[continued from previous message]   
      
   >>> Highland Tech Glen Canyon Design Center   
   >>> Lunatic Fringe Electronics   
   >>   
   >> If I wanted to build a very low distortion 1k Hz oscillator then I'd build   
   the circuit below.   
   >> I'd be happy to see 90dB in reality if that could be measured.   
   >> If better than 90dB can be measured then that's fine but the expense of   
   doing so will likely   
   >> start to increase exponentially.   
   >>   
   >> Bill will be along soon to tell us who designed what and how he did this or   
   that but we   
   >> all know that anyway.   
   >>   
   >> I don't claim any credit for this circuit except for the Q1 Q2 circuit   
   which I arrived at   
   >> by doing online research into how it might be implemented. AS3944 might be   
   a good   
   >> choice for that circuit.   
   >> (This won't stop Bill, for the nth time, pointing out how further   
   refinements were added   
   >> but who cares.)   
   >>   
   >> I wanted to see whether it could be done without thermistors, opto devices,   
   lamps, FETs   
   >> or similar devices. And also without large capacitors. After all if you   
   wanted to implement   
   >> it all on a chip then you don't want those kinds of components in the   
   circuit if you can   
   >> avoid them. And none of the FET based circuits I ever saw could match the   
   LTSpice   
   >> performance of this circuit.   
   >>   
   > 16 to 20-bit DACS are available. Given a sine lookup table in ram, one   
   > could fine tune it for sub-PPM distortion. Or use a 16-bit DAC and sum   
   > in another 16-bit DAC that nulls the distortion, even the distortion   
   > of output amps or whatever.   
      
   You left out dithering. Cheap twenty bit sigma-delta A/D converters have   
   been around since 1993 - my 1996 millidegree thermostat paper relied on   
   one. The cheap ones weren't good for 1kHz, but the hi-fi business   
   offered faster devices.   
      
   Twenty bits is about 120dB. You don't need to block all that much of the   
   fundamental to be able us a 20-bit ADC to directly measure the harmonics   
   of interest.   
      
   > The problem is how to measure the distortion to close the loop.   
      
   Not necessarily.   
      
   --   
   Bill Sloman, Sydney   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)