From: info@turneraudio.com.au   
      
   Last night I said I had a plan in mind for dc current control of the lamp   
   resistance.............   
       
   """Then I built a high input Z rectifier to produce a max +Vdc of +3Vdc. This   
   is applied to a base of a common emitter stage which inverts the at the   
   collector output so that a Vdc can be derived to power a BC640 bjt with Re   
   180r so that as it turns on    
   it feeds collector current up to 22mA through 4 series lams each rated for 12V   
   x 50mA.    
   Suitable R + electro Cs slow down any change in lamp DC current and its the DC   
   though the lamps that causes the most change in their resistance, which varies   
   between 27r cold to up 115r with about 22mA.    
      
   I have not tried my 3 bjt Vo regulator yet, but it does seem to behave far   
   better than the j-fet, and up to a slightly higher Vo.    
      
   The ß just goes over 0.33 at 3.5Vrms input to my circuit.    
   But at turn on, ß = 0.1, so oscillations will be max and distorted until lamps   
   warm up SLOWLY.    
      
   This NFB is a much lower resistance circuit so I expect no troubles getting   
   all F between 100kHz and 1MHz    
      
   I should get far lower THD, because the VNFB produced has no visible THD on my   
   CRO, unlike the j-fet which produced several % 2H.""   
      
   This is all very well, and a fabulous plan and while there wasn't any   
   oscilator amp hooked up, the experiment showed it might work, but   
   unfortunately, time constants got me yet again when trying to implement a   
   circuit with long time constants into    
   another amp circuit ( the oscillator amp ) with very high open loop gain, even   
   at near DC.    
   The result was that the Vo slowly waved up and down like very slow amplitude   
   modulation because the DC servo circuit was acting as a LF oscilator to vary   
   the lamp resistance.    
   Back to the drawing board. I doubt if any simulation program would tell me it   
   won't work before I built it because I doubt anyobe would know how to enter in   
   so many interactive variables.    
      
   But at least it works with a j-fet, except for the highish THD. I will   
   reserach a little more, before returing to the j-fet idea.   
      
   Sorry to bore you all to tears. Its OK, but wenya tryta build sumping good,   
   with high aims in mind, then a special Law of Life applies :- COWPAT = 1 / N   
   squared, where COWPAT = Chance Of Working Perfectly Any Time and N is the   
   number of things you didn't    
   know might stop it working, or what you didn't know, or you dropped a spanner,   
   or the dog ate your data notebooks :-)   
   Patrick Turner    
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)