floating ground on 1st smoothing capacitor

[yunger]

has anyone ever tried this? i'm not understanding the logistics and hoping someone can explain.
if i wanted the dropping resistance to be 1k effectively, would the positive and negative resistors be 500 and 500 or 1k and 1k? wattage rating the same on both?
any safety issues at the rectifier?

[eschertron]

[ATTACH=CONFIG]32339[/ATTACH]
has anyone ever tried this? i'm not understanding the logistics and hoping someone can explain.
if i wanted the dropping resistance to be 1k effectively, would the positive and negative resistors be 500 and 500 or 1k and 1k? wattage rating the same on both?
any safety issues at the rectifier?

That graphic is from The Valve Wizard

Be sure to read the linked PDF file, if you haven't already

[yunger]

That graphic is from The Valve Wizard

Be sure to read the linked PDF file, if you haven't already

yes, that's where i copied the graphic from. my questions remain

[Malcolm Irving]

Interesting question! Not sure if this explanation is right - but here goes. If there is a ripple voltage (relative to ground) at the bottom connection of the first cap (nearest the rectifiers) then the normal (unbalanced filter) will distribute that ripple voltage all the way along the ground bus. In the balanced filter version, the ripple voltage will almost all drop across the resistor in the ground path, leaving the rest of the bus clean.
How do we get a ripple voltage there in the first place? I think that would be a result of stray capacitance or mutual inductance in the wires from the rectifiers or inter-winding capacitance in the PT itself.
1k would split into 500 + 500 and each would have the same current as before, so wattage rating could be reduced to half if you want.

[yunger]

one more question...
if the two resistors are not balanced, would the B+ voltage be affected?

i'm sure they're out there, but i can't find any examples of this configuration in action.

thanks

[yunger]

Interesting question! Not sure if this explanation is right - but here goes. If there is a ripple voltage (relative to ground) at the bottom connection of the first cap (nearest the rectifiers) then the normal (unbalanced filter) will distribute that ripple voltage all the way along the ground bus. In the balanced filter version, the ripple voltage will almost all drop across the resistor in the ground path, leaving the rest of the bus clean.
How do we get a ripple voltage there in the first place? I think that would be a result of stray capacitance or mutual inductance in the wires from the rectifiers or inter-winding capacitance in the PT itself.
1k would split into 500 + 500 and each would have the same current as before, so wattage rating could be reduced to half if you want.

thanks for the explanation. quite helpful

[Malcolm Irving]

I think the B+ would stay the same. It's just that the rectifier and PT circuit is now elevated above ground.

[R.G.]

Here are some things I think about this.
1. Note where the ground symbol is. It's on the negative terminal of the second filter cap. So all voltages are measured from the minus terminal of the second filter cap.
2. The rectifiers dump current pulses into the first filter cap. That is, they make the positive side positive, and the negative side negative, then let the current flow. The charged first filter cap then discharges through the resistors into the second filter cap. Since the first cap has mostly converted the charging pulses into a reasonably smooth DC level, the current flowing into the second filter cap is more DC than charging pulses, and so it's smoother by the RC filtering of the series resistor and the second cap. A techie would say that the ripple current in the second cap is reduced. So is the voltage ripple.
3. Given that you define "ground" as the minus side of the second filter cap, there is no net effect at all on the conditions in the second filter cap depending on the relative sizes of the resistors in series from the first filter cap to the second filter cap. None at all. You can have all the resistance on the top, bottom, evenly split, or any ratio. There is a result from the math describing circuits that says that as long as you don't look at or use the internal voltage distributions, the order of items in a series string makes no difference at all.
4. Notice that this reduces to a standard power circuit when the resistors go to zero and the filter caps are just in parallel. By inserting those resistors, you get better filtering, but lose some DC voltage. You lose the DC current in the amp flowing out of the second filter cap (since the net DC current into/out of the second cap must be zero once it's charged) times the sum of the two additional resistors. This means that you probably want the resistors to stay small to avoid lowering your B+ much.
5. The total power in the two resistors is I²*(R1+R2). The power is distributed by the ratio of the resistances, and the total power changes not at all as long as the sum of the two is the same. The resistors dissipate power in proportion to their resistances.
6. Given that the circuit ground point is on the second cap negative, the ground noise is not affected at all, at least to a first order. The second through N orders depend on what the wiring and side effects of the charging pulses and loss of voltage from the added resistors do. This doesn't reduce ground noise, other than (1) knocking the high peak nature of the charging pulses down some before they get to ground and (2) absolutely forcing you to make the negative side wire of the rectifiers go the the negative side of the first filter cap that is connected to circuit ground.

