Single power supply - need help

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

I want to build an amp with continous 150w and max 300 watt power, and +/-30v. I have some circuits for it. My problem is the power supply because I have a 400va transformer, but it isn't symmetric, so it has not center tapped secondaries. It has only one secondary, so i have to make a virtual ground .

Could you give me a good, complete circuit between the secondary and the amp?
OK for 150 watt rms into 8 ohm you need 34 Vrms output. Do-able with a bridged amp running off +-30v.

For 300 watt rms you need 49 Vrms, or 69.3v peak, add 5v for losses, and make it 75 v rail in practice, for a bridged amp, or +- 37.5VDC, and that's assuming perfetc regulation, so your'e out of luck right out of the gate with +-30v.

Into 4 ohm You'll make it IF the transformer regulation holds up for 300 wat peak. For 150 watt continious, you should be good into 4 ohm. All this assumes a brided amplifier.........


I want to make an amp with 2 lm3886. (I know it is a chip amp, and not a solid state, but my problem is the psu)

Here is my transformer:

the front side, with the primaries

the back, with the one usable secondary

I created a little circuit to provide the 'virtual ground'.

Is this circuit good? Will be Unequal Virtual Ground Splits a problem???
How can I solve this problem- if it is a problem??
Your schematic for symmetrical supply (+ and - supply) is ok. It just needs double the storage capacitors for the same ripple as a full bridge set would.

However, your transformer seems less than perfect. It has a lot of secondaries, but I can't see any current rating for each secondary. If you have a 400 VA transformer, the rating is split across all these secondaries. You can't choose for yourself, which secondary you want 400 VA from.

Based on the (little) available info, I don't have a good feeling about this project. sorry.

If you look at the data sheet for the LM3886 there is an example of how to use the chip with a single rail supply. This would be far better than a virtual earth arrangement. The catch is the need for a high quality, large capacitor in the speaker output.
yes, i know this transformer is not the best, but I don't want to spend more money on this project so i will try it out.

There are 2 10,5V outputs on the transformator, with 3 and 0,5 amp rating. Maybe the 43V output can deliver 3 amp. And if it can, it is enough for the 2 lm3886 chips.

I will double the capacitance of those caps.

I'm afraid the output volts, the (+) and the (-) won't be equal (maybe it can be 2 to 5 volts difference). But tell me i'm not right.

The original circuit That I modified has voltage regulator ic's on each rail.

original circuit

The PS schematic is essentially a voltage doubler, which will give you something like 2 x 60 Volts. This should blow the chip amps to pieces!

If you want to use this particular transformer and these chips, single supply and bridging seems the way to go

Thank you for your reply.

I corrected the 2x60V mistake so it will be 2x30V with the circuits used below.

Those 2 circuit below are use batteries, I change them to bridge rectifier.

Here is the simplest circuit :


In this circuit the transistors are not appropriate, so i change them to 2n3055 and mj2955.
maybe the better circuit

Which circuit is better? Are these modified designs good?? if not , can somebody post me a virtual ground ps circuit?

Are the voltages on the output equal? +30 :: -30 or something like this: -34 :: +26 ???
The reason for the single supply being better is you are effectively only using halfwave rectification, which is going to mean you will have a lot of 50Hz ripple in the supply which will be difficult to smooth out without large, fast capacitors. You would end up spending more on capacitors than you would on a centre tapped transformer.
jaycee said:
The reason for the single supply being better is you are effectively only using halfwave rectification, which is going to mean you will have a lot of 50Hz ripple in the supply which will be difficult to smooth out without large, fast capacitors. You would end up spending more on capacitors than you would on a centre tapped transformer.

If using the same kind of capacitors, you'll just need twice as many, due to the slower re-charge cycles. (taking longer before they get recharged). Nothing fancy there.

As for the transformer, the 42V rails are not suitable, unless the chip can handle +/- 60V (which I understand it can't).
You might use the 10V secondaries for a proof-of-concept construction (to see if you got it right and everything works (except for lower outpot power due to lower supply voltage). For "real" use, it's not good.

Also, if using voltage regulators, make sure they can handle the involved voltage & current ( == power loss). Also, they have a higher output impedance then a "simple" supply with transformer and caps.

for Jennice:

I will use 2 x 15000 uF Capacitors for smoothing(is it enough?), and 2 x lm338 for voltage regulation.

