Simple power supply design for reduced screen voltage

[nuke]

I've had this kicking around for while as a means to provide a lower voltage power rail for screen supply in pentode/beam-power output amplifiers.

This first schematic is the standard half-wave rectifier power supply that's been around since the beginning of tubes amps. I drew it with silicon rectifiers, but it is the same functionally as a tube rectifier. The interesting thing here is the "capacitor input, pi-filter" topology. The output of the rectifier is fed directly into a capacitor. This kind of power supply approaches 1.41 * Vrms of DC output. ( 1.41 is sqrt(2) ) and with a 360-0-360 transformer or about 500v, which sags a good bit with load to 470-480v or so.

The downside is that the plates and screens are operating from essentially the same voltage. That was OK, since mostly, amp makers were looking for as many watts as possible and didn't worry much over tube life. So everything ran a bit hot. In the old days, tubes were cheap, and robust and handled this well and there was an endless desire for watts output.

standard-halfwave.png

Today, tubes are expensive, not as robust and we don't need a lot of power, in the increasingly fewer venues that allow guitar amps at all. So, it makes sense to conserve tube life, reduce power, and to maintain good tone provide a good, stiff screen supply power rail.

A good method to do this is simply add a DPST standby switch in place of the normal SPST standby switch, and two silicon rectifier diodes, which then feed directly into the choke, forming a "choke input filter power supply". This means that the output of the rectifier is fed into a choke.

split-choke-input-screen-supply.png

A choke input filter power supply produces lower voltage and better regulation, but at much lower voltage. So with a 360-0-360 VAC transformer, you get about (2*(sqrt(2))/pi * Vrms, or about 324 VDC.

By splitting the power supply, we can feed the heavy current needs of the output tube plates with the capacitor input filter power stage, as usual. A choke-input filter for this power level would be large and expensive.

However, the usual Fender 90ma 4H choke, or the 5-7H 120ma chokes found in Marshall amps, is already reasonably priced and available. Ideally, a bigger choke is more ideal, but we can get away with it here as long as the total current through the choke isn't approaching its saturation current limit. For a pair of most common output tubes, like EL84, 6V6, 6L6, EL34 should do fine.

[Roe]

I've had this kicking around for while as a means to provide a lower voltage power rail for screen supply in pentode/beam-power output amplifiers.

This first schematic is the standard half-wave rectifier power supply that's been around since the beginning of tubes amps. I drew it with silicon rectifiers, but it is the same functionally as a tube rectifier. The interesting thing here is the "capacitor input, pi-filter" topology. The output of the rectifier is fed directly into a capacitor. This kind of power supply approaches 1.41 * Vrms of DC output. ( 1.41 is sqrt(2) ) and with a 360-0-360 transformer or about 500v, which sags a good bit with load to 470-480v or so.

The downside is that the plates and screens are operating from essentially the same voltage. That was OK, since mostly, amp makers were looking for as many watts as possible and didn't worry much over tube life. So everything ran a bit hot. In the old days, tubes were cheap, and robust and handled this well and there was an endless desire for watts output.

standard-halfwave.png


Today, tubes are expensive, not as robust and we don't need a lot of power, in the increasingly fewer venues that allow guitar amps at all. So, it makes sense to conserve tube life, reduce power, and to maintain good tone provide a good, stiff screen supply power rail.

A good method to do this is simply add a DPST standby switch in place of the normal SPST standby switch, and two silicon rectifier diodes, which then feed directly into the choke, forming a "choke input filter power supply". This means that the output of the rectifier is fed into a choke.

split-choke-input-screen-supply.png

A choke input filter power supply produces lower voltage and better regulation, but at much lower voltage. So with a 360-0-360 VAC transformer, you get about (2*(sqrt(2))/pi * Vrms, or about 324 VDC.

By splitting the power supply, we can feed the heavy current needs of the output tube plates with the capacitor input filter power stage, as usual. A choke-input filter for this power level would be large and expensive.

However, the usual Fender 90ma 4H choke, or the 5-7H 120ma chokes found in Marshall amps, is already reasonably priced and available. Ideally, a bigger choke is more ideal, but we can get away with it here as long as the total current through the choke isn't approaching its saturation current limit. For a pair of most common output tubes, like EL84, 6V6, 6L6, EL34 should do fine.

I was thinking along the same lines. For the standby switch you can add power resistors across it (47k-220k). With a single pole standby you can lift the ground IFF you add a resistor across it?

To adjust the screen voltage, add a small cap after the rectifier. try 220+nf

[R.G.]

Good idea.

As you tinker, keep in mind some quirks of choke input filters - they have a current-controlled variation of output voltage. If the load on the capacitor after the choke is nearly zero, the thing acts just like a capacitor input filter and the output voltage is high. As you suck increasing current out of the cap after the inductor, the voltage drops until a critical current is reached and then remains essentially constant with increasing load until something burns up.

This means that there is a relationship between the inductance and the minimum load that the filter supplies. You either need to pick a "critical inductance" to meet the circuit's natural minimum load, or add minimum loading with a bleeder resistor to keep the inductor conducting constantly and hence keep the output voltage constant.

I am lazy. I might opt for the simplest possible MOSFET source-follower regulator - a couple of resistors, a cap, a zener or two, and a BFM. You get any output voltage you want, and it's constant(ish) up to any current you can supply from the rectifier and PT and heat sink the MOSFET for living through it. A TO-220 can dissipate up to 2W in free air with no heat sink at all. More that that, you need a sink, which can be pretty small and simple for a few watts.

