[Amps] 3.5 kV 2A REGULATED Power Supply: Schematic ?

[R.Measures]

>R. Measures wrote:
>
>>**  Ian -- For AB2, yes, but for AB1, control grid V-stability matters
>>not since there is no grid current.
>
>Yes it does matter, because you cannot guarantee that grid current will 
>always be exactly zero.

**   I adjust the grid potential so that there is no grid current under 
max drive.  There can be grid current only if a more powerful driver is 
used.
>
>Even if you use speech processing or ALC, the transient response of 
>filters and ALC systems means that occasional spikes of excess drive 
>cannot be totally eliminated. This situation is real life, and the grid 
>bias supply must be able to handle it.

**  Which is why I adjust the grid potential for zero grid-I with a pulse 
which produces maximal output from the driver.  
>
>Also, some tetrodes show significant reverse grid current at lower drive 
>levels. 

**  Now there's a new one.

>If the bias supply can't handle this situation too, there will 
>be an unwanted shift in operating point.
>
>A grid bias supply whose output voltage changes with even a small trace 
>of grid current will allow - correction, will *cause* - serious IMD.
>
**  Which is why I adjust for zero grid-I with max pep drive.  
>
>
>>>To achieve such good voltage regulation, you need a transformer with
>>>very low winding resistances. Voltage doubling is not a good idea for
>>>high-current supplies, because it *always* has worse regulation than a
>>>full-wave bridge unless the winding resistances are extremely - no, make
>>>that extraordinarily - low.
>>
>>**  Transformer secondary-winding resistance is inherently low with a FWD
>>becaise only half as many secondary turns are required for the same
>>output potential -- which means fewer layers of paper insulation are
>>needed for the secondary.  Less paper means that more space is available
>>for copper.  The result is a transformer that provides the same potential
>>as a FWB configuration transformer but is more efficient because it has
>>less copper loss.  Also, the FWD configuration has the benefit of ripple
>>cancellation since, as one half of the filter is charging, the other half
>>is discharging in the opposite direction.
>
>That isn't really how it works, for several reasons.
>
>1. The situation you describe is only true for the relatively short time 
>while capacitors are being charged. All the rest of the time, the caps 
>are discharging.
>
>2. In the doubler, the voltage across half the capacitor stack is going 
>down while the other one goes up. In the bridge (or biphase with a CT 
>secondary) the whole capacitor stack gets charged.
>
>3. Even in a so-called "full wave " doubler, each half of the capacitor 
>stack is only charged on alternate cycles. With a 60Hz supply, each half 
>is discharging for almost a whole cycle (16.7ms) before it receives 
>another boost. In a full wave bridge or biphase circuit the whole 
>capacitor stack is recharged every half-cycle (8.35ms).
>
**  Sure, it's not perfect cancellation, but it does cancel.  The FWD 
anode PS at:
http://www.somis.org/pb.ps.gif
uses 20, 300uF, 450V capacitors.  The net C is 15uF @ 9000V.  Many 
builders considered this inadequate during construction, however,  no 
ripple could be heard on the air and the amplifier produced 1200V-pk into 
50-ohms.  

>A good situation to compare the two configurations is where you have a 
>transformer with two identical secondaries (or two identical 
>transformers) and you use the same two capacitors connected in series. 
>You have the option to connect both secondaries in parallel and 
>voltage-double, or both in series and use a bridge. In that situation, 
>the bridge *always* gives better regulation.
>
**  However, a same-core transformer that was designed for FWD service 
would have less secondary R than both windings in parallel in the above 
example.  

>The other side of the argument, as Rich points out, is that a 
>transformer for bridge use requires more insulation and is generally 
>more expensive.
>
**  Cu costs more than paper.

>I won't deny that voltage doublers are good value, and can be made to 
>give adequate performance - especially in high-V / low I applications. 
>But let's not kid ourselves that the voltage regulation is better than 
>bridge or biphase. In any fair comparison, it's always worse.
>
**  Your comparison was apples vs. oranges.  
>
>Later:
>>>25 - 30mA will often be OK, but it won't prevent runaway in all possible
>>>cases. Some tubes - or pairs of tubes - will generate larger negative
>>>screen currents than that.
>>
>>** Wow.   Ian must be uing some humungous tetrodes.  Perhaps these are
>>the ones that have a  chain hoist loop on top because handles simply
>>wouldn't do?
>>
>Quite the opposite - it's the small tetrodes like the 4CX250B/R, 4CX350A 
>and 4CX400A that seem to be the worst. Also, some are notably worse than 
>others, especially after they have already suffered some overheating of 
>the screen.
>
>For all of those tubes, Svetlana recommend a screen current sinking 
>capability of 15mA per tube, so 25mA might not be enough to guarantee to 
>keep a pair of tubes out of runaway. Other manufacturers are less 
>specific, but that recommendation is probably a good design value for 
>other makes also.
>
>With larger tubes, negative screen current seems to become less 
>important, relative to the normal positive current.
>
**  I was talking about tetrodes with 1000 - 2000 V screens.  For 250 - 
800 V screens, the zener-string screen shunt regulator fed through a high 
resistance from the anode supply is as foolproof as it gets.