Differential vs Single

Hmmm, Maybe I don't know what I'm talking about ;/ Looking at a schematic for an ADC that what looks like a single to differential conversion uses a buffer and two differentiators in "parallel". I guess I'll try and re-read some books ;/

I've never heard of a "single ended op-amp". A buffer perhaps, but those are usually some sort of op-amp with the output tied to the -in anyway. Even so, if it has gain, you could contrive something with, say, its power pins bootstrapped as the -in. It's impossible to get away from a differential amplifier, simply because voltage IS differential (a.k.a. relative, delta, ... synonym potpourri here!). Of course to be fair, that may not be how the designer intended it, and the very performance you're interested in may suffer as a result...

An SE ADC, offhand, sounds like something that accepts something against ground. It gets buffered or whatever, internally, and the ADC works away at that to generate the digital form. In contrast to a differential input, which might be little more than an op-amp, probably with a feedback / "comp" pin provided to set gain.

A differential amp has an output voltage O = G * (x - e*y). The error in one or the other's gain (e is a number nearly equal to one) is the common mode rejection ratio: if x = y, then O = 0 should be the result, regardless of the value of x or y. Any difference in the value of x and y is, of course, multiplied by the gain factor, G. IC op-amps are designed with gangbusters gain (>80dB or >10,000) and relatively bad stability, so they aren't useful as open loop amplifiers.

Two transistors, a constant-current tail, and some load resistors -- like the CA3026(?) -- makes an amp of reasonable proportions, including gain. Collector-base (or gate, or grid) and emitter (or source or cathode) degeneration can be applied to reduce gain and improve distortion and frequency response, making a less general substitute for the explicit feedback resistors applied to op-amps. Because bandwidth is excellent, gain suitable and outputs balanced, such gain stages are commonly seen in oscilloscope circuits.

Tim

-- Deep Fryer: A very philosophical monk. Website @

used

the

There are fully differential amps around (Burr Brown, now part of TI, makes them - the OPA1632 for example)

That page also has links to some excellent app notes.

The noise reduction advantage of differential pairs is rejection of common mode signals. As an example, take a single ended signal referenced to ground (at signal frequencies) and pass it through a noisy environment. The noise it picks up can't be (properly) eliminated and even some noise reduction takes some circuitry. A differential pair, on the other hand, in this environment, will induce the same noise into both wires, assuming a perfect pair. As the difference between the two (for a theoretically perfect pair) is zero, the noise is effectively eliminated because what we care about is the difference between the signals.

No diff pair is perfect, of course, so we spec the common mode rejection

- see any op amp data sheet and it will spec the maximum common mode voltage and usually the rejection.

A diff amp can indeed be used as a single ended amplifier by tying one input to a fixed level, and that's precisely what is often done with op amps.

Cheers

PeteS

I guess I wasn't clear. What I don't understand, and can't seem to find any information on, is how the inputs differ to the different types.

Suppose, for example, I have a guitar pickup which is essentially an inductor. I have two wires comming from it. Can I connect one to the + and the other to the - in a differential scheme and it will be ok?

Could I ground one and then it will be single ended?

essentially something like

+----------- (-)\\ | | \\ Pickup | >

| | / +----------- (+)/

Is that essentially a differential configuration(I didn't add the feedback).

If I were to ground the + side(or even the -?) do I then get single ended?

I understand theoretically why differential inputs are better than single ended w.r.t noise in an op amp but I do not really understand how hooks up single ended vs differential inputs.

We know that an op amps output is essentially O = G*(I1 - I2).

Now if noise enters the picture will will essentially effect I1 and I2 equally. So if N is the noise we have

I1' = I1 + N I2' = I2 + N

and O = G*(I1' - I2') = G*(I1 - I2)

so the noise is rejected.

But I don't how the inputs themselfs are configured for single ended verses differential. All I can see is that in the case of single ended I2 is grounded so that one has O = G*I1'. Is this simply the case? Essentially to ground or not to ground?

In the case of the guitar pickup I wouldn't need to ground and therefor I should get pretty good noise rejection along the cable?

Thanks, Jon

Yes. Inductors are perfectly good at doing either. There are some finer caveats that can apply; shielding for instance. A balanced inductor is typically centertapped, and the CT connected to shield/ground. The cable might be a twisted or straight pair, with braided shield ("twinax" I believe). This way the voltage with respect to the surroundings (which presumably are around ground level) is equal and opposite for each signal line, which are both shielded anyway for good measure. You can send it down two wires, but the common mode basis isn't well defined -- if you send the coil to the op-amp, like this,

the voltage will drift around and do nonlinear things to the amp. (An ideal amplifier, with inifinite common mode range, will accomodate this; essentially, even though the common mode voltage is undefined, we could care less because we have infinite range to sit in. Leakage resistances (which are likewise undefined) will stabilize it somewhere!)

If you don't have a CT, you might add two (equal!) resistors, in series, across the balanced leads, to generate a virtual ground reference. If not at the pickup and cable shield (the best place is where it starts), this might be at the op-amp itself.

Note that, if your pickup's body is grounded to the "negative" signal lead, it adds capacitance to one side and becomes a problem. This is one of those finer things I mentioned.

Grounded +in gives an inverting SE amp, grounded -in gives noninverting.

verses

to

Yes, that's how noise is gathered. Note also that coax carries the same induced noise voltage on both leads, but it is introduced on the shield, so that by grounding the shield, the center conductor's voltage follows, for all intents and purposes. (This is most valid at RF, where the cable's length is significant with respect to signal wavelength, and where its length inductance is large enough to be ignored. Indeed, you can hook up a piece of coax backwards, send a pulse down it and find it inverted at the other end.)

Only if the pickup is appropriate (as mentioned) and the cable is likewise balanced (twisted pair would be typical). Shielded cable is still advantageous anyway. If it is designed for SE operation, you are better off with coaxial cable.

Tim

-- Deep Fryer: A very philosophical monk. Website @

hmm... Suppose I did want to use a guitar output that uses coax like(not sure what the cabling it uses is actually called) for its output and I hook it up to an op amp that is suggested in any book. Either non-inverting or inverting(I heard there is a difference though). Is it wrong? Is it bad? It seems like what your saying is that what introductory books teach is not really correct or just over simplified(which I wouldn't doubt)?

I think this is usually done though? Maybe the electronics of a guitar can be improved greatly?

Thanks again, Jon

Oh, I guess I didn't read that part ;/ Heh, thats what I wanted to know.

Thanks, Jon

hook

Assuming the amplifier is stable and has a useful gain (typically 100 to 10k for guitar apps), it'll just amplify whatever comes down the line. External noise (that is, whatever tries to couple crud into the cable) is shielded (or shunted to circuit ground, same difference*), so it's about whatever comes off the pickup.

*This is about the only grammatically correct way to use this expression, by the way. I HATE IT when people say "same difference" when they mean "same thing"!

This is effectively the same thing as balanced lines, where noise couples equally to both lines (due to their geometry), and since the differential signal is low impedance (at least, it's supposed to be!), it can't really have high impedance noise stuff coupled in. The mutually open geometry of a twisted pair is naturally balanced, just as the enclosed coax geometry is naturally single ended.

What, exactly, do they teach?

can

SE is more convienient. If you wanted to go to the trouble of making a balanced pickup, you could send it through balanced cables, and might get somewhat better noise rejection. You do have to be careful of where the metal bits go (like the pickup's body and magnets), which should be to ground (cable shield).

Tim

-- Deep Fryer: A very philosophical monk. Website @