RIP) who did good work in high linearity design. Not to prolong this veer but I am often suspicious of published distortion specs for uber opamps when they get this extreme and often involve gimmicky measurements. Bench test equipment is probably the last frontier for high performance design.

The good news is there isn't much point in arguing about artifacts down at those levels, while some still do. Once you get past a couple zeros to the right of the decimal point, it mostly about bragging rights.

JR

PS I nominate Sam for our resident sine wave generator expert, he's melted more solder in the pursuit of this already than most do in a lifetime.
 
Sam, can you post links to your circuits?

I hope some of you will post links to software implementations too ...  which simulate a knob.

When I say 'knob', I mean some device that allows easy & precise manual control of frequency.  So mouse or thumbwheel = knob.
 
ricardo said:
Sam, can you post links to your circuits?
IIRC Sam hosts a website with many of his designs published.
I hope some of you will post links to software implementations too ...  which simulate a knob.
I have made digital sinewaves  and changed frequency in response to button pushes in connection with my tuner designs. The steps were in terms of musical intervals, and the code probably instructs very little out of context of the larger software architecture, and specific hardware/firmware.

A software knob promises the capability to deliver an optimal scale, and even adaptive response from say a rotary encoder, that can adjust response based on recent command history.
When I say 'knob', I mean some device that allows easy & precise manual control of frequency.  So mouse or thumbwheel = knob.

You mean PC/MAC software? I write microcode for lower level devices.

I have long speculated about revisiting my old TS-1 using modern technology, but I try to stay away from products that can done with smart telephones and easily copied software.

JR.
 
Samuel Groner said:
I have not (possibly not yet) made the final schematics public. However the thread www.groupdiy.com/index.php?topic=26461.40 gives a good overview of my initial research in this field.
:eek:  I see all the heavy hitters chimed in!  I grovel at your collective feet.

Sam, are you going to publish schematics of your track & hold / sample & hold method?  Your pic of 1kHz settling on page 5 is awesome.

My own puny efforts are at Baxandall & Vanderkooy level and even that is me mostly wanking  8)
I have made digital sinewaves  and changed frequency in response to button pushes
For the times when I want manual control with a knob, stepped frequency no matter what the resolution, won't cut it.  One example is finding a very high Q resonance in a speaker.

Actually stepping back as far as I can, many of the things I want a knob for, can be done using more sophisticated methods and 21st century digits.  I go back to the VERY early days of so called Thiele parameters using a manual oscillator, freq. meter, valve voltmeter & calculator.

In the late 80's I worked out the theory for using log sweep to find the response & distortion of a speaker in the theoretically shortest possible time for production.  Computers and especially good digital Codecs were far too expensive to make it viable for what I wanted on the factory floor.

When I emerged from the bush after a decade as a beach bum, I found computing power far greater than my wildest dreams in da previous millenium.  It's now ridiculously easy to do this.  It's Prof. Angelo Farina's method but I'm happy for him to take the credit as he popularized it.  Excuse my wanking

...  but I would still like a software oscillator with a stepless knob ...  preferably free  ;D
 
ricardo said:
...  but I would still like a software oscillator with a stepless knob ...  preferably free  ;D

Well, the stepless part is, uh, not so simple, but if you were so inclined, you could write an Apple Audio Unit to do you oscillating.

Oh, wait, writing an Apple Audio unit is also somewhat ... not simple.

-a
 
Back
Top