PRR said:
> it is hard to twist tracks on a PCB
On PCB you can "NOT" get low hum. The AC in the heater tracks is throwing hum at grid and plate nodes, and you (generally*) have negligible cancellation.
You know multi-program broadcast consoles. Say you have the Hummingbird Hour in one mike channel. For whatever reason you are told to route the football show line output right next to these low-level nodes in a poorly balanced cable. You are going to have "GOAL!!" echoes behind the little hummingbirds.
*) Duke's twisted via PCB trick is novel to me. It isn't clear that any reasonable number of vias will really cancel the field when heaters are 0.2" (5mm) from grid and plate pins. A quad layout may work, if vias are utterly free.
There are sundry PCB tricks... I've even seen coils made from ribbon cable jumpers inserted into a PCB forming a loop, with the pcb traces connecting adjacent wires on the bottom of the PCB to complete the coil.
Chasing crosstalk and hum out of mixers is part of the job, especially if it's a powered mixer with a couple hundred watt power amp, and a spring reverb for good measure... Try to keep the hum out of the spring reverb magnetic pick-up a few inches from the power transformer, but I'm veering off topic.
One PCB layout technique that an engineer who used to work for me in my mixer design group, and sorry I forget what he called it, but he was trying to reduce the loop area for a signal pair (send and 0V return). In theory to reduce the voltage generated by a magnetic field passing through that loop area you want to make the wires as close together as possible to minimize the loop area exposed to the magnetic field. His technique that worked even for single sided PCBs was to run one trace for the signal send, and then bracket that one trace with two return traces. As he explained it to me, the effective centerline for the two outer traces is right in the middle of them, exactly where the first trace is. So in theory, there is zero effective loop area between the send trace and virtual return trace. ;D
I never bench tested this myself, but I know he did, and I presume it works (at least on the noise rejection end). Tweaking the last noise out of such designs often takes multiple iterative passes, he was my best senior mixer design engineer.
If you think about it, this virtual wire center line is kind of how star-quad mic cable works. The two pairs of twisted wires in parallel, will have an effective center in the middle between each pair. The tight symmetrical twist of all four wires means their effective center-lines should overlay each other perfectly, again forming a zero loop area condition. (I have bench tested star quad and it really works).
JR
PS: @Duke I spell it bus... if you look up "buss" you will find a verb describing something you do with a (nice) girl.. enjoy.
IMHO, hand-wired twisted-pair is an effective old-school technique for many tube inputs. In today's market DC supplies (up-rated to cover cold-start) are affordable. Both DC-drop and PCB layout favor series heater circuits whenever possible. If you have an odd number of heaters (relative to standard supply voltages), a series resistor is dead heat but also limits start-surge and gives a place for filtering. 20-50 Ohms with a low-ESR 5uFd cap will cut a lot of 20+KHz hash before it passes through your sensitive nodes.
And of course much depends on the client's metering. When I had low-power headphones jacks and VU meters I could have significant hum and not know it; on older systems maybe not even in final monitoring (all systems hummed). And of course some power-hum was masked by typically larger room-blower rumble. Now that everybody has a DAW with "meters" which go down to -96dBfs, and some work in blowerless studios (bedrooms), a wee trace of hum looks bad, even if unheard.
