Bass guitar preamplifier in the Ampeg SVT tradition

Power supplies

Regulated filament power supply

The regulated filament power supply board is designed to supply 12VDC @ 750 mA. The preamp contains 2 x 12AX7 and 2 x 12AU7 filaments, each nominally drawing150 mA.   The regulator is good for 3A, and the filament windings are rated for 2A.  The limiting factor is the heat sink dissipation.  I consider this a conservative thermal design, because at 750 mA and 40C ambient temperature, the regulator junction temperature is less than 40C below its maximum rating.  The following description presents the details.
DIY SVT preamp regulated
                      filament power supply assembled
The original Ampeg SVT had an unregulated filament supply.  The Ampeg SVP-CL provided a 12 VDC regulated supply with a full wave bridge rectifier implemented with 1N4007 1kV diodes, followed by 2 x 1500 uF filter caps, followed by an MC78M12CT fixed regulator.  The combination of 1N4007 diodes and standard design fixed regulator requires that the power transformer filament winding to supply 14 VAC under load, because two diodes per leg give a 3.4V (2 x 1.7V) forward voltage drop, and the regulator needs an additional 2.5V above the output voltage.

The power transformer I selected, the Antek AS-05T280, provides only 12.6 VAC under load from the filament windings, which is exactly right for an unregulated AC filament, but insufficient for a regulated 12V supply using the same diodes and regulator as the SVP-CL.  To obtain a regulated 12 VDC from the Antek's filament windings, it is necessary to use both a low dropout regulator (LDO) and low forward-voltage-drop diodes.  I chose MUR-410G diodes and a LM1085-12 regulator.  The forward drop from the diodes is only 2 x 0.9V = 1.8V, and 1.5V from the regulator, giving a total DC drop of 3.3 V.  Ideally, the minimum AC voltage for this arrangement to stay in regulation is (12V + 1.8V + 1.5V) / 1.41 = 10.85 VAC.  We should therefore be able to stay in regulation With the AC mains 10% low, or 12.6 VAC x 0.9 = 11.3 VAC.

It is possible to obtain a 10 dB improvement in the regulator's ripple rejection performance by using the adjustable version of the LM1085, rather than the fixed version, because the adjustment terminal can be compensated with a bypass capacitor.  Iteration 2 of the filament supply might use the LM1085-ADJ instead of the fixed LM1085-12 to take advantage of the better ripple rejection, and to allow the output voltage to be set to 12.6V instead of 12V, if desired.  I don't think a few mV ripple is significant for a DC filament supply, and it is perfectly acceptable to supply nominally 12.6V filaments with 12V (5% low).

The regulator requires a heatsink.  Given an input voltage of 16V, an output voltage of 12V, and an output current of 0.75A (4 x 150 mA filaments + 150 mA for margin), the power dissipated in the regulator = (Vout - Vin) x Iout = 3 W. The LM1085's maximum junction temperature is 125C.  Therefore, we need a heatsink that can dissipate 3W with a temperature rise of less than 125C - T ambient.  I chose to work the problem backwards, by finding the thermal resistance of a 1" high TO-220 heatsink (so it would fit in a 1U box), and calculating the thermal rise.  I chose an Aavid Thermalloy 531002B02500G (Mouser part 532-531002B25G), which has a thermal resistance of 13.4 C/W, i.e., each watt dissipated into the heatsink will result in a rise of 13.4 C.  So dissipating 3W will result in a rise of 40C.  Adding in the thermal resistance of the TO-220 package (0.7C/W), and the thermal pad that electrically insulates the regulator from the heatsink (0.2C/W) gives a thermal rise of (13.4 + 0.7 + 0.2)C/W * 3W = 43C.  If we let the ambient temperature inside the 1U box be 40C (104F), then the junction temperature will be 40C + 43C = 83C, which gives us a thermal margin of 42C before the regulator chip hits 125C. The vertical stack up of a 1" heat sink on a 1/8" board mounted on a 3/8" standoff doesn't leave any clearance between the top of the heatsink and the top cover of the 1U case.  I gave it vertical clearance by using a 1/4" standoff in place of the 3/8" under the filament power supply board.

The negative terminal of the filament output is referenced to a voltage divider on the B+ supply giving 1/4 of the B+ voltage (77V).  Raising the filament reference voltage reduces the heater to cathode potential for the direct coupled preamp stages.

The power-on indicator LED is driven through a current limiting resistor from the regulated filament supply.  The 20 mA draw from the LED serves as a minimal load for the regulator if all the tubes are removed for troubleshooting.

High voltage power supply

The high voltage power supply is designed to supply 320 VDC @ 20mA for the B+ of the preamplifier, and provide a well-filtered reference voltage of 1/4 the B+ voltage (~80 VDC) for the tube filaments.

