From: info@turneraudio.com.au   
      
   On Tuesday, 17 September 2013 14:03:13 UTC+10, Phil Allison wrote:   
   > **Hi, my old valve Wien bridge oscillator has an interesting quirk. During   
   the warm up period, the frequency drops by a small amount. About 6Hz at 2,000   
   or 34 Hz at 10,000 for a 10 degrees C rise. So the tempco is negative, about   
   300ppm per degree C.    
   If I play hot air onto the MF resistors mounted on the range switch, the   
   frequency rises by a similar amount. Has the ( 1940s) tuning gang got a   
   significant negative tempco ? Seems hard to believe, given the very small   
   coefficient of expansion of    
   aluminium and brass. .... Phil   
      
   Probably you may be the first person to question the stability of an ancient   
   wien bridge oscilator. 6Hz change at a nominal 2kHz is a change of 0.3% and   
   not something that is going to get noticed easily. So what happens after the   
   unit has been on for an    
   hour? any huge changes? I suspect your MF r may have changed value ever so   
   slightly but then your unit is a tubed unit, and temp inside the box will rise   
   after turn on thus affecting the C and causing minute movement of the moving   
   plates. But many ppl    
   rely on variable tuning caps especially for oscilators in HF radio where a   
   very slight C change of an LC tank at 7.1 MHz might move you off the station   
   and maybe certainly require re-tuning if the signal is SSB and you have a BFO   
   going. Well, folks just    
   keep twidling knobs to stay tuned. But in an FM set I have, the oscillator is   
   1/2 a 12AT7 at F around 100MHz, and the set WILL drift after turn on as the   
   internals rise about 20C. Hence the need to have another 1/2 12AT7 with its   
   anode used as a variable    
   reactance C which is operated by the Vdc recovered at the ratio detector. Its   
   a simple remedy, but very effective.   
      
   The normal old fashioned ruggedly made double tuning cap used in 1,001 AM   
   radios and in countless old oscillators has always been regarded as so much   
   more stable than just about everything else you can think of, except perhapds   
   a crystal, and even the    
   frequency of a crystal can be made to shift a bit with added bits and peices.   
   Presumably, a digital frequency producing chip which has a crystal reference F   
   to relate all other F to would be more stable than a humble wien bridge   
   network.   
      
   For best stability of F, you need to have a regulated temperature, so a fan   
   plus heater + thermostat is needed, all far too much trouble and all   
   completely un-neccessary for an average audio tech.   
      
   I have often thought that possibly an LC circuit used instead of the wien   
   bridge network may give less F drift. ( Think of the 26 tube Racal multi band   
   radio set which had impressive F stability even at 65MHz and with 3 oscilators   
   ie, multible    
   heterodyning. In many radios, to make the oscillator more stable, a very   
   slightly temperature dependant cap is put between the 36-360 pF gang and the   
   the coil, so as T rises, the added cap compensates for change of L due to T   
   rise, and any change in the    
   tuning cap is reduced. All a very ingenious solution to my father's generation   
   which purchased radios and who wanted them to be stable, after spending such a   
   huge amount on what we now think is a tiny amount of purchasable media.   
      
   But, there may be a good advantage for a wien bridge oscilator based on an LC   
   circuit. The Q of a typical parallel LC will be many times that of the low Q   
   wien bridge. This means that if you have an LC set up at 1kHz and its Z-in is   
   a typical 5k0, then    
   the feed in R could be 10k0, and PFB voltage to the amp would be about 1/3 of   
   the amp Vo. So the normal NFB thermistor and trim pot should be able to work   
   to control Vo.   
   Suppose the amp makes 5Vo, and OLG amp 2H = 2% = 0.1V, ( typical for tubes )   
   then suppose ß = 0.33 approx, then 0.033V 2H appeas at NFB port. But at the   
   PFB port the attenuation of 2H with LC might be -12dB, so 2H = 0.0083, and   
   phase shifted.   
   The 2H difference between both input ports is about 0.033 - 0.0083 = 0.024V,   
   and if the tubes have total gain of 1,000, then the "error signal" created =   
   1,000 x 0.024 = 24.66V, which of course does not actuall exist, unless you had   
   found that **with NFB connected**, the THD was 2%, but I know you would not   
   find this.   
      
   The effectiveness of the NFB towards reducing THD would have to be MUCH better   
   if the PFB network has a high Q, and harmonics are much more attenuated than   
   by the wien bridge.   
      
   Consider 2H. The PFB network causes A times the 2H to be reproduced at Vo, and   
   its ß for RC is slightly less at 2H, maybe 0.3. The NFB network causes the 2H   
   to be amplified in opposite phase and the ß = 0.33, so the 2H at output is   
   reduced because there'   
   s a large amount of oposing phase of 2H than the orignal Vo 2H.   
      
   But we could say overall ß' = NFB 0.33 - PFB 0.30 = 00.033, and if the    
   open loop gain at 2H = 1,000, then 2H reduction factor after NFB = 2H / ( 1 +   
   [ 1,000 x 0.033 ] ) = 1 / ( 1 + 33 ), ie, 2H of amplifier is reduced by factor   
   1/34and if there was 2% without NFB, with NFB 2H should be 0.058%, which one   
   may find is typical    
   for a tubed wien bridge oscillator.    
      
   With LC, the PFB ß for 2H could be 0.083, and effective ß' = 0.333 - 0.083 =   
   0.249, and the 2H reduction factor = 1/250, and 2H would be lower at 0.008%,   
   and if someone used an op-amp instead of a couple of tubes, then expect low   
   THD with an LC tank.    
      
   But pigs wiil fly before anyone could verify my figures because there is more   
   than one thing which creates THD.    
      
   And should anyone make a variable F oscilator with LC tank, they may find the   
   variation of Q with F for any LC will mean the effective ß at the Fo in the   
   PFB loop will vary considerably, thus Vo level will vary. A typical Hartley   
   tube oscillator with say    
   1/2 12AU7 attempts to control its Vo level by means of the negative bias at   
   the grid. So as Vo rises, so does negative grid bias which reduces energy made   
   by the tube to the tank. This energy is in the form of highly non linear   
   current pulses lasting for    
   a portion of each wave, so the Vo has considerable HD, despite the high Q of   
   the LC. However, this HD can be drastically reduced if a suitable sized   
   resistance is placed between the tube cathode and the tap on the LC at about   
   1/4 of the turns, and the R    
   value tweaked for low 2H, 3H etc, while making sure the tube will always start   
   to produce oscillations. Perhaps Vdc amp could improve bias control of Vo in   
   such an oscillator. In radios, the tuned anode plus tuned grid with 2 tuned   
   circuits gives cleaner    
   Vo.   
      
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)