Virtual ground in power amp applications

[aspringv]

OK, so I was running around getting a quote on some transformers for a F5 amp build, and I came across a transfo (625VA, 38-0, 38-0 secondaries) which was for sale for a song... so I bought it figuring it'd be good for a class AB build. The I got to thinking (always dangerous!) about my f5 build and started wondering about whether I could use the transfo I have somehow...

So the F5 wants rails of +/- 24V and with 38V secondaries I'd end up with an unloaded rail voltages of around +/- 53V.... which is almost double what I actually need... So I started wondering about voltage doubling rectification, and so arrived at the following questions.

Is there a way to (exactly) halve the voltage output after rectification?

Crazy idea I had was to use a truckload of diodes (I have, oh, about 1000 N4004's on hand). Each one would knock ~1V off the rail voltage and I can then drag the rail voltage down to +/- 24V. I can wire them in series/parallel so that they can deal with the current they'll be seeing.

Another one was to apply some kinda voltage regulation, but my understanding is that with a Class-A build, voltage regulation is a bit of a pointless exercise, and I'd be asking the regulator to dump a -lot- of power as heat whenever it's running. I haven't looked up the current draw of an F5 as yet but i already feel sorry for whatever sort of regulation scheme I come up with...

Anywho, I'd normally experiment a bit, but playing with that kinda power alarms me somewhat, so a question seemed in order!

Any suggestions?

[AndrewT]

Originally Posted by aspringv

OK, so I was running around getting a quote on some transformers for a F5 amp build, and I came across a transfo (625VA, 38-0, 38-0 secondaries) which was for sale for a song...

a superb transformer for a two channel 140W+140W into 8ohm amplifier.
It would even do for ~250W+250W into 4ohm amplifier.
It's too low VA for a single channel KSA100.

[godfrey]

Hi

Wasting half the voltage seems like a bad idea - and there would be twice as much heat to get rid of.

Maybe you could do something like shown below, with the left and right channels each having their own supply (one transformer secondary per channel).

It looks a bit funky, but Quad uses this arrangement in some of their amps so I guess it works OK.

I'm not sure what resistor values would be best - with 100R as shown, they'll each waste about 6W as heat. For a class A amp that won't make much difference though.

Cheers - Godfrey

[aspringv]

That makes sense to me - you're making a kind of virtual ground in lieu of a CT.
. As you say, I can then use one secondary per channel...

Nice! I'll give that a shot when my f5 boards arrive. Cheers!

[AndrewT]

hi,
those 100r will allow a low current unbalance for an opamp or similar low current circuit.

Each F5 could require 5Apk of unbalance when still in ClassA and more if the amp is taken into ClassAB.

I think you need an active virtual ground.
Can some recommend a current capability for the active virtual ground?

[aspringv]

... <browses internet> Virtual Ground Circuits is a pretty good primer around what you're talking about as I understand it, except that we're talking about 20x the current.

Ok, so I need to find some sort of high current active virtual ground schematic... Or would finding some sort of regulated PSU circuit that doesn't feature a CT on the transfo be an easier task?

Hmmm...

[aspringv]

Pros/Cons on Split supply vs. Virtual Ground

Some reading on the same idea... I'll have a proper look at the circuit detailed in page 2 of that thread ASAP...

[godfrey]

Originally Posted by AndrewT

hi,
those 100r will allow a low current unbalance for an opamp or similar low current circuit.

Each F5 could require 5Apk of unbalance when still in ClassA and more if the amp is taken into ClassAB.

I think you need an active virtual ground.
Can some recommend a current capability for the active virtual ground?

Hi Andrew

The high AC currents will go through the capacitors, so the resistors only have to deal with DC offset current, which should be very low.

Performance should be acceptable although an active solution would be better.

The circuit below could be good. The opamp is used as a unity-gain buffer. R3 and R4 simply limit it's loop gain for stability.

R5 is included to ensure that the bulk of the current flows through the caps, so the opamp will only have to deal with the relatively small DC offset current (and some low frequency ripple).

R5 should also help with opamp stability, and is low enough to ensure minimal DC offset voltage.

Regards - Godfrey

[AndrewT]

Originally Posted by godfrey

.........the resistors only have to deal with DC offset current, which should be very low.

Performance should be acceptable although an active solution would be better.

your solution shows a 7A chipamp. Why, if it only has to handle the very low DC offset?

[godfrey]

Originally Posted by AndrewT

your solution shows a 7A chipamp. Why, if it only has to handle the very low DC offset?

In the active circuit (post 8), all of the loudspeaker return current flows through
C3//C4//R5. Most flows through the capacitors but at low bass frequencies a significant amount flows through R5 (and the opamp).

For example, with the values shown:
C3 = C4 = 10000 uF, R5 = 1 ohm.
At 32Hz the reactance of the capacitors is about 0.5 ohms each, i.e. 0.25 ohms for both in parallel, so the current through R5 will be about a quarter of the current through C3//C4, i.e. roughly a quarter of the total loudspeaker current.

e.g. for a loudspeaker current of +-2A (1.4A RMS) at 32 Hz, the current through R5 will be about +-0.5 A (350 mA RMS).

The suggested LM3886 may be overkill, but some sort of power device is needed, ideally capable of at least a couple of amps peak current.

As always, component values can be changed to taste. e.g. although I'm not sure what the recommended PSU for an F5 looks like, I expect it involves more than 10000 uF per rail.