What do you think the simpler or the more complicated circuit is the better? :confused:

The last two circuits provide +/- 30v, with voltage regulator +/-24
Only the first, hand drawn circuit gives 2 x 60V.
So I can use the 43v secondary...
I don't think the circuits you have shown will work. By doing your amp. that way the speaker current will need to be passed by the voltage divider. Those circuits can't pass that amount of current without adding a lot more capitance. If you must use the transformer you have then just build the single supply version of the circuit in a bridge configuration, and it will work for sure.

For that matter if you build a bridged amp you don't need the output caps. Just reference both chips to the same reference voltage. The speaker will be at half the rail voltage ref. to grd. but both amp outputs will be at that voltage so no problem.


The two later circuits won't work for power applications, but why shouldn't the first one work.
There may be a problem with the peak voltages being too low for the regulators, but I wouldn't use them anyway for this. Apart from the regulators, the original circuit should be ok I think.

However, if more power is needed (than can be used with this supply), I agree on the dual/bridged amplifier idea. In that case, I agree that the output cap is not needed. (I would check for output off-set voltage, though!)

The 2x 15000uF should be fine I think (both for full-wave and half-wave rectification).

You could instead just run in bridge configuration - full wave rectify the ac, to get ~60VDC, then buy an extra pair of 3886s and run them in bridge configuration - that way, no need for the 338s or an output capacitor, and also no need for heaps of capacitance on the supply rail. Cost will probably be about the same - the 3886s are only a few dollars more than the 338s, but you will save this when you don't need to buy extra power supply capacitors for the output capacitors. Oops - someone has already said this.

On another note, using the 338s won't work here anyway - you need isolated transformer secondaries to achieve this. If you are thinking that you could just regulate the 60V to (say) ~25V and the other ~50V to get a ±25V supply, it won't work - current cannot flow from the 50V regulator into the 25V one (as required for a positive output) - current can only flow out of the 338s.

Look at the data sheet for the single supply application. Now, some components here you don't need: get rid of the big 4700uF output capacitor. You shouldn't need the parallel inductor/resistor combination at the output either. BUT you MUST keep the capacitor Ci (connecting the feedback loop to ground), otherwise the output won't be centred around mid supply rail.

You need two amplifier stages, and you want the output offset voltages to be pretty much the same. Thus it would be a good idea to reference each 3886 (of the one channel) to the same voltage, i.e. the voltage divider formed by the 91k and 100k resistor, the 0.1uF capacitor to ground, and the transistor. So, it would be a good idea to only build one of these voltage divider circuits to supply both 3886s. However, the increased current requirement means we should half all resistors - just put two 91k in parallel for the top and two 100k in parallel for the bottom. The capacitor is a little bit too small in value in my opinion. Increase it to 2.2uF. So now you can reference two 3886s to this voltage divider (sort of like a virtual ground if you will).

Now all that needs doing is to invert the input to one of the 3886s. Just use a dual opamp package, connect one as a voltage follower, connect its output to one of the 3886s. Connect up the other opamp as an inverter, connect its input to the voltage followers output, and its output to the other 3886. That should be it!

I will try drawing some schematics, give me a few minutes.

Ok, here's the schematic.

For the opamp you can use something like a TL072 should be fine. If you must, you can use something more exotic, but the values of some of the resistors around them will have to be changed if it isn't a FET device - i.e. the 100k input resistor should be smaller at say 33k, and the ground at the noninverting input of the lower opamp should be connected by a 5.1k to ground instead of directly to ground (to balance bias currents).

Also please note that the opamps will require a dedicated suppply of ±15V. This should be easy with all those extra windings on the transformer - just use a LM317 and LM337 to regulate the positive and negative rails respectively. The 'yellow' and 'black' windings look ideal for this (assuming I read correctly that the yellow output is ~20V).

The 2N3904 transistor could probably be substituted with another small signal type - eg a BC546, just be careful of voltage ratings (i.e. probably want Vce,max to be >60V - 546 has 80V max).

Also, before connecting a speaker to the outputs, make sure that the voltage between them is less than around 0.1V.

I know that the diagram is a bit messy and that some of the diagram from the other side of the sheet has scanned through, so if you can't see what I have drawn, please ask!

My god that file size limit is annoying - I'm actually just under the limit and it still won't work!

Finally got it.



  • 3886 bridged single supply.gif
    3886 bridged single supply.gif
    82.3 KB · Views: 117
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.