Hiwatt power supply:Capacitive divider balast resistor values?

[SoulFetish]

I was wondering about a common design used in Hiwatt power supplies;
It's a bit strange to me, and it seems wrong. In the amp on my bench there is a series configuration of (essentially) a supply dropping resistor of 100Ω to a 50µF capacitor in series with 220µF. This creates a voltage divider where VOUT= VIN (C1/(C1 + C2)), however, there are also balast resistors in parallel with each capacitor of 220k.This is weird to me. plus the 50µF is part of a can cap with another 50µF and a shared negative terminal. Here is a schematic showing what mean....


What do you make of this?

[glebert]

I think that thinking about the caps as a voltage divider is a bit of a red herring. That is only for the AC, but the whole point of this supply it to shunt any AC to ground. If anything just think of series capacitance as Ctot=(C1*C2)/(C1+C2). The 220k resistors (I would call them "balancing" rather than "ballast") are there to keep the DC voltages across the 50uF and 220uF roughly equal. As far as where the caps reside (discrete or in a can) I think that is just a matter of economy, The node in between the 50uF and 220uF is still a pretty effective AC ground point for the other 50uF cap.

[Helmholtz]

You can't make a DC voltage divider using caps, as they don't pass a DC current (apart from an initial charging surge).
DC voltage dividing is solely accomplished by the parallel resistors.

Edit: Missed post above.

[g1]

For this particular circuit, the only thing they would need to worry about voltage division for would be if one of the series caps was much lower voltage rating than the other. Then they would need to play with the values of the balancing resistors, like in some of the newer Fender amps where they have a totem pole arrangement with different value caps.
This for example is the 65 Deluxe re-issue.

[SoulFetish]

Okay, that answered my question. I was wondering if the different capacitance would've created an offset in DC shared voltages due to the different charge capacity. Which, I suppose it would (were it not for the 220k resistors).
So, the 220µF capacitor is there to build up the voltage rating of the 50µF caps, and being of large enough capacitance so as not to decrease the total value of the stack too much. (probably around 40µF or so at HT supplies 2 & 3?)
Thanks fellas.

[SoulFetish]

For this particular circuit, the only thing they would need to worry about voltage division for would be if one of the series caps was much lower voltage rating than the other. Then they would need to play with the values of the balancing resistors, like in some of the newer Fender amps where they have a totem pole arrangement with different value caps.
This for example is the 65 Deluxe re-issue.

g1, this is the example which came to mind when I was thinking about the Hiwatt. Fender actually uses Resistor values in their divider to provide the exact V-out as the capacitor divider does

[Helmholtz]

There is no capacitor divider with DC.
A voltage divider can't work without a (common) current through the series-wired impedances.
It's worth taking the time to derive the voltage divider formula just using Ohm's law.

Resistor values in the Fender schematic correspond to the cap voltage ratings.

[g1]

There is no capacitor divider with DC.
A voltage divider can't work without a (common) current through the series-wired impedances.

Maybe it's just a terminology issue here, as far as 'divider' goes. Won't the impedances for the different value caps will be different? What happens to the voltage division in the Fender example if there are no parallel resistors (at junction of the 2 caps)?

[Helmholtz]

Maybe it's just a terminology issue here, as far as 'divider' goes. Won't the impedances for the different value caps will be different?

The impedance of a cap is proportional to 1/f. As DC means f=0, cap DC impedance is infinite (ask your Ohmmeter).
Infinite DC impedance means zero DC current.

What happens to the voltage division in the Fender example if there are no parallel resistors (at junction of the 2 caps)?

Cap voltages will settle according to the (hopefully) very high but unpredictable leakage resistances.

[catalin gramada]

The "divider" you're talking about is a matter of capacitors leakage and not of capacitive value, as much as I know. The balance resistors will not guarantee nothing as much it did not ensure a certain current through in respect of the leakage current of the caps. I.ll bet the differences will be much reflected in used caps than in new ones. Still think better use high rated voltages than nominal necessary needs to provide enough room just in case of voltage derive.

[Helmholtz]

Yes, from what I stated above it follows that balancing resistors should be much (at least 10 times) smaller than worst case leakage resistances.
I think that for somewhat aged ecaps 220k is on the high side and lower values would be safer.
Marshall often used 56k balancing resistors.

[mhuss]

FWIW, I've never seen a significant voltage imbalance in a HIWATT I've serviced, before or after a recap.

The 220uF series caps were used (a) because they already were using them elsewhere in the circuit and (b) 500v electrolytic caps did not exist at that time.

I note with current line voltages, I often see B+ in excess of 500v anyway.

[trobbins]

It's worth appreciating the current through the balancing resistors, and the likely leakage current through the caps, and what that all means.

For the Hiwatt example, HT-3 could reach above 500V for a hot switch on, but would obviously be lower, so up to 500V/440k= 1mA through the balancing resistors. In contrast, well formed e-caps such as the 50uF 450V and 220uF 350V should have idle leakage currents << 100uA. Any imbalance in the idle leakage currents will push the mid-point voltage up or down a bit from 50% of HT-3 rail dc voltage.

But there are a few important variables that can upset the apple cart and apply significantly more voltage drop across high or low side e-caps, and hence could cause concern. The balancing resistors have an initial tolerance - perhaps 5-10% - so the mid-point voltage could by chance be up to 18% away from anticipated nominal right from the time of manufacture. In some amps, the balancing resistors operate at near their max voltage rating, and are at risk of degrading (I've come across such failed resistors) and typically need to have one end unsoldered to confirm they are still ok.

The e-cap leakage current rises with age and temperature, and although initially should be at most 10% of the balancing current, that can slowly lead to increased unbalance.

During power up, there is an initial leakage current surge, and that surge can be substantial and dominate any thoughts of balance should one cap or the other want more initial re-forming current. This includes the issue of having 50+50uF as the top cap, and 220uF as the bottom cap - their power up charging will cause an unbalanced voltage divider for starters. Over decades, years or months of no use, the e-caps loose 'form', and any initial power up event can be quite stressful on unbalance. This stress can also occur if the amp has been operating from a low mains voltage, and then powered from a high mains supply - the e-caps are not used to the higher rail voltage, and any 're-forming' to the higher rail voltage can be quite substantial.

Btw, the 100 ohm is there to minimise the surge current into the 50,50-220uF caps when standby is turned on. WIth 500V available, that surge can have an initial peak of 5A (ie. a 2.5kW surge in that resistor) !