From: Winfield Hill
Subject: Re: three-terminal regulator question
Date: 10 Jan 2003 04:32:04 -0800
Organization: Rowland Institute
X-Newsreader: Direct Read News 4.20
Dale Chisholm wrote...
> Winfield Hill wrote ...
>> A better HV-regulator circuit using common parts is more like
>> this, based on an old posting of mine (which I haven't tried),
>> 200V FET _____ LM317 1.2 to 125V
>> IN o----+----, ,-+-----+--| |--+---+----+-----o OUT
>> | | | | s | | | | | | 0 to 200mA
>> | _|_V_|_ | |_adj_| 240 | |
>> 2.2M ----, | | | | | 0.47uF
>> 1/2 W | '-||--+-----+ | ===
>> | | 68pF | | | 250V
>> '--------+---|<|-----------|---' |
>> / 10V zener | 2.7 ohms
>> 25k |
>> pot gnd
>> - Win
> "There's nothing new under the sun save that which is forgotten."
> I believe something awfully close to that very circuit was the subject
> of several articles in "Audio Amateur" magazine (later "Audio
> Electronics" & now "Audio Xpress") in the early or mid 1980's. They
> were regulating the plate supplies in vacuum tube power amps; probably
> Dynaco or Harmon-Kardon units. I think they had some pretty extensive
> (for a "hobby" publication) design notes and circuit analyses. Even
> down to measured values of dynamic output impedance versus frequency.
> They may have used a BJT horizontal deflection transistor, or 2n3439,
> rather than the MOSFET. I think they also concluded that the zener
> voltage wanted to be close to the LM317's max in/out differential
> (35V?) but I don't recall why.
Hah! Dale, it's too bad you don't have that 20-year-old article
handy for comparison, but it appears the design you remember has
many differences from mine above. As they say, the devil's in
the details, and a detail like using a MOSFET instead of a BJT
in a circuit like this can have a huge effect. [Otherwise, surely
we'd both have to yield precedence to some high-voltage supply
designed in the 40s, using a pair of series vacuum tubes. :>) ]
BTW, my earlier version of the above design had 1.2 to 500V output,
which if implemented with tubes would likely require two in series.
Let's think about the operating range of these circuits. Consider
the FET's gate pullup resistor. In a BJT design, this resistor
must supply enough current to drive the transistor's base at full
load current. High-voltage transistors are notorious for low beta,
e.g., the 2n3439 you mention has typical beta of only 40 at 200mA,
even less minimum. If you choose the base pullup resistor to supply
5mA base drive at say 400V out with 450V in, that's 6.8k ohms given
a say 35V difference. But now think of the 0V output case, that
6.8k resistor sources nearly 75mA and dissipates 37 watts! Clearly
such a design is a non-starter.
OK, you implement a darlington pair with another 2n3439 transistor.
This one operates at 5mA for full load, and you see a typical beta of
60. But you aren't playing the beta^squared game without a safety
margin, so you settle on a 350uA minimum base-drive capability for
400V out, which means a 100k pullup resistor. Now the pullup's
sourcing current at zero output is a more reasonable-looking 4.5mA,
dissipating 2 watts, so you choose a mean-looking 5-watt resistor.
But, oops! for open loads or low loads at 0V, that 4.5mA current has
to be sunk by the adjustment-pot circuit, and in fact it's painfully
interferring with that circuit functionality all along. Big trouble!
For example, say you add an extra load resistor to handle the light
load case near zero output, say 2.7k for 4.5mA at 12 volts. First,
this choice means your regulator's minimum output voltage is now 12V
rather than 1.2V, and second it means that at 400V out this resistor
draws 150mA and dissipates 60 watts. Furthermore, your unregulated
supply has to provide 350mA to get 200mA into the load. That idea
is also a non-starter, and you have to either entirely give up on
circuit operation at low output voltages, or require a minimum load
(which simply transfers the messy problem elsewhere: most electronic
load circuits shutdown at low voltages), or add a constant-current
sink circuit with high-voltage capability. All painful choices.
In my circuit, by contrast, the FET's 2.2M pullup resistor sources
only 80uA near zero volts; that's easily handled without any effect.
In conclusion, my circuit's 1.2-500V capability, with 0-200mA load,
is not easily matched, or even approached with a common BJT design.
Especially if matching its 5V dropout voltage is added to the spec!