PS Dropping String Layout

[David Root]

Normally we go sequential in a single line, power tube plates on thru to V1.

Occasionally you see two feeds from a single node, eg feeding the reverb driver from the screen node, and so on.

Anyone see anything to be gained by feeding the PI and all preamp nodes directly off the screen node?

Would likely need additional cap uF on the later preamp tubes as they don't get the smoothing benefit of the earlier ones, or you could just make the screen cap a bit bigger than you normally would to compensate. That would not hurt the screens either.

I'm thinking that maybe fanning out the dropping string might have beneficial effects on the caps' charge/discharge cycle, ie tighten that up a bit because they are not going thru each other to get to the supply node?

Or is this just wishful thinking?

[Alexo]

The only time I do that is when I have multiple nodes and don't want progressively lower voltages.

IME, raising the value of a cap does not have nearly the benefit of adding another node - thinking back to my champ builds, putting a Pi filter in front of the A node beats out doubling the UF of that cap any day.

..re the charging and discharging cycle, I may be misunderstanding... although I suppose that if you are fanning everything out from a single node, you have to use larger dropping resistors, which raises the impedance of your downstream nodes, meaning you can get away with smaller caps, or rather, big caps act like even bigger caps.

..fwiw, I used to run around in circles on the grounding when I did this sort of thing. The key is to treat parallel PSU nodes as single PSU nodes when it comes to the grounding, if you catch my meaning.

I suppose one benefit would be a generally more hi fi power supply. If one node is sagging, it won't draw the voltage down on the downstream nodes. But most sag happens in the power amp filtering stages anyway.

[Colossal]

I agree with Alexo's comments. In a parallel arrangement, smaller caps can be used with large dropping resistors and filtering can still reach a very low cutoff (say 1Hz as a target which is well below anything we'll hear). If you go parallel, you can "isolate" each node with a diode so if one node is demanding more current, it can't rob the other strings.

[David Root]

"I suppose one benefit would be a generally more hi fi power supply. If one node is sagging, it won't draw the voltage down on the downstream nodes."

That's what I was getting at. Guess I could have put it better!

The diode trick is a good addition too.

[Alexo]

In that case, you might be better off just building a whole new power supply rail for the pi and preamps, independent of the output section.

...could be cool. You'd retain a little more punch to hit the output section with at high volumes. Might just make things sound louder... that could be fun.

[Super_Reverb]

Anyone see anything to be gained by feeding the PI and all preamp nodes directly off the screen node?

One feature you would lose is the progressively increasing filtering (ripple reduction) of the typical series RC power supply.

Another way to look at filtering is in frequency domain, whereby each RC combination creates another pole in a low pass (<<120Hz) filter. A simple RC filter creates a rolloff at a slope of -20dB/decade, so stacking poles give you a multiplicative effect (-20dB/dec + -20dB/dec, etc) as opposed to making the R or C larger, which simply reduces the frequency at which the filter starts to roll off, but doesn't increase the slope of the frequency response.

One benefit of the approach you suggest is maintaining higher preamp voltages. You could use a series approach of 4.7Ks and 33uFs and create several 1Hz poles as Colossal suggested.

The series diode is a good idea to isolate current surges in power amp from PI / preamp

my $0.02

rob

[David Root]

Alexo, that is a good idea. But expensive! I have a big chassis I could try that on though.....!

Super Reverb, that is a good counter argument, always always compromises!

[Firestorm]

I've seen two or three old designs where they split the supply into two separate rails, presumably to prevent current interaction between specific stages. But like I said, these were old designs where the intent was minimum distortion, minimum sag, maximum accuracy of reproduction. More modern designs like Trainwrecks and Dumbles wouldn't sound like themselves without that interaction.

The concept is interesting, but I think you'd have to be very deliberate about how the PS was designed: what stages should interact and in what way; are they in phase or out of phase; what relative signal levels are they seeing. Interesting, but potentially very complex.