From: info@turneraudio.com.au   
      
   Phil says more about Quad-II.....   
      
   ** More info:   
      
   In order to find the " open loop " THD of a Quad II output stage, I   
   removed the two EF86s and connected an output tranny primary to   
   the + supply and both plate pins.   
      
   The secondary was then driven by my bench oscillator, which has   
   very low THD -  circa 0.002%.   
      
   This gave 52Vrms from grid to grid on the output valves - enough to   
   drive the stage to 10 W at 8 ohms and 1kHz.   
      
   Results:  THD =  0.76%,   DF = 1.   
      
   ***This is all to be expected because with Quad-II OPT strapped for 8r0,   
   and load = 8r0, the RLa-a = about 4k0, so each KT66 sees a class A load of 2k0.   
   Above 10W, Po is class AB1 to a maximum of about 22Watts. If a 16r0 load is   
   used at 8r0 strapping, then each KT66 sees RLa = 4k0 and RLa-a = 8k0 and Po =   
   about 20Watts, and is all class A1 if Iadc = 70mA. Perhaps if you test with   
   16r0 instead of 8r0 the    
   THD would fall even lower than you say it is.    
      
   Now if we ignore winding losses, KT66, the 4k0 load Voltage across anode turns   
   and cathode turns = 200Vrms, and Iac = 50mArms = 70mApk. And if the tubes have   
   Iadc at idle = 70mAdc, then the amp is in pure class A. While this is the   
   case, the 10% CFB    
   windings make the two KT66 tetrodes behave almost identically to when they are   
   strapped in triode mode and have no CFB windings. BUT, the CFB connection with   
   fixed Eg2 allows the Va swing to reach down to virtually the same Ea min to   
   when KT66 are used    
   in pure tetrode mode. Hence more maximum Po is available beyond the first 10   
   Watts.   
      
   Also tried were Electro-Harmonix and GE 6L6GCs with closely similar   
   results, the EHs producing 0.8% and the GEs about 0.7%.   
      
   ***Typical expected results.   
      
   OPT has ZR 4,000 : 8 = 500:1 so if you measured DF = 1, the Rout = 8r0, so at   
   primary the Ra-a must be 500 x 8 = 4,000 ohms and this is equal to RLa-a.   
   Indeed DF without GNFB is 1. It means effective Ra of each KT66 / 6L6GC is   
   vastly reduced from 33k    
   approx to 2k0. Not quite as low as triode connected KT66, but    
   is lower than plain old UL. Mind you, when OPT winding losses are included in   
   considerations the Rout comprises Rw, and then you'd find the effective Ra is   
   closer to the 1k6 of triodes.    
      
   With EF86s restored to their sockets and the OP tranny removed,  THD   
   dropped to 0.036% ( 10W, 8ohms) and DF improved to a respectable 10.   
      
   ***So reduction of THD you measured with input stage and its GNFB loop is from   
   0.7% to 0.036% which indicates D% red factor = 1/19.4 = - 25.6dB.   
      
      
   My conclusion is that the two EF86s reduce the THD of the output stage in   
   line with the level of NFB being used ( ie 26dB ) while adding virtually   
   *NONE* of their own.   
      
   ***OK, I agree with your 26dB. I'd have to check my notebooks to see if I ever   
   measured 0.036% at 10Watts, but I cannot now recall ever seeing THD that low.   
      
   The THD will probably be, or SHOULD mainly be 3H because 2H and other even H   
   should be cancelled by balance action. If the phase of 3H produced by EF86 is   
   opposite that produced by KT66, then 3H cancelling is happening. Most class A   
   output stages make 3H    
   which shows up as slight flattening of wave peaks. But pentodes can produce   
   more peaky waves. Pentodes do odd things, and the 2H and 3H produced by a   
   single pentode change in amplitude with different RLa, and 2H can be similar   
   to a triode for low RLa and    
   as RLa becomes higher, 2H declines to zero then rises after a certain RLa   
   value. But the phase of 2H above the null is opposite to below, and I suspect   
   when a balanced pair are used the outcome gives cancelled 2H but 3H with peaks   
   in phase with H1 signal    
   phase.   
      
   Have you worked out how much applied GNFB occurs in Quad-II ? I don't recall   
   ever measuring 26dB.    
      
   If Va to Vk = 100Vrms for each KT66, 6CA7, whatever, then each Va = 90Vrms+   
   and at cathode Vk = 10Vrms-, and if Vg1 = 26Vrms, then Vg1 to Vk = 16Vrms, and   
   one might say there is CFB gain reduction from 6.25 to 3.85, so about 3dB NFB,   
   but its more because    
   of the UL screen FB between g2 and k. Its a bit tricky to quantify FB because   
   there are 2 loops involved. Its simpler if each KT66 g2 is bypassed to k with   
   its own cap, then action becomes pure beam tetrode with 10% CFB, and the Gain   
   red is more. But the    
   fixed Eg2 probably gives some reduction of odd H that is more than with g2   
   bypassed to k. Of course, g2 on one KT66 can be bypassed to k on opposite   
   KT66, so tubes are working with 20% UL and with CFB.   
      
   So, say CFB including the 10% UL from fixed Eg2 gives a total of 6dB CFB, then   
   total FB used is 26 + 6 = 32dB, which I do not believe exists until indicated   
   otherwise by some scholar here. Total FB is more like 20dB I recall; I am not   
   sure, been awhile    
   since I did the numbers.    
      
   I just dug up another schematic of mine - a 2010 reformed Quad-II with 6BX6 /   
   EF80 used as a pure diff pair, and KT88, and dynamic bias stabilization which   
   makes a 35 Watts max, and very good first 10 Watts. So Quad-II can be made to   
   perform as well as    
   Quad-II-Forties.    
   Patrick Turner.        
      
   Nice.   
      
      
   ....  Phil   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)