Quick Op-amp Questions

[Raven Luni]

Greetings,

For op amps that require a dual supply, is a simple resistor divider enough? If so, what values should be used for best results (high or low?) and can the same divider be used for more than 1 amplifier or do they each need their own?

 

[KrisBlueNZ]

When you're using an op-amp that requires split supplies, you generally have a "half-way ground" rail that is half way between the two supply voltages, and you can usually generate this using a resistor voltage divider, often with a decoupler from the half-way ground to the negative rail if inputs and outputs are actually referenced to the negative rail. If they are, in audio applications, you use series capacitors to couple a signal that is referenced to the negative supply rail into the op-amp circuit, and back out.
You can also use a supply splitter such as the TLE2425 or TLE2426 (not sure which is which, one of them splits the rail at 2.5V and the other splits it in half). These generate a "firm" ground rail, and you can draw some significant current from it. With a resistor voltage divider, if you draw current from it, the rail "moves around" relative to the actual positive and negative rails, which may or may not be a problem depending on the design.
The lower the resistances are, the stiffer the rail will be, but the more power you will waste in the resistors. Values of 10K+10K are common. You normally use the same half-way ground rail for all ICs.
Look at some application notes (National Semiconductor, Texas Instruments, Maxim etc) for examples of how it's done.

 

[Harald Kapp]

Her's someting to read up on power supply splitters or virtual grounds:
http://tangentsoft.net/elec/vgrounds.html

Since the output of a dual-supply OpAmp is usually referenced to GND, any output signal will return to GND. This means that the output current will flow to the virtual GND in your application. This current will develop a voltage drop across the divider resistors (I think that's what Kris meant by "stiffness" of the virtual GND). So the virtual GND will follow the signal from the OpAmp's output - not desirable. Therefore I suggest you use one of the buffered techniques.

Harald

 

[Raven Luni]

Ah so no easy solution then. I remember using a variation of the buffered op amp circuit from that page in my bat detector but I remember having to increase the feedback resistor and capacitor values significantly to stop it oscillating. It also needed some large capacitors on the output where that page says not to use them.

I might order a bunch of those splitters to make things simpler. Thanks for the info

 

[BobK]

What are you trying to do? I have often used opamps with only a resistor divider to make a 1/2 V reference. The power pins are connected to V+ and ground. The only problem with this configuration is that you cannot output ground (or V+). If the output can stay near the mid voltage range (like in an AC amp that is capacitor coupled) it works just fine.

Bob

 

[Raven Luni]

Its for my PWM projects. I'll be using op amps for various things including amplification, wave generation, comparison, active filtering etc. I'll also be performing logical operations on some of the outputs so I'd say thats going to need a pretty stable supply

 

[BobK]

Okay then, you are probably better off with dual supplies.

Bob

 

[Harald Kapp]

Or with a rail-to-rail OpAmp using a single supply.
If you need to generate e.g. a sinusoidal waveform, you can always add an offset of 1/2*Vcc to the signal which gives you a signal swing from 0 V - VCC with 1/2*Vcc being the logical zero for the signal but without the hassle of a dual supply. I've been doing this for 1% exact measurement circuits.

Harald

 

[KrisBlueNZ]

Another option is a negative rail generator, that generates a negative supply rail for your op-amps from the positive supply.
You can do this with an inductor-based switching regulator (see National Semiconductor, Maxim, Linear Technology etc), or a switched-capacitor charge pump device like the ICL7662, which is available from Maxim. Maxim make quite a range of switched capacitor devices; have a look at http://para.maxim-ic.com/en/index.mvp?tree=powersupplies under "capacitor charge pumps".

These articles should also interest you:

"Virtual ground circuits": http://tangentsoft.net/elec/vgrounds.html

"Provide a negative bias voltage to low-voltage amplifiers to improve their near-zero-volt performance": http://www.eetimes.com/design/autom...s-to-improve-their-near-zero-volt-performance

"Considerations In Designing Single Supply, Low-Power Systems Part I: Designs Using Ac Line Power": http://www.analog.com/library/analogDialogue/archives/29-3/consider.html

 

[Raven Luni]

Lots of useful info - thanks. I went back to the splitter I used in my bat detector and made some refinements - stripped out some stuff and took the feedback directly from the output. It seems stable with light loads but I've no idea how it will perform with audio stuff...

 

[KrisBlueNZ]

I'm not too keen on that op-amp/discrete supply splitter circuit... I think the four capacitors on the output will make the design very sensitive to supply noise. It's not normal, and I think a very bad idea, to have capacitance to ground (and/or positive supply) connected straight to the non-inverting input of an op-amp (unless it's being used as a comparator). This (a) delays the negative feedback from the op-amp's output, which should cause instability, and (b) injects any power supply noise directly into the op-amp, bypassing any negative feedback. Any op-amp experts out there have an opinion?

 

[poor mystic]

One way to go ahead is to use a multiple op-amp package, and use one of the amps as a voltage follower, as Harald has suggested. Use a ladder of 2 equal resistors between V+ and 0V as the voltage which your follower will follow, and hey presto! A virtual ground!

 

[poor mystic]

... like this... (see attachment)

 

[Harald Kapp]

In fsct I did think along the lines of the circuit as shown by Raven. A one transistor circuit as proposed by by has the disadvantage that it can't regulate downwards, meaning that there is no path for current to V- from the output of this stage.

Harald

 

[Harald Kapp]

In fact I did think along the lines of the circuit as shown by Raven. The one transistor circuit has no path for current to V-. Thus any current flowing into the virtual GND will raise the potential there.

Harald

 

[poor mystic]

Yes, that's true.

 

[Raven Luni]

Well I made a simple mic amplifier to test it and it works great - no oscillation and almost no noise when switching in additional loads