Started working on my first PCB (guitar amplifier) and so far have just roughly placing a few of the components and ran some tracks that I've color-coded based on purpose, not necessarily the layer they will be routed on. I plan on externally mounting potentiometers and switches and any footprints are placeholders for molex connectors. But I had a few questions and concerns before I actually commit to a layout or start adjusting vias, trace width, spacing, etc....

GREEN = Signal, RED = Local Ground Star, ORANGE = Ground Bus, BLUE= 100-250VDC 3mA RMS (or filament pins 4,5,9 6VAC 300mA) enter image description here

  1. Is there a significant cost or performance difference between 90 degree, 2x 135 degree, or a 1mm-10mm filet for bends in traces at lower voltages that don't experience corona?

  2. As it is just for myself, cost isn't a huge concern. I'd like to keep noise to a minimum. But I'm pretty overwhelmed with trying to visualize the layers and pours and how to properly utilize them. I've gotten accustomed to point-to-point wiring and planes and pour zones don't compute in my head yet. Do I want to start routing everything on the bottom layer furthest away from where the components will mount through on the top-side, and only move up a layer if traces cross? Or should an un-utilized copper plane be layered in between anywhere traces cross to shield from cross talk? If someone could list a few possible layouts for a 2 and 4 layers board that would be a huge help in visualizing my next steps. I don't understand how starting with pouring your "power plane" or a "ground plane" like everyone says to start with online won't just end up with a bunch of loops, but I can't help but feel that some of my traces are longer than anything I've seen in person.

EDIT: Added schematics and visualizations for clarity, grounding/current visualizations from https://www.valvewizard.co.uk/Grounding.pdf EDIT 2: added another layout I made with side by side comparison if it helps with clarity, I/O should end up on the same edge if possible enter image description here enter image description here enter image description here enter image description here enter image description here

  • 1
    \$\begingroup\$ Hello and welcome. It's generally very difficult to look at just a PCB without understanding the circuit, which is why circuit diagrams are so important, they convey the purpose of each thing, not just how big it is and so on. Can I suggest you add your diagram so we can understand a bit better? \$\endgroup\$
    – jonathanjo
    Commented May 27 at 15:10
  • \$\begingroup\$ @jonathanjo Sure! I'll update the original post with my schematics and a PDF I've been using as a reference \$\endgroup\$
    – Tyler B
    Commented May 27 at 16:20

1 Answer 1


Vacuum tube circuits are actually one of the cases where you don't want a tight ground plane, at least not like for a digital circuit. Circuit impedances are high, and bandwidth low (100s of kHz is doing well) so microstrip (trace over GND plane) geometry simply looks like capacitance.

You also need fairly generous spacing between conductors, including to ground, due to the high voltages (particularly around the output, which might peak to 1kV or more in higher-powered designs).

Don't worry about star grounding. There is a proper way to do it, and audiophile circles almost universally get it wrong -- they are poor reference. The biggest problem is trace length, making huge loop areas for induction -- it's an RF nightmare, and no wonder gate stopper resistors are almost universally employed to help prevent high-frequency oscillation.

You can at least shorten the path like so:

enter image description here

The main thing is to have PSU currents flow into the capacitor, which I guess is already a minimal problem as this wouldn't be raw rectified DC, but RC or LC filtered to this section.

At (1), note that the grid circuit can return to the cathode circuit just fine. Cathode is the ground reference for the amplifier stage itself, and makes the priority ground location. Not to mention, the grid resistor is massive, a 100k resistor isn't going to create any sensible ground-loop voltage at all. For that matter, ideally a (floating) input would return (GND) here as well.

Even better though, you can just fill the whole bottom edge of the board with ground, or wherever else it's convenient -- make strips around and between stages, perhaps -- and get a ground impedance far lower than any star ground would afford.

High-voltage effects aren't relevant at these voltages (corona etc.), but just ensure there's plenty of clearance. I'd probably set a rule of 50 mil (~1.2mm) and be done with it; most traces can be spaced out much farther anyway. I'd also use fatter traces (20-40 mil / 0.5-1mm?), I guess just to ease the electric field around them, but it really doesn't matter for signal traces. You may want larger still for heater connections; it hardly matters on just a 12AX7, but typical power tubes draw up to a couple amperes where it will help.

There are also techniques to deal with input ground loop. I have no idea how this isn't universally known and understood in audiophile circles, but... well, then again, I kind of do.

enter image description here

This makes considerations far beyond your scope here (RFI, etc.), but is (...should be..) the gold standard in input wiring. The trick is to treat the input as a differential source: what your amplifier is measuring is the voltage between terminals, that's it. It should ignore (reject) voltage between circuit and input grounds (common mode; hence the measure, common mode rejection ratio).

The easiest way to do this is a "long-tailed" or differential pair. We still need a ground reference, we can't leave it completely floating or the circuit would malfunction, so we use a resistor (1M to GND) to set the idle potential. This draws minimal current from the input if GND voltages don't match, so we don't incur voltage drop along the cable ground conductor.

At high frequencies (e.g. ambient noise sources: SMPS, cellphones and other radios, sparking circuits, etc.), we need a low impedance to honor the unbalanced nature of the connection: shunt signal GND to chassis GND with a modest (some nF) capacitor. This capacitor could resonate with the cable's inductance, so a ferrite bead might be added to the left, or a CMC (common mode choke) to the right; the CMC also adds inductance so should be dampened with the subsequent capacitor, hence the 100Ω + 4.7nF. Finally, audio cables tend to have poor CMRR at RF, so some differential-mode filtering is desirable: use a ferrite bead and 1nF, say.

This level of RF filtering is almost certainly already overkill for vacuum tube use, as the rectification behavior is pretty mild. It is of critical importance to BJT amplifiers, where the base-emitter junction easily rectifies even 10s of mV of RF into a sensible DC offset, or if varying (amplitude modulation), signal error. We would prefer such a circuit in that case.

Finally, to deal with potential safety issues like large transients (ESD?), crossed wiring (say 120V somehow gets on an input) or other paranoia, a TVS diode enforces common mode voltage range. (If crossed wiring is an actual concern, the input can even be fused, to isolate it in case the TVS is activated by a low impedance source. The TVS will most likely fail shorted in the process, so both it and fuse will need replacement. A TVS and fuse on the signal line would also be needed.) Which, for a LTP with a CCS tail, the CMR can be massive, hundreds of volts -- an automatic advantage of tube circuits that takes a little more work to solve with solid state, not that such a huge range is often required. (That said, professional equipment is sometimes exposed to wild voltages: stage wiring can be hasty, and grounds highly dubious, or even crossed; audio isolation transformers are a common solution!)

Finally, circuit grounding would be made in the usual way: a simple chain from stage to stage, with the rectifier at the far end where its peaky ground-loop current is well separated from signal paths. (Think of this as a cascade of star grounds, where each star is the output of one stage, feeding the input of the next; the voltage between stages that aren't directly connected is irrelevant, so we don't need a more central star, as you often (most likely erroneously) see in projects.) Chassis ground is tied at a single point, perhaps near the input, or the middle; with the excellent CMRR at the input section, and little current flowing through this link (normally), it really only serves to reference the two grounds together. A hard connection is better than a high impedance, for safety purposes, however: suppose a wire comes loose inside, you don't want 500V zapping something (potentially electrifying the chassis, or the inputs with respect to it), it should find a low-impedance ground path, short out and blow the fuse.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.