Improving the capacitance multiplier

[Mr Evil]

In my continuing quest to do things slightly differently to everyone else, I've come up with some ideas to improve on the capacitance multiplier. I'm developing it to use with the horn subwoofer I'm planning, which being very efficient will want the lowest possible ripple to keep hum down. I've rejected the idea of a normal voltage regulator on the grounds that they are boring.

The improvements were inspired by R.G's thread on amp output protection where he mentioned using P-channel MOSFETs on the positive rail as switches. This made me realize that the same idea could be applied to a capacitance multiplier to lower the dropout voltage (NPN BJTs need a couple of volts; N-channel MOSFETs need even more), further improving what is already one of the main benefits of capacitance multipliers.

The extra drive circuitry necessary to accomplish this also has the benefit of increasing the effective multiplier due to high input impedance. This allows the use of a smaller 'base' capacitor (what was the base capactor anyway - it's not connected to a base now); small enough to make a film cap possible here. Base capacitors can of course still be large for ridiculously low ripple, but it will then take quite a while for the output to ramp up to full potential.

Using low on-resistance MOSFETs, such as IRF5305/IRLZ34N, dropout voltage can be really very low. As low as 0.1V is possible, but more is needed if very high currents are required. This means vanishingly small power disipation, which is nice.

With appropriate component choice this circuit should be able to provide tens of amps with only mV of ripple and very relaxed heatsink requirements. Regulation is not good, but that's because it's not a regulator. It's a bit more complicated than the standard design, but still has a reasonable component count.

[AndrewT]

Hi,
have you tested this circuit yet?
How about an extra CR filter on the supplies to x1 & x4 gates to reduce ripple even further?

[jcx]

why not just add voltage references and call your circuit a discrete low drop out series voltage regulator?

in a subwoofer amp you really can't hope for much dynamic headroom from the power supply when the major freq components at the output are much less than the rectified line freq - so a cap multiplier really isn't suited to this app at all

[Sch3mat1c]

Originally posted by jcx
[B]why not just add voltage references and call your circuit a discrete low drop out series voltage regulator?

Because...that would be......... BORING!!!!!!1


It always mystifies me how people always want to do roundabout things and totally trash performance for the sake of boredom.

Tim

[Mr Evil]

Originally posted by AndrewT
Hi,
have you tested this circuit yet?
How about an extra CR filter on the supplies to x1 & x4 gates to reduce ripple even further?

It's only simulated at the moment since I don't have any spare MOSFETs. Come to think of it, I don't think I've ever used small-signal MOSFETs for anything before.

Filtering the multiplier's supply doesn't help much since the output is determined by the voltage across X1/X4 gates and ground. Thus ripple rejection is more effected by putting current sources in the tails instead of a resistor, but it still has only a very small effect.

There are improvements to be had by bypassing R9/R13 with large capacitors, and splitting the input filter in two to make it 2nd order, as seen in the ESP project.

Originally posted by jcx
why not just add voltage references and call your circuit a discrete low drop out series voltage regulator?

in a subwoofer amp you really can't hope for much dynamic headroom from the power supply when the major freq components at the output are much less than the rectified line freq - so a cap multiplier really isn't suited to this app at all

Well you're right that there is no advantage for dynamic headroom, but there are multiple other benefits (and disadvantages) for a capacitance multiplier vs. a regulator. For instance:

• Controlled turn-on: it takes a few seconds for the output voltage to ramp up, which reduces current inrush and may help reduce thumps from the amp.
• Reduced power disippation: Allows use of relatively low-power components and small heatsinks, which reduces cost.
• More benign ripple spectrum: Regulators tend to reduce the magnitude of all harmonics evenly, resulting in the output ripple being the same (sawtooth) shape as the input, whereas capacitance multipliers attenuate high frequencies more, resulting in a smooth output ripple.

Also, there is no particular advantage to maintaining a constant supply voltage to most power amp designs, so that's one less benefit a regulator would have.

Originally posted by Sch3mat1c



Because...that would be......... BORING!!!!!!1


It always mystifies me how people always want to do roundabout things and totally trash performance for the sake of boredom.

Tim

Yay!

Hey, wait a minute! It's hardly totally trashing performance!


P.S. I feel I should point out that the transistors in the top current mirror are upside down.

[Mr Evil]

Since it's going to be a while before I can order those MOSFETs, I made a BJT version just for testing, one side of which is shown in the attached schematic.

Resistor values decreased a little to allow for lower input impedance, and capacitors increased to balance. Removed the current mirror because the extra gain makes no difference.

Not much intersting to say about its performance. As the sims predicted, ripple remains very low (just a few mV) when pulling several amps through it, as long as the dropout voltage is high enough. Required dropout is much higher for the BJT version.

[454Casull]

Any prototypes made with this?

[Mr Evil]

Originally Posted by 454Casull

Any prototypes made with this?

I made a thread when I built this: Finished capacitance multiplier. I suppose I should have put a link to it in here when I did that. I've been using it for 6 years now, and it has served me well.