[nuke]

This was born out of my current thinking of, "use what I have".

I have a lot of really good mosfet switches, but not a lot of really great linear mosfets.


I'm staring at this amp chassis and I notice:

1. I already have the Marshall-style DPST power/standby switches.

2. I already have a reasonable choke.

3. I already have 100 or more silicon rectifier diodes in the parts drawer.

So I shook the box of ingredients up and got the idea to just re-arrange things a bit.

I searched Ultra-Librian, Mouser and Digikey and RS/Allied for a 100H choke with .0001 ohms of DCR and came up empty.

So, I figured I'd employ what's sitting on my bench already. And this would work equally well with a tube rectifier, just tack the diodes on to the circuit at the HT leads.

I like the linear-pass regulator idea. Maybe extend it to the screens on a node, the PI on a node and the preamps on a node, each with an active mosfet regulator in the power supply. While we're at it, might as well regulate the filaments too.

Ok, back to shaking the massive box of parts I already have....

[R.G.]

This was born out of my current thinking of, "use what I have".
I have a lot of really good mosfet switches, but not a lot of really great linear mosfets.

I get it!

I like the linear-pass regulator idea. Maybe extend it to the screens on a node, the PI on a node and the preamps on a node, each with an active mosfet regulator in the power supply. While we're at it, might as well regulate the filaments too.

Ok, back to shaking the massive box of parts I already have....

I would only do the screens, and let the RC dropping chain fall where it may.

Mouser says they have suitable MOSFETs for under $1 each. There is still the delivery cost, I know.

[LOUDthud]

Many modeling programs freakout when you simulate choke input filters. When the diodes (or the switch) stops conducting, the choke generates a giant Voltage trying to keep current flowing. In the case of the Full-Wave CT connected diodes, the choke tries to push the junction of the two diodes negative so the other diode catches the transient Voltage. The switch in your schematic isn't so lucky. It will probably arc and eventually fail. I'm not sure what protection circuit might work. A diode to prevent the input side of the choke from going negative ?

[Roe]

Many modeling programs freakout when you simulate choke input filters. When the diodes (or the switch) stops conducting, the choke generates a giant Voltage trying to keep current flowing. In the case of the Full-Wave CT connected diodes, the choke tries to push the junction of the two diodes negative so the other diode catches the transient Voltage. The switch in your schematic isn't so lucky. It will probably arc and eventually fail. I'm not sure what protection circuit might work. A diode to prevent the input side of the choke from going negative ?

yes, a diode here may work. Also, consider a small cap to ground connected to the rectifier (e.g., .47uf 630v) and power resistor across the standby switch (total silence usually calls for 150k or higher values)

[nuke]

It is a BAS. Big-ass-switch.

The collapsing magnetic field of the choke needs to dissipate to somewhere. But in the normal case of switching to standby after you stop playing, there's not a lot of current in the choke.

If we take the formula to get the joules stored in the inductor:

J = 1/2 L I^2

I'm using a 4H choke, rated for up to 90ma. I'm probably looking at maybe 30ma. The energy in the choke would be:

J = 1/2 * (4H) * ( .030 A )^2

= 0.0018 Joules.

Even at the full 90ma rating of the choke, it's 0.0162 Joules. So, not that much energy.

But enough that it might arc at the contacts and wear them prematurely.

But what to do about it is simple:

Place a capacitor across the switch. Same as we do in breaker-point ignition systems that all the cars used to have prior to electronic ignition.

With the switch closed, the capacitor will be completely discharged. When the switch open, the capacitor will appear as a "short" as it begins to charge. We just need enough capacitance to dissipate the stored energy in the choke core, until the switch contacts separate far enough to prevent any arcing.

If you work the capacitor formula for energy storage:

J = (C * V^2)/2

re-arrange the algebra:

C = 2J / V^2

Plug in some numbers: C = .02 / (350^2) = 160nf

So just need a 0.2 µf capacitor, rated for 400vdc and preferably one with good current handling. Or anything really from 0.1 µf would work.

[R.G.]

Even simpler to convert the standby switch to muting the output, not switching B+.

There is a fair mount of reasoning that standby power switching is unnecessary in modern tube gear. What it's used for is mostly muting for a while, and having the amp come back up semi-instantly. A reasonable alternative is to remove the standby switch from the B+ wiring, and instead use it to mute the audio going to the PI or output tubes.

[nuke]

Yeah, but I'm used to the standby switch. I find them kind of handy for development builds. AKA, prototypes, aka 'homebrew' amps.

The one thing that I'm concerned about is any chance the screens are powered before the plates are...

[TUBEDUDE]

I think Merlin recommends the use of a small cap, say 470nF, shunting the inductor input to ground. It's not large enough to become a cap input supply, but it will maintain a voltage on the coil long enough to keep the field from collapsing between cycles.

[Roe]

split-choke-input-screen-supply.png

Some details that may be adjusted:
- there is lots of voltage across the standby switch, consider adding 100k-220k power resistors across it. OR switch AC voltages instead!
- the screen supply must draw current to drop voltage. consider a resistor to ground to drop and stabilize voltage (100+k) together with a zobel network of .1u/100R as well as a .22uf cap before the choke
- the preamp voltages will be too low if feed from the screen supply. Use the mains supply instead

See https://ampgarage.com/forum/viewtopic.p ... 3&start=60