Vcc/2 Generation

[oddist]

Hi,
Just wondering if anyone here has ever used a transistor to buffer Vcc/2 for biasing purposes?  So, it would be similar to using an opamp, but the voltage divider would be designed to generate just over Vcc/2 to account for the Vbe drop of the transistor.  My main concern would be - is the transistor buffer able to provide as "stiff" a reference as the opamp.  If not, how bad, relatively, would it be? 
Thanks for any input.

[.Mike]

I'm not sure of the answer to your question, but...

TI makes a precision rail splitter that will create a virtual ground voltage from anything between 4 and 40 volts, and can source 20mA of current. It's available in TO-92 or in an 8-pin DIP.

They're $1.17/ea at Mouser: http://www.mouser.com/_/N-scv7?Keyword=TLE2426&Ns=Pricing|0&FS=True



Mike

[idiot savant]

the lm386 makes a good reference in a pinch since the output automagically biases to 1/2 Vcc.

[aziltz]

at what point should a design use something more than a simple voltage divider to give Vcc/2?

[Ripthorn]

The only reason I see using anything other than a simple resistor-based voltage divider is if you are planning on there being a lot of current going through it and if your load has a relatively low impedance.  Other than that, the simple divider should work fine.  You could also use a zener diode, but the cool thing about a divider is that even if your Vcc+ changes a little (as some do) you still get one half whatever you are sending to the rails.  Of course, if you are drawing very much current or have a relatively low load impedance, you should try something a little more stable and reliable.

[earthtonesaudio]

Yep, Ripthorn said it.  It depends on the current you want to draw from the Vcc+/2 reference.  The big advantage (in my opinion) to using a transistor (or op-amp) to buffer it is that you can use higher value resistors for the voltage divider (reducing quiescent current consumption and allowing for better RC noise filtering) while still supplying enough current at times of high demand.

The main advantage to an op-amp over a single transistor is the op-amp will most likely sink/source current symmetrically.  With single transistor + emitter resistor, the transistor supplies current from the positive supply, but the resistor has to supply the negative current.

[rustypinto]

In my humblest of opinions, if you're going to apply that much effort in making such a virtual ground, why not just make the "-V" like most op-amps are supposed to be biased at?

The main reason to do the resistor divider is because its cheap, and works in most applications. But all those nice common mode and power supply rejection ratios go right out the window and your noise immunity is all but gone.

I'm sure the virtual ground maker from TI has some decent noise immunity characteristics, but at $1.17, you can still get a charge pump around the same price to make the -V, where you now have increased headroom, better noise performance and better signal to noise ratio. All at the cost of extra current required from your single 9VDC adapter!

Awesome topic!

[Eb7+9]

Just wondering if anyone here has ever used a transistor to buffer Vcc/2 for biasing purposes?  So, it would be similar to using an opamp, but the voltage divider would be designed to generate just over Vcc/2 to account for the Vbe drop of the transistor.  My main concern would be - is the transistor buffer able to provide as "stiff" a reference as the opamp.  If not, how bad, relatively, would it be?  

I've used emitter-buffered pots many times in lieu of an op-amp buffered half-rail - see my Nyquist Aliaser schematic for an example ... the key is to get an idea of the impedances being driven into that ac ground from other sources in the circuit ... an op-amp can source a good 25mA of current  easy while here you're limited to about 1/10 of the bias current flowing through your buffering device ... with higher current you get noise though, hence the advantage of op-amp in high-current loading applications - but the need for this is rare in small-signal processing and you can also get some ground ringing from an op-amp configured with full NFB - which some audiophiles try avoiding ...

another way to look at it is in terms of relative currents - if you can estimate the max ac current passing through any branch to that ground and figure that the sum total is well under 1/10 of your emitter follower bias current then you can guarantee that it will never approach shutoff and thus count on a fairly constant Beta (hfe) ... to establish emitter bias current simply divide hald rail voltage by emitter resistance - ignore Vbe since it can be tuned out to give you exact half rail if you wish it

knowing this allows you to divide the driving point resistance of the pot (ie., less than half the pot's value) by a rough estimate of Beta ... for most high gain transistors Beta is well over 100, so on a 10k pot we'd get a Zout of less than 5k/100 = or 50 ohms ... if you want a much lower Zout on the bias source then Darlington that device and you're looking at Zout less than 5k/Beta^2 or .5 ohms in conservative terms ... in reality it will be even lower using high Beta devices

still, even if you're using low current-gain devices this "ball-park" 50-ohmAC target is way less than what we see in the non-inverting path of most op-amp circuits, or at least it can be designed for it ... usually you don't see a negative leg with less than 500 ohms of resistance (filters, noise sources, oscillators ...) ... so in many low-power circuit cases a 50-ohmAC ground is plenty good - i.e, the super low Zout of an op-amp would not give you much of an advantage or produce a significant sonic or operating difference ...

the idea can be taken further and used to produce some interesting tube-transistor hybrid signal-path topologies
funstuff ...

[nexion777]

A noob question: it is possible "rawly" insert the buffered voltage divider (using an opamp like tl071) in those circuits like the TS, where Vcc/2 is connected to the filter of the feedback loop,  and  to output volume pot?