Converting bipolar supply to single

[jdub]

I've seen several threads that talk about converting a bipolar supply to single, but haven't seen any that fully clarify exactly how to do it (though maybe my search-fu is a bit off today).  Say, for instance, that you have a circuit powered by a +/-9V supply (i.e. +9V, 0V & -9V)... if you wanted to convert this to a single supply, would it be as simple as, for instance, keeping +9V the same, setting a vref of 4.5V to replace 0V (midpoint) and having 0V replace -9V?  Am I correct in assuming that in the bipolar supply one would have a rail-to-rail swing of 18V, and that if you replace that supply with example I gave you would be cutting your headroom in half (i.e. 9V)?  If that is the case, assuming the component voltages allow it, can one then use a single supply of 18V, 9V and 0V to preserve the headroom?  Am I on the right track here?  Thanks!

[Gurner]

You are right on all counts.... but bipolar has some advantages over single supply, the main one for me is that the impedance of the ground (0V) is as good as you can get, whereas once you start shifting the voltage up to mid-point you are exposed to the strengths/weaknesses of the components (&/or design of the circuit) providing the mid-point voltage. To give you an example...imagine an audio amplifier cct with a heavy current draw...the power rail sags/rises in sympathy with the signal current...which can mean your midpoint voltage (virtual ground) wavers too, which can feed into the signal itself .....that doesn't happen when your ground is true ground (unless your pcb ground layout is terrible!)

[jdub]

Thanks, G...just wanted to be sure.  I'm toying around with a bipolar phaser cct and want to try it with a single supply so I don't have to muck around with the extra stuff for a bipolar supply.  Tried powering with an LT1054 doubler/inverter but apparently the current draw of the cct is above the 50mA that it can provide  ... 

[Seljer]

You can however use a buffer on your reference voltage to lower its impedance. Even a spare half of an opamp might help.

[ashcat_lt]

You may already realize this, but for completeness...

The input signal is usually swinging around 0V, which works fine in a bipolar circuit.  If you want to work on the whole signal in a single supply unit, this needs to be biased up to the half-supply point.  This means a connection to the bias voltage at a place where you wouldn't find a ground connection in a bipolar circuit.  

Then, we normally don't want that DC voltage to "escape" from our circuit where it might mess up the action (or damage) of the source and/or load.  So you need at least two coupling caps.

Also, it seems like you need to be more careful about coupling between stages inside your circuit, since small offsets times large gains can equal large offsets.  This can mean more caps, and more connections to Vbias.  I'm not sure why this would be a bigger concern in single-supplies than bipolar.  It might not even be true, but you don't see as many interstage coupling caps in bipolar units.

[jdub]

Thanks for the pointers...I'm very much a hack, but tryin' to learn...so please forgive if my questions are less-than-educated!

@Seljer: Where would such a buffer be applied?  I'm thinking in terms of a resistive divider- would this entail simply buffering the output of the divider?  Could one use a standard transistor voltage follower for this purpose?

@Ashcat:

You may already realize this

Highly unlikely, my friend, highly unlikely...

If you want to work on the whole signal in a single supply unit, this needs to be biased up to the half-supply point.  This means a connection to the bias voltage at a place where you wouldn't find a ground connection in a bipolar circuit. 

Do you mean at the signal input, e.g. just before the input buffer?  Not sure I'm getting this.  I assumed that, in converting, I would just connect to vref all the points that were at ground with the bipolar supply.

So you need at least two coupling caps

Believe it or not, I think I actually understand this.  In this particular circuit (it's the Maestro PS-1A, BTW), I'm not sure whether I would need additional coupling caps between stages, but it's something I hadn't thought about, so thanks!

Once again, thanks for the help, folks.

[ashcat_lt]

I guess it depends on the buffer.  Some transistor buffers maybe self-bias without conversion?  An inverting opamp I think would work as described because the - input will follow the output up to the Vref at the + input.  A non-inverting opamp buffer will not work without an added biasing resistor because it will be trying to wiggle around zero.  Even a rail-to-rail opamp will only pass the top half of the wave.  Most real world opamp's won't get close enough to 0V to allow much of any real world guitar level signal through.

[jdub]

Just now got a chance to reply...thanks for the help Ashcat.  I haven't had time to play with it for the past few days but will try this weekend.