amp picking up noise

I recently designed a 10Watt amp for my electronics design lab. The amp has a diff amp input, class A stage, and final class B output. I have been playing music through the amp and the sound is excellent (especially for my first amp design). The only problem is that noise is being picked up somewhere in the system and amplified when there is not a cd player attached to the input. Does anyone know how I can eliminate this noise? I was thinking a large resistor and a cap in series from the output to ground. Would this work or would it ruin my frequency response?
I'd say to hook the circuit up to an oscilloscope and start prodding the innards. Squeals are Bad Ju-Ju...
Most amp circuits aren't really that sensitive to being out of a case--the Aleph 2s I've been fiddling with are sitting in the middle of the floor without a stitch on, naked for all to see. Preamps and other low level circuitry are more prone to trouble, but that's not the case here.
Does this circuit have an RC, RL, or RCL network at the output?

The input impedance is about 40Kohm (20K in series, 20K to ground). The sensitivity of the amp is very high 100mVpk-pk input = 27Vpk-pk output. Possibly if I reduced the input impedance and the sensitivity? What is the maximum voltage from a preamp usually? 4Vrms?

This amp was an assignment for a class of mine and not necessarly an audio amp, but I am trying to tweak it for audio since it sounds nice (when everything is connected).

It does not have any networks at the output.
I don't see any reason that 40k at the input should be a problem. The Alephs on the floor behind me are 47k (modified input network, that's not the stock input Z), and my tube circuits are 100k. I've seen numerous amps with 470k inputs (tube circuits, mostly), so I wouldn't lose a lot of sleep over that.
I can make up a scenario or two where excess sensitivity might be a nuisance, particularly if you're in an RF prone area, but I'm not sure that's the problem, either. However, I am curious as to why you designed it to be so sensitive--we've talking something on the order of 48 dB gain, here. Most amps are between 20 and 30 dB, with a tendency to cluster in the 26-27 dB range.
1) Why so much gain? Do you have a particularly low-output preamp?
2) How much feedback are you using? (N.B.: We could easily start a war, here. My feeling is that NFB is best in small doses; some is [usually] necessary, but it does subtle, unfortunate things to the sound quality. I use 10 dB NFB in my tube circuit, and the frequency response is already out to much more do you want? Others feel that NFB has no downside and the sky's the limit. It's a cheap fix for those who want low distortion figures.)
3) Why no output network? Most (not all) solid state designs require one for stability in the real world.
4) Are you using a cap at the input to decouple from any possible DC input?
5) What's the upper -3 dB point for your circuit?
Back to you...

This amp was an assignment and the gain that I have was a part of the specs for the assignment. Obviously, 48dB is a little much for an audio amp. I will be changing the gain so that it can work more practically as an audio amp.

Right now the feedback is a 120K from the output with a 100uF cap and 560ohm in series to ground. So right now the gain is really high. I will probably change that to more like 20K feedback with a 1K and cap to ground. I'm not sure what you mean by NFB.

What do you mean as far as an output network, like a RC series from output to ground? What calculations are use to determine the components for that?

Right now I am not using an input coupling cap. Are they always used? I was worried about hurting my bandwidth so I left it out.

The bandwidth is almost DC to about 30K (-3dB point). Remembering electronics 2, I think if I reduced the gain that should increase the bandwidth.

Thanks for your help. I have done speaker building and design for a long time, but amp design is pretty new to me.
If 48 dB is what the teacher wants, then give the teacher 48 dB...get the grade first, then modify the circuit later.
NFB is common shorthand for Negative FeedBack--saves having to write it all out.
Output networks are just there to stop oscillations. If, by some miracle, you've got a circuit that's stable without one, then by all means don't go mucking up the sound with a bunch of extraneous nonsense at the rump end of things. Basically the intent is to short high frequencies to ground with a cap and/or stop them from going out to the speaker with an inductor.
Input coupling caps are to be avoided, if possible. The best cap is no cap, and all that. The concept is that you want to block possible incoming DC that might either change the operating point of your front end or show up at the output of the amp as a really huge DC offset that will turn your woofer's coil former into something resembling popcorn. The decision is yours to make--sound quality vs. safety. Me, I regard all components as guilty until proven innocent. I've heard far too many things chipping away at my signal quality that "shouldn't" be doing so. Good quality caps aren't nearly as bad as some other nasties, but I still prefer to avoid them whenever possible. Besides, they're bulky and expensive. On the other hand, I lost a subwoofer driver once upon a time when the rail in a crossover got loose and went through the amp. You pays your money and you makes your choice.
Regarding bandwidth vs. gain, I assume you're referring to a Bode plot. Yeah, all things being equal, lower gain means wider bandwidth. But...things ain't always equal. The "Specifications Are Everything" crowd run the gain up to where you can look someone on Mt. Everest eyeball-to-eyeball, then knock it back down to something reasonable by using lots of feedback. They feel that NFB does no harm to a signal. I've got a different view, in that I've played with circuits and listened to the subtle things that NFB can do to a signal. Most solid state amps get rather spastic if you try to run them open loop. Tube amps are much more polite beasts and are quite happy to run with no feedback whatsoever. All you do at that point is add a pot in place of a fixed resistor and play to your heart's content (being sure to readjust the volume, as adjusting the NFB, by definition, will change the gain of the circuit). My advice is this:
--What kind of music do you listen to? Rock/pop/rap, and the like? Most (there are a few exceptions, but not many) recordings of that type lack the kind of detail that NFB destroys. Go ahead and use all the NFB you want. Bluegrass, classical, jazz? Use only enough feedback to get the job done.
--How's the rest of your system? Party machine? Mid-fi? High-end? Choose lots of feedback for party machines as the damping will increase. If you have now, or intend to have in the future, a high-end system, consider going lightly on the NFB.
Beware--if you shoot for mid-20's dB gain, and intend to get there from where you are now using feedback only, you're building an oscillator, not an amp. It'd be really easy to tune the gain in at, say, 26 dB by adding 22 dB feedback on top of what you've already got. Your bandwidth will go from DC to light. Somewhere around 50-200kHz you'll pick up a nasty oscillation that will make short work of your tweeters. There are various ways to approach this, but a good place to start is to build in a rolloff at 100-150kHz, then add a network at the output.
The more elegant solution is to design less gain in from the git-go (thus hitting lower on the Bode than you would otherwise), then use NFB sparingly to fine tune the circuit, rather than as a hammer to bludgeon the poor thing into submission.
But the more immediate goal is to get the grade. Turn in a circuit that makes your teacher happy.

1) it could be possible that the noise is picked up at your connectors: in this case you can compare the "signal"-input with the ground of the connector (you can do this with a simple opamp)

2) I've had the same problem with a mixer: be sure that your power supply isn't placed in the neighbourhood of connectors or wires. If this is the case, you can also use a kind of COAX-cable to