The original schematic used three separate Hammond iron components: a Hammond 369AX as the main power transformer providing the high-voltage B+ secondary and heater windings, a Hammond 167S6 as a secondary transformer for an additional heater supply, and a Hammond 193B filter choke in the B+ supply for CLC pi-filter smoothing. This worked but was heavy, expensive, and took up a massive amount of board and chassis real estate. Most of the design work has centered around consolidating and improving the power supply while keeping noise low enough to not defeat the purpose of building a tube amp in the first place.

The first major design change was replacing both the 369AX and 167S6 with a single toroidal transformer. The idea was straightforward: one core with all the secondaries I need instead of two separate chassis-mount transformers. I needed a high-voltage center-tapped secondary for the rectifiers, a 6.3V heater winding rated for the total heater current draw of all four tubes (the 6AS7 alone pulls 2.5A of heater current), and whatever additional heater or isolation winding the 167S6 was providing. Toroidals have lower stray magnetic field means less hum coupling into the audio path, lighter weight, and a more compact footprint. But they also come with higher inrush current on startup which is a problem with tube circuits since the B+ comes up immediately while the tubes are still cold and not conducting. This necessitated adding a soft-start circuit, either an NTC thermistor or a timed relay with a series power resistor that gets bypassed after a short delay. 

The Hammond 193B filter choke is around 5–8H at the DC bias current the B+ supply sees. It works well for ripple filtering in the CLC topology but it is a big, heavy, expensive part. I explored two alternatives:
The first option was a board-mount inductor. The challenge here is finding something with enough inductance at the required DC bias current in a board-mount package. Most board-mount inductors in the multi-henry range are either rated for very low DC current or are physically almost as large as the Hammond choke, which defeats the purpose. The second and more promising option was a capacitance multiplier circuit. A capacitance multiplier uses a transistor and a relatively small RC filter to electronically simulate a much larger capacitor. The transistor's emitter follows the filtered base voltage, effectively multiplying the capacitance of the base filter cap by the transistor's beta. This can replace the choke entirely while taking up a fraction of the board space and costing almost nothing. The tradeoff is that unlike a passive LC filter, it drops voltage across the transistor (which dissipates as heat), so there is an efficiency penalty. For a tube amp where the B+ current draw is relatively modest this is very manageable.

The full schematic was built in KiCad. One of the more involved parts of the design was setting up proper high-voltage clearance rules for PCB layout. This is a circuit where 300V+ B+ rails, 120V mains connections, 6.3V heater lines, and millivolt-level audio signals all coexist on the same board, and the clearance requirements between these nets are very different.
I set up five net classes with specific clearance and trace width rules:

HV_BPLUS — 4.0mm clearance, 1.0mm trace width for all B+ supply nets, rectifier outputs, and plate supply nodes
MAINS — 8.0mm clearance, 1.0mm trace width for the primary side of the transformer and anything connected to line voltage
HEATER — 0.5mm clearance but 5.0mm trace width to handle the high heater current (especially for the 6AS7)
SIGNAL — 0.25mm clearance, 0.25mm trace width for the audio signal path
CONTROL — 0.5mm clearance, 0.25mm trace width for soft-start logic and bias control nets

Custom DRC rules were written so that KiCad flags any trace-to-trace violation between net classes automatically. For example any signal trace that gets routed within 4mm of a B+ trace gets flagged, and any copper within 8mm of a mains net gets caught. This is critical because it is very easy to accidentally route a signal trace too close to HV when you are focused on getting the topology right and not thinking about physical clearance.
The heater traces being 5mm wide is worth noting — the 6AS7 alone draws 2.5A through its heater and undersized heater traces will cause voltage drop and uneven heating which introduces hum. Getting the heater routing right and keeping it physically separated from the signal path was one of the more tedious parts of the layout.
Challenges and Current Status
The most annoying part of this project has been the interdependency of the power supply design decisions. Changing the transformer changes the available secondary voltages which changes the B+ rail voltage which affects the operating point of every tube stage which changes the current draw which feeds back into transformer and filter component selection. This is similar to what I ran into with the liquid rocket engine design where chamber pressure, injector geometry, and mass flow rate are all codependent, you pick starting values, iterate, and refine.
 

This project is still in progress. Photos and board renders will be added as the layout is completed and boards are fabricated.