There are some subtle reasons why one might want to set up a power supply this way, but they are that - subtle - and not always applicable.

[yunger]

thanks for the in depth.

in my case, i have a small amp i made using 12AU7 parallel triodes as a single ended power tube. i've learned the hard way about ripple noise since i only accounted for 2 filter caps....1 for the power amp and 1 for the preamp.
i can squeeze another filter stage in there and can also stand to drop the voltage a bit so this seems like a good application for me. at a minimum, using the balanced filter will allow me to fit physically smaller dropping resistors in a tight space.

[R.G.]

Yep single ended will show off any ripple

And yes, you can spread out the power dissipation into more resistors.

Here's a radical thought: since you can stand to lose some voltage, have you thought about regulating the B+? This could take the form of a power MOSFET and some smallish other components. The worst part of this is that you'd have to worry about getting enough heat out of the MOSFET, but that might not be all that bad.

You set up the MOSFET with its drain to the first filter cap, then tie a resistor and a string of zeners to ground. The gate of the MOSFET is connected to the resistor/zener connection, and it operates as a source follower, providing a DC voltage out the source equal to the zener string minus about 3-5V. There are some other nits to worry about, but you can get rid of about 30-40db of ripple (that is, 1/20 to 1/100 of the original ripple) in one step. It takes some careful building, but these things have worked.

[Tony Bones]

As R.G. says, the effectiveness of ripple filtering is not changed by moving some of the resistance the negative leg. However, the common mode rejection is improved by splitting the resistance equally. Common mode noise is coupled through the PT capacitively from primary to secondary (unlike the 50Hz/60Hz power which is coupled inductively.) Usually, it's noise on the power line generated from EMI, other appliances, dimmers, whatever and is higher in frequency. Sometimes this sort of thing really can reduce unwanted noise. But again, it has nothing to do with ordinary ripple reduction.

EDIT: I'm such a dumbass. Putting a filter like this before the rectifier works as I said. After the rectifier...?

[R.G.]

Yeah, it would possibly help with common mode. I tend to try to suppress incoming noise before the primary, but after works too.

The other thing is it might have been intended to suppress rectifier noise. That would better be done by snubbers on the diodes, IMHO.

[Malcolm Irving]

I think if the components and wires are 'perfect' there is no effect either way (except the rectifier circuit would be elevated above ground). But, if we allow for the (very small) resistance of the ground wires, then the ripple current in them will translate into ripple voltage. This is not a problem at the rectifier end, but causes buzz if it gets through towards the more sensitive pre-amp end.

[R.G.]

It's far worse than that. Even fairly thin wires have resistances measured in the milliohms per foot. The resistors have tolerances, probably in the 5% range, so a 500 ohm resistor that's 5% off will be either 25 ohms high or low. That's a gross imbalance that's far from the resistances, or resistance differences in the wire. So real balance is hard to do, needs precision resistors.

The real, screaming advantage is cleverly laying out the wires so that the capacitor charging pulses do not flow through wires that the circuit uses for references. In this case, the fact that the rectifier charging pulses feed into the first filter cap, but only a DC current with minor ripple on it goes out to the second filter cap means that even ground-wire caused ripple voltage is much reduced. That's what I was alluding to when I mentioned this circuit forcing you to wire the place where rectifier pulses went.

[yunger]

well, i wired in another filter cap and lifted it with 560 + 560 resistors. when i turned the amp on, it was so quiet i thought for sure i had miswired something and the amp wasn't working.
not sure if using top/bottom resistors had anything to do with it, but i'd call this a successful fix

thanks for all the help