coupling caps and a bipolar power supply?

[gaussmarkov]

i've been learning about capacitors and i have a question that i hope someone will answer
about coupling caps.  suppose that i have a simple booster circuit where the amplification has
no DC offset.  as i understand it, i can do this with a bipolar power supply, one pole at 4.5V say
and that other at -4.5V (two 4.5V batteries in series).  if that is correct, then would it be sensible
to remove the coupling caps on the output and input of the circuit?  or do they do something
more electronically than "containing" a DC offset inside a gain stage?  maybe this it is theoretically
o.k. to omit the coupling caps but still needed in practice?  like real batteries may not have
ideally equal voltages?

i understand that the coupling caps have audio consequences, too, and that i might want them
for those effects.  this may not be a well-posed question ... so i appreciate any help with just
asking my question.   

cheers, gm

[slacker]

Yes if there's no DC on the input or output you don't need decoupling caps. You can do this with opamps and probably transistors using a bipolar supply. You can do it with jfet circuits using a single supply because most of the time the gate of the jfet is only connected to ground via a resistor and has no DC on it.
You need to bear in mind though that if you plug something into the input or output that has DC on it, like an old pedal with leaky caps for example, then this will add a DC offset to your circuit which might upset it.

[R.G.]

Good question and a good set of theorizing about the answer.

Your suspicion that caps may be required in practice is accurate.

Opamps have a datasheet spec for "input offset voltage". This is the amount by which the two inputs are mismatched, so that if they are both at the same voltage, they still appear to have the input offset voltage between them. Opamps used to be so bad that there were pins set aside for an offset adjustment pot. The idea was to null out the input offset with this pot so you could have 0V input really be 0V output. Later, opamps got so good that the input offset voltage was trivial. It's usually in the range of 1-25mV today, depending on which opamp you have.

If you use the opamp at a gain of one, then for an input of 0V differential, the output does not sit at 0V, but at the input offset voltage. If you use the opamp at a gain of 100, the output sits at 100 times the offset voltage. This is not too bad for a single opamp, but if you have a chain of opamps with gain in each one, the DCoperating point can be pushed to one side by the input offset voltages being DC -amplified. Inserting a capacitor between stages breaks this chain of amplifying offsets, and helps keep things in bounds.

Circuits other than opamps may be used in bipolar circuits. In general, these have worse offsets than opamps, and must be AC coupled to avoid offsets which mess things up. The exception is when a circuit is set up with an opamp "servo" circuit in which a DC-accurate opamp measures the DC offset on the output of the circuit and inserts an offset voltage at the input to make the output offset be nearly 0. It's kind of an automatic DC nulling pot.

Did that help?

[gaussmarkov]

great!     thanks very much.  little by little ... 
 
i did not pick up on the op amp datasheet spec.  that's nice to know.  i suspected that my question
was missing something.   

cheers, gm

[R.G.]

There's another way to get offsets as well.

Opamp inputs don't need much bias current, but they do need some, for bipolar opamps anyway. An LM741 opamp is listed as having an input bias current of 1.5uA. So each input sucks in maybe 1uA from somewhere. If the DC resistance path from the input to wherever the bias current comes from is not the same for both inputs, that may add to or balance out part of the input offset voltage. One uA times a 1K imbalance is one millivolt, so a 6K imbalance in source impedance for the input bias can double the rated input offset voltage.

That's why you will see resistors between the + input and its bias voltage in inverting-only stages. It's there to balance the DC bias on the inputs. The - input is generally supplied its bias through the output of the opamp.

[gaussmarkov]

That's why you will see resistors between the + input and its bias voltage in inverting-only stages. It's there to balance the DC bias on the inputs. The - input is generally supplied its bias through the output of the opamp.

this is the sign of a great teacher:  take a reasonable, but confused, question and turn it into an interesting question.   thanks for this, R.G.  i have to laugh (at myself) because i thought op amps were much, much simpler to understand than transistors.  maybe they still are, but this clears up something i had not reached the point of even musing about.   

and i see now that my concern about different battery voltages was off.  their values affect how far the signal can swing in each direction, but not DC offset. hmm, i guess it wasn't even clear that i was confused about that.

