question about power supply nodes

[sean_thornton]

The general rule of thumb seems to be one node (rc filter) for every pair of triodes but i seem to be needing an awful lot of filter caps. previously i have wired 4 triodes per node with no problems. I am currently working on a design that uses 2 x ecl86s in the power stage, 2 x 12ax7 in the preamp, and 1 x 12ax7 for reverb. My rail is 47uf, 33uf, and then FOUR 22uf caps, which seems like overkill. The 22uf caps are supplying the ecl86 triodes (ltpi), reverb, v2a&b, and v1a&b.
Does this seem excessive to anybody else? My voltages on Duncans PSU Designer look great, right where I wanted them, but it still seems a lot of caps.

[sluckey]

The Fender AB763 and Vox AC-15 both run six triodes from a single PS node.

[sean_thornton]

I know, thats why it seems like such overkill. I have decided to run the PI and reverb off one node and v1 and v2 off another.

[vibrolax]

The ampeg SVT preamp had 6 triodes running off a single filter node.  Only the first gain stage ran off its own node.  I think what you can get away with depends on how the triode stages interact.  If you have cascaded gain stages, feedback, and/or high current then you might want to be conservative.  The rule of thumb may also have evolved when filter caps were typically <10 uF.  It may be that other forms of unintentional coupling between stages, e.g. lead dress, conductive fiberboard, were "fixed" by adding power supply filtering.

It would be interesting to see if "motorboating", "ghost notes", and other defects commonly attributed to inadequate filtering or power supply decoupling could be observed in an LTSpice simulation.

I don't mind dropping in a 10 uF and 5+k resistor in the low current sections of the preamp because the filtering effect of the RC cascade benefits each stage down the line. 

[sean_thornton]

v1a&b are paralleled, with a relay operated "fat boost", then the gain pot, into v2. V2a is setup as a cathode follower driving a TMB tone stack, and then into v2b. the reverb (relay switched) goes between v2b and the MV. Strictly speaking i don't NEED v2b in the circuit but i am playing with the idea of having it on a switch, just to try it out. So v1 and v2 could be fed by one node along with the reverb, it's not a very high gain design, more a test platform for some ideas of mine. The PI needs its own node. My rail voltages are on the high side but thats as planned (nothing over 290v), but I kinda like the tone from the higher voltages in the preamp.

[EL34]

You get better separation by using separate nodes, but in the case of a Fender, you are not using both pre-amps at the same time
That's a parallel type situation
Two pre-amps in parallel



Notice how the 5F6A bassman has one less node than an old Marshall
Both circuit are the same, but Marshall decided to separate the pre-amp with one more node


The Marshall's power supply is superior, IMO


If you are doing a higher gain type amp and the stages are all in series, I would do separate nodes
If not, then just copy old amp designs

[sean_thornton]

Thanks Doug, useful info there. This is my "own" design (not much new in tube tech really), not a high gain amp at all, single channel, basically just trying a few ideas.
Thanks for the help and input everybody.

[HotBluePlates]

Pick a single channel of any design. Count the triode per filter cap.

Note where the output signal is taken. A plate output inverts the signal at the grid. A cathode output does not invert the signal at the grid.

If you have 2 series triodes, each with a plate output, the 2 outputs (and tube plate currents) are moving in opposite directions when a signal is applied to the first grid.

If you had 3 plate-output stages, the 1st and 3rd stage currents/outputs are moving in the same direction. That opens the risk of feedback through a shared power supply connection.

Sometimes you'll see 3 series stages on a single filter cap, but 2 are plate-output while the 3rd is a cathode follower. No risk of feedback there.

As Doug pointed out, additional channels don't count, because they generally weren't designed to have the same input signal applied to each at the same time.

Other things use many more filter caps than what you think is "normal." I have some tube microphone preamps made by a reputable designer. The power supply is kept in a totally separate box from the preamp-proper. The plate power and heaters from that power supply box are both regulated. However, once the power enters the preamp chassis, each tube stage has its own 22uF filter cap (even though that might not be technically necessary).

The mic preamp's circuit uses a negative feedback loop for the EQ controls. I think the designer wanted to make the supply as perfect as reasonable/possible to eliminate sneak-paths through shared power supply nodes, given the feedback loop that was already present.

[EL34]

Good advice HPB

I like lots of filter caps and stage separation

I just finished doing the layout, schematic and BOM for the Hoffman Vox AC30

Check out the crazy power supply that Vox used
Right after the choke, the power supply splits 3 ways in parallel

Usually the power Power supply goes down the power rail in series to each stage in most amps

Hit refresh after loading the PDF to see the latest version
http://el34world.com/Hoffman/files/Hoffman_AC30.pdf

[HotBluePlates]

Check out the crazy power supply that Vox used
Right after the choke, the power supply splits 3 ways in parallel

And Vox didn't do that "just because..."

The supply voltage is low feeding the EL84 plates (at ~320vdc). The voltage after the choke would be similar, due to the relatively low d.c. resistance of the choke.

But if the rest of the power supply was laid out in series like Fender/Marshall amps starting with 450vdc+, there wouldn't be enough voltage left at the early preamp stages. That's because each dropping resistor shaves off some amount of B+ voltage.

Notice that Vox had to use relatively high 22kΩ dropping resistors for most of the preamp. They had to do that because they didn't have the benefit of successive stages of power supply filtering, as in a Fender supply. So the big resistor working against the filter cap improves hum reduction. But big-resistor also equals big voltage drop.

Overall, the Vox has poorer filtering all-around compared to a Fender or Marshall amp. But that is a necessary trade-off of having a lower supply voltage to start with.

[sean_thornton]

I have decided to feed 4 triodes per node. 
One node will feed the LTPI and Reverb, and the other node will feed V1 and V2. This takes me down to 2 x 22uf caps in the rail. I use the 47uf and 33uf to feed the power section. This combination seems to work well for me. I also use SS rectification (4 x UF4007), never cared much for tube rectifiers. My voltages are all still within my limits so I will carry on. Thanks to everybody for all their input and help. I will post a schematic soon, but it really is just a basic little amp with a few ideas i wanted to try.