The high voltage winding of the Antek AS-05T280 supplies 290 VAC @ up to 90 mA  to 4 x UF4007 silicon diodes forming a full wave bridge rectifier.  The RC filtering stages are 3 x 47 uF electrolytic caps separated by 1.2 K ohm resistors supplying about 320 VDC @ 20 mA to the preamp board. 

Following these filter stages is a voltage divider consisting of a 100K and 33K resistors, supplying 77V to the filament reference.  It is filtered with a 10 uF electrolytic cap.  The voltage reference supplies negligible current.  I followed a rule of thumb to keep the total resistance of the filament reference divider less than 150K.
high voltage power supply
                      board assembled
Ampeg's SVP-CL B+ supply used 3 x 100 uF filter caps.  I used 47 uF caps without any misgivings because 100 uF @450V axial lead caps were not readily available, and the additional capacitance is not really effective for a circuit that draws only 20 mA (the rule of thumb is 1 uF per 1 mA).  The Ampeg design feeds the next 7 preamp triodes from the same B+ node, with only the first gain stage B+ being decoupled by another 100 uF cap.  This is a poor use of capacitance, as a single 12AX7 triode draws less than 1 mA, leaving the other 7 triode sections with a poorly decoupled plate supply.  It is not considered good practice to feed more than 2 gain stages from the same power supply node, as the triode sections can have undesired interactions through their common power supply nodes, e.g. "motorboating".  Even though this isn't a high gain preamp, I still didn't like the idea of running 7 triode sections from the same power supply node.

My preamp board B+ supply uses a 10 uF / 6.8 K ohm RC filter for the first gain stage, a 10 uF / 1.5 K RC filter for the next 3 triode sections, and another 10 uF / 1.5K RC filter for the next 2 triode sections. Cascaded RC stages are much more effective at reducing voltage ripple than adding capacitance to fewer stages, as only a few nanovolts remains in the B+ supply at the critical first gain stage.

Because this unit will be used in Europe, a externally accessible voltage selector switch is provided.  The voltage selector switch connects the dual primary windings of the power transformer in parallel for use with 120VAC mains, and in series for 240 VAC mains.

Simulating the high voltage power supply

As with all of my prior projects, I used Duncan Amplification's Power Supply Designer II, a free Windows application that makes it easy to graphically design and simulate typical unregulated linear power supplies in a few minutes with a bit of pointing and clicking.  You can achieve the same result with simulation tools like LTSpice, but it might take a few minutes longer starting from scratch.   Here is my PSUD2 design file for the preamplifier's high voltage supply.  The output of the program looks like this:
Duncan's Power Supply
                  Designer II screen capture
The tabular section of the output shows the voltages and currents at each node of the circuit.  I've scrolled the output to the voltages at each of the filter capacitor nodes.  The simulation results agree with the preamp as-built to within a few volts.  Note that the predicted power supply ripple, Diff at V(C6), the first stage preamp, is 176 nanovolts.  That is what you get from 6 cascaded R-C filter stages.  The basic rule of thumb for calculating what rectified DC output voltage under load you need from the power transformer is approximately 125% of the DC voltage after the final filter stage.  In this case, 376 VDC after the rectifier is filtered down to 291 VDC at the first gain stage.  The loads are represented as fixed current taps. The values of the fixed current taps are the sum of the plate currents for all the tube sections attached to that power supply node.  The plate currents may be obtained easily from an LTSpice simulation of the preamplifier, using a fixed DC voltage supply.  I simulate the power transformer, rectifier, and initial filter caps separately from the rest of the preamp because it takes the several seconds for the power supply to reach steady state, much longer than the preamp circuit.

Constructing the power supply boards

Full-sized layouts are cut and taped to the blank circuit board to serve as a template for marking the positions of the screw and turret holes. Note that some component values shown on the layout templates were changed in the actual build.  Refer to the complete build layout document for the correct values
full sized layout is tapped to the board
                      for center marking
A spring loaded punch is used to mark the locations of the screw and turret holes, then the layout paper is removed from the boards
Use spring loaded punch to mark the turret
                      and screw hole locations
The boards are drilled on a drill press.  The drill size is #33 (0.113"), which is just 0.001" larger than the turret body for a very snug fit.
poiwer supply boards are drilled
The turrets are swaged into the holes.  The Keystone turret lugs are Mouser Electronics part 534-1509-4.  I like using these turrets because the large hole in the center can accomodate several component leads. I set the turrets on my drill press (not plugged in) using the Keystone staking tool 534-TL-8, also available from Mouser.  I recommend using the Keystone tool rather than a pointed DIY swaging tool, because the profile on the Keystone tool flares the end of the turret without damaging it.
turrets are staked into the board
The appropriate turrets are laced together with #22 soft tinned buss wire, using the Doug Hoffman technique explained here.
some turrets are laced together
Here are the assembled boards attached to one of my prototyping fixtures ready for subsystem testing.
power supply boards attached to prototyping plate
              for testing