[R-390] Re: R-390 Voltage Regulator

[Drew Papanek]

On a soiled-state replacement for the series pass tubes in the R-390 non-A  
Roger L. Ruszkowski  wrote (not in the order I quote):

>Third good thing is a zener reference and series pass type transistor 
>regulator will work fine.

Someone used that approach and published the circuit and results in, IIRC, 
Electric Radio some years back.

>We will have to do the math on an R390 B+ line for the real current and the 
>real voltage that is
dropped.

IIRC, current drain is about 200mA or so.   The input voltage to the 
regulator has a high ripple content because of the Non-A's puny filter 
capacitor.  The ripple is not a problem; regulating action removes it.  
Ripple peaks reach about 300 volt, IIRC.  Finding Regulator power 
dissipation calls for calculus.

>A 2N3055 will do a 100 watts.

Only sorta'.  That would be when perfectly thermally coupled to a heatsink 
whose temperature was 25 deg. C.   Any reasonable heatsink would get hotter 
than that.   Two or more  pass devices operated in parallel would be 
preferable.  Devices in parallel need some emitter ballast resistance to 
share current equally.  I would venture a guess at somewhat less than 10 
ohms for each device.

The 2N3055 hasn't sufficiently high VCEO to run at voltages encountered in a 
'non-A.  For easy availability and overkill on voltage rating one could use 
a TV horizontal output transistor, though somewhat low on hfe.

>Rather than play with the 47 ohm resistors under the chassis, solder a wire 
>into the socket pins for >each 6082 cathode (a total of four wires) and 
>each plate (another four wires). <snip>
>This will draw the current through all 4 of the 47 ohm resistors.
>This also keeps the current spread over all the tube socket pins. Rather 
>than
>pass the full current through any one socket pin, keep the power spread 
>out.
>Your new plug in wiring may take the load, but consider all the original 
>wiring
>under the deck. Its so much easier to just build a mod that will work and 
>play
>well with a little prior planing than to rewire a burned chassis.

A single wire would handle 200 mA easily.   For simple plug-and-play, 
though, 4 wires would effectively place all of the 47-ohm resistors in 
parallel and thereby reduce the output impedance and improve voltage 
regulation.   Just one wire to a cathode pin could be used if its 
accompanying 47-ohm resistor were jumpered out under the chassis (so much 
for plug-and-play, but you were planning to tear down the radio for 
refurbishment anyhow.)

>The next operation is to build up a suitable reference voltage. The 5651 is
>only 85 volts. The solid state regulator wants a reference .6 volts above 
>the
>desired B+ voltage. My choice would be for a couple 5 watt zener diodes in
>series with a proper filter cap across the zener diodes and a voltage 
>dropping
>resistor between the high power supply voltage and the junction of the 
>regulating
>transistor base and zener reference.

Power zeners can be tough to get and are harder to mount than lower powered 
devices.  An NPN with zener collector to base, a couple K resistor base to 
ground , emitter grounded and collector used as the power zener cathode 
substitute would work.

The reference would be more free of ripple if a diode (1N4007 suitable) were 
inserted in series with the reference dropping resistor.  That would keep 
the reference's filter cap from discharging back through the dropping 
resistor during ripple valleys.

>Depending on heat sink preference, and the regulating transistor, 
>everything can be mounted on the >relay octal sockets and plugged in.
[in place of the 6082's]

A heat sink to fit in that space would need fan cooling.  A small fan could 
be mounted atop the plug-in and powered off the filament pins.  A small DC 
computer-type fan (with rectifier and dropping resistor) would serve.

I would prefer Roger's "put the heatsinks outside the radio with a long 
cable" approach.  With heat thusly relocated, evah' li'l bit he'ps.

Instead of bipolar pass devices, power MOSFETs could be used. The IRF820 
(available from Mouser) would suffice.

That approach is used in a high voltage regulator in a tube tester project.  
Go to members.aol.com/sbench101.  Click on RAT Tube Tester.

Same considerations about sharing power dissipation between several devices 
apply.  Use a few ohms ballast resistance in series with each MOSFET's 
source.  Connect 1k (exact value not critical) in series with each gate 
terminal to quell oscillation tendencies; locate the resistor right at the 
gate terminal with short leads.

The high gate impedance (zero DC current) makes a reference using low power 
zeners practical.
Stack low-power zeners to, say, 200 volts for a reference.  Bypass with  
.1uF poly for zener noise.   Regulator input goes to 1N4007 then to 10 uF 
250v electrolytic then to 47k (2 mA zener current; must be above the zener 
knee), 47k to zener stack.   Strap a few hundred K pot across the zener 
stack and connect wiper to resistor'ed  gate terminals.  Pot will adjust 
output voltage.  Selected fixed resistors could be used instead.

For a voltage reference more in keeping with the theme use a few neon bulbs 
in series to replace the zeners.  The parallel capacitance would have to be 
kept low to prevent relaxation oscillations.  NE-2's would be run at about 
half a milliamp.  Pretty to look at.

Another interesting approach can be found in the National Semiconductor 
Linear Data book.  That circuit uses the LM317 sort of cascoded with series 
pass transistors to stand off the high voltage.

That's mine 2 cents worth. Thanks go to Roger for bringing up this 
interesting topic.  Any more ideas to share?

Drew