LM3886 Power Supply Modification - Might Work?

[mattthegamer463]

So I bought a 60VCT transformer for $15, and the LM3886 amp I bought but don't have yet wants a 28-0-28 (56VCT) transformer. The board says it needs a 28-0-28 @ 5.5A transformer, which is 308VA I believe.

I'm also only running a 25W 6-ohm pair of speakers off it so less than 28VRMS is a good idea. Here is a voltage-cutting circuit I designed using what I believe the amp circuit uses.

The Amp on ebay:

http://cgi.ebay.ca/2-x-68-Watt-LM388...d=p3286.c0.m14

My idea image, placing some resistors in between the rectifiers and the diodes, is attached to this post. I had no clue what to put for the amplifier's total resistance, so I put 10k ohms. With calculations, R = V / I so R = 56 / 5.5 = 10.2 ohms, maybe that is more accurate. In that case, the voltage would be closer to 22.7V, says LTSpice.



So, is this kind of thing actually feasible? I think theoretically it could work, but the power-rating on the resistors could be huge. Would the resistors be ridiculous or something more reasonable like 10W?

Thoughts on this idea would be great. Thanks.

[iko]

While in the plot window in ltspice, press CTRL-A and plot V(n003)*I(R3). Check out the power dissipated by R3. You better turn C2 around.

Basically, it wouldn't work as shown.

[akis]

When you design a potentiometer you need to account for the current you will be drawing from it. Good practice says use the times ten rule, ie let the potentiometer carry 10 times more current than what you'll ever going to need, so that the voltage will not drop. But if you load the potentiometer too much, then the voltage will of course drop. If you intend to have aconstant load, then you can do the calculation so you end up with te exact voltage you need, but in that case you do not even need a potentiometer, simply a series resistance.

If you want the voltage to be "constant" then the times ten rule applies, and yes you will have to expend 10 times more current than what you're planning to use, in other words if you planning to use 5 Amps you'd better draw 50 Amps. Ouch.

Another way to do what you're trying to do is with a simple zener and a couple of transistors. You could replace the zener with a potentiometer, but the zener is better, will keep the voltage more constant.

[akis]

To explain the circuit I posted, it is a simple bridge + fitler capacitors. Ordinarily that will give us a DC voltage (on no load), say for example +/- 35 volts. But we need no more than say +/- 25 Volts. We use a zener of about 27 volts. We choose a resistor so that we have the appropriate current flowing through the zener, eg a 5W 27V zener wants 50ma, in that case the resistor would be something like (35-27)/0.050 = 160 Ohm. I'd use a 150 or 120 Ohm, since on load conditions the voltage from the filter capacitors will drop a few volts too. The zener now generates our 27 Volts and we feed that to the base of the transistor, and we have 2 transistors each will drop like 0.6-0.8V, say a total of 1.3-1.6 Volts, so the net out put will be something like 27 - 1.5 = 25.5 volts. Depending on the current we will then need to calculate the transistor ratings and heatsink. We may even need 3 transistors, depending on the amps needed and the zener used. One problem with the above circuit is it has no protection, if you short the terminals then you are risking burning the transistors, so you'd need some sort of a series sensor resistor and a few more components to short the zener in case of a short at the output (2 more transistors and 3-4 resistors).

[pacificblue]

Get a different transformer. If you don't, you will pay more on your future electricity bills, than you can save on the transformer, whether you waste the energy in resistors or regulators.

[akis]

Originally posted by mattthegamer463
So I bought a 60VCT transformer for $15, and the LM3886 amp

I had no clue what to put for the amplifier's total resistance, so I put 10k ohms.

Thoughts on this idea would be great. Thanks.

If the LM3886 were 10K load you'd be laughing, you'd need something like a 1K potentiometer.

But the LM3886 at full swing would be basically your speakers, ie 6 Ohms, not 10,000 :-)

Of course it is AC and not DC, so it would not really stay at 6 Ohms for too long, but you get the idea.

[mattthegamer463]

Hmm. These zener diodes are interesting, but I would need two in the realm of about 30W it seems, which is a lot of heat to get rid of.

What about something less beautiful but a little easier to implement, like a household light dimmer switch? What about removing 1/3 of the toroid's secondary winding, bringing it down to 20V RMS and 28.28V peak?

[pacificblue]

Originally posted by mattthegamer463
Hmm. These zener diodes are interesting, but I would need two in the realm of about 30W it seems, which is a lot of heat to get rid of.

Use akis' schematic. The transistors will have to get rid of the heat, not the Zeners.

Originally posted by mattthegamer463
What about something less beautiful but a little easier to implement, like a household light dimmer switch?

You will have to find one that is suited for transformer dimming. Then you will find that they don't cost less than a new transformer.

[mattthegamer463]

Originally posted by pacificblue
You will have to find one that is suited for transformer dimming. Then you will find that they don't cost less than a new transformer.

Well put.

Looking into this transformer unwinding, it sounds to me like the right way to go. Since the transformer is 420VA theres enough room for the increase in current, and wire gauges shoudlnt be an issue.

[annex666]

There are a few things I would consider changing about akis's design:

1) Change the Darlington pair to a complementary pair.
2) Move the reference to the regulated side.
3) Include an op-amp comparator between the reference voltage and a potential divider on the output.

All of the above will change the design from open to closed-loop - allowing you to compensate for varying Vbe drop with current draw that affects the current (no pun intended) circuit.