7f829311   
   From: apogosso@tpg.com.au   
      
   "Patrick Turner"  wrote in message   
   news:091af74a-e21c-42c1-927f-4e4d80d8d815@30g2000yql.googlegroups.com...   
   >I didn't really want to proceed to using the Gilbert Cell for an AM   
   > generator although I got it to work a bit, but not much better than   
   > the pentode thing I built over 10 years ago.   
   >   
   > I tried using 3 x MJE340 transistors with a +41V supply.   
   >   
   > Two were placed in an LTP with CT coil of about 400uF and a two gang   
   > tuning cap fitted to give 500kHz to 1,700 kHz for MW.   
   >   
   > One base was taken to a fixed supply of +20V and the other biased 20V   
   > but through a 100 ohm R.   
   > The commoned emitters were taken via a 1K0 to the collector of a   
   > common emitter gain stage with its base biased via  27k to a fixed   
   > +8V.   
   > The emitter of this bjt has a 4k7 to 0V.   
   >   
   > The value of the Re determines the idle current through all devices.   
   > 1.6mAdc was found to be ideal and any more or less made severe   
   > distortions of one kind or another to appear.   
   >   
   > The LC tank has a secondary of about 1/8 of total P turns, with   
   > grounded CT and two live ends for a balanced output if needed.   
   > It was found that when tuned, the LC impedance was so high that   
   > serious distortion was too likely, so the LC had 33k strapped across   
   > it to load the LC. This also meant the sec ended up with about 700   
   > ohms Rout, end to end, or 350 ohms for 1/2 the sec.   
   >   
   >   
   > RF was fed to the live side of the LTP from a 1k pot. Only 30mV was   
   > needed Rin to the base was quite low, less than 100 ohms, so if the   
   > feed signal is large, say a volt or two as it is from my existing sig   
   > gene then the 1k pot turned low was fine, and helped keep the signal   
   > linear, due to low nonlinear bjt input impedance.   
   >   
   > AF to modulate the RF was fed to the bottom amp base via cap coupling   
   > to an AF gene.   
   > The large Re of 4k7 means that the base Rin is quite high, and also   
   > that the current change applied to the common emitters of the LTP is   
   > also linear. The 1k R allows measurement of signals but is not really   
   > needed because Rc collector impedance is effectively megohms.   
   >   
   > The above set up needed 5.1Vrms applied to the bottom bjt base. AC   
   > current change was about 1.2mA.   
   >   
   > The AM thus generated looked slightly better than the AM produced by   
   > one pentode, and I could go right up to 100% without wave flattening.   
   >   
   > The max Vo at LTP collectors was nearly equal to Eb to supply, about   
   > 19V pk-pk which is the about the maximum possible swing. About 1/8 of   
   > that appears across each 1/2 secondary.   
   >   
   > The circuit worked better than the Gilbert, but was rather too   
   > sensitive to control pot settings compared to the easier to use   
   > pentode modulator. Distortion varied a lot depending on the tuning,   
   > more Dn when tuned high, less THD when tuned low, but it works.   
   > Once set up, RF coud be adjusted to any level below 30mV and then AF   
   > set to modulate it at any level. The RF wave has virtually no   
   > distortion because the tank Q was found to be over 200, and harmonics   
   > are low. least envelope Dn is when both RF and AF produce a wave of   
   > about 1/2 of the maximum possible, ie, between 0.5V and 1V rms max.   
   >   
   > It seems to me the AF merely changes bias current in both LTP bjts. AF   
   > cannot appear differentially between collectors because they are   
   > shunted by 400uh, a very low Z at AF.   
   >   
   > The linear change in Ic causes the differential gain of the bjts to   
   > change where there is a differential input. In this case, one base is   
   > grounded and the other is fed the RF carrier. There is a CCS common   
   > emitter sink, so equal but opposite RF waves appear at each collector.   
   > The change in modulator current applied to commoned emitters swings   
   > the bjts each side of their square law non linearity, and this sums to   
   > linearly alter the differential gain. Hence the AM has a good looking   
   > envelope, and better than a Gilbert Cell. Setting up is much more   
   > fiddly than for a single pentode or for a pair of PP pentodes because   
   > a bjt only needs a very small Vb-Ve change to give a huge current   
   > change. I found that gain between the one base input and Vc-c was over   
   > 200, even at 1.7MHz, and bjts are renowned for being slow at RF, but   
   > the Ft for MJE340 is well above where I wanted to use them I may have   
   > tried BF469 which are video transistors with Ft at 60MHz which would   
   > have been better. But I only had a lot of BF472 with are PNP, and   
   > which require me to build the circuit "upside down" which I didn't   
   > fancy doing.   
   >   
   > The LC was easy to make. I googled up an inductance calculator and   
   > dialed in the dia of the 42mm od PVC former and wire size of 0.45mm Cu   
   > dia and wound a coil with slightly more turns than needed, 162t. I put   
   > 5 taps on each end of the coil to allow inductance trimming if needed   
   > The 20t sec was placed over 0.2mm insulation and centered over the   
   > primary coil CT to give little capacitance effects, but good magnetic   
   > coupling.   
   > The CT of the primary is taken to +41V and the 2 gang cap has its   
   > frame to 0V and its fixed plates to each end of the primary via 0.01uF   
   > to prevent 41Vdc appearing across the tuning plates. If ever the   
   > plates short, then the 41V supply is shunted and there's a problem.   
   >   
   > I have an HP 606A oscillator which has a similar modulator using a   
   > pair of 6CL6 pentodes with 2 gangs of a 4 gang tuning cap and CT tank   
   > coil. It has a 6B4 to drive the commoned cathodes. There is a balanced   
   > RF input to the 6CL6. The 606 also has a fancy PP tuned oscillator   
   > using the other 2 gangs. And another tube which fiddles with oscilator   
   > biasing to keep the output level constant. But that 606 oscillator is   
   > such a huge thing. Must be 2 cubic feet. I suppose they made all that   
   > sort of thing big way back then to make it look impressive so they   
   > could sell if for squillions of bucks to military and Govt   
   > institutions. Anyway, I think I will haul the old girl out and hooker   
   > up nearer to my bench and try to get it working OK if it doesn't after   
   > being asleep for the last 20 years. Its probably better than anything   
   > I could build. I got it given to me well after I built the simpler   
   > pentode modulator. Its also got meters on the front and nice knobs.   
   > There are no pissy little bjts inside.   
   >   
   > Patrick Turner.   
      
   You are on the right track Partick. As I promised, the simple differential   
   pair (you like to call it "long tail" for some reason) is more adequate for   
   AM.   
      
   Now some improvements.   
   1. Add common base cascodes on top (3...5V above) of each of your   
   transistors -- you will get rid of LC tank tuning affecting the setup. Mr.   
   Miller will simply go away. It is better to use small generic BC847 or any   
   crap in the diff pair (they have far better beta), and make your favourite   
   MJE340 become the cascodes.   
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)