[R.G.]

The human mind is a wonderful thing. An engineer might have built it. The mind works well when it has just the right amount of information for the level of complexity of the task involved; and it is very capable of working at higher and lower levels of complexity when needed.

For instance, using amplifiers. At a very high level, we want, say, an amplifier with a voltage gain of 10, and the details of how we get that are not too important, because we're designing a bigger thing.

When we get the big picture thing done, we then need to think - OK, how am I going to actually make that gain-of-ten amplifier? So we unplug our minds from the big picture, and go one level deeper, and design a gain-of-ten amplifier. We have many choices - single and multiple bipolars, single and multiple JFETs, compounds of the two, opamps, dedicated amplifiers, lots of things. From our knowledge about amplifiers, we choose what we need to build and make it the simplest, cheapest thing that will do the gain, voltage swing, and frequency of the amplifier we need. At this level, opamps ARE simpler than transistor circuit.

Then we go to a deeper level yet. What is the effect of this amplifier on the circuits next to it. What is its input impedance, output impedance, DC input and output levels, and power supply needs. Can we simplify the overall thing by choosing to make the DC output of the opamp be what is needed for the following circuit to operate and remove capacitors? And at that level, we need to know the details of exactly how the opamp as a circuit on its own needs bias voltage and current, how those might drift, what noise it makes, and so on.

There is a hierarchy of different levels of abstractions. Not being able to see the forest for the trees is a GOOD thing if what you want to do is find a nice walnut tree to make some fine furniture with. At that level of abstraction, seeing the forest gets in the way and if you're not able to see the trees for the forest, you have a problem.

Being able to work at and shift from level to level in a hierarchical structure of abstractions is one thing I've seen in really good designers. These guys can argue the philosophical advantages of common protocols, shift hardware abstraction layers in the computer that does the commo, then think about the workload partitioning schemes in the OS and how it shifts commo loads from processor to processor as users instantiate and dissolve virtual processors among the real processors and what that effect will be on bandwidth. Or something similar... 

Think in layers. At one layer it's a vehicle, just a way to get from point A to point B. At the next layer, it's a personal vehicle, and you can make it go from A to B anytime you personally want to. Next layer down, it's a car/truck/SUV/chick magnet whose configuration and syling matter; next layer down it has certain acceleration/deceleration/reliability/economy features; next layer down are the subsystems - engine, gauges, brakes, transmission, unsprung weight, shock damping ratio, body stiffness; next layer down it's mechanical fits, bearings, lubrication. Next layer down you're getting into chemistry, metallurgy, heat transfer, wear resistance of surfaces, corrosion, and so on.

The human mind has what amounts to an infinitely adjustable magnifying glass. We can think about things at the level of the entire universe at a one moment, sub-sub-atomic particles the next, and propose bridges between the two.

[Ben N]

Wow--that post that is applicable to a whole lot more than stompboxes.  Food for thought--thanks, RG.
Ben

[gaussmarkov]

yeah!  the mind is an amazing thing.  i have always been fascinated by its ability
to conjecture.  sometimes intuition yields amazing insights.  of course, sometimes
intuition is way off, too.  but it's an intriguing phenomenon all the same.

cheers, gm

[idlechatterbox]

QUOTE:
"sometimes intuition yields amazing insights.  of course, sometimes
intuition is way off, too.  but it's an intriguing phenomenon all the same."

Agreed. Intuition, if there is such a thing, poses a real puzzle, since it suggests that we can know something without knowing that (or how) we know it.

Some believe that logic and mathematics are "intuitive," in the sense that we don't so much learn them as we are "reminded" of what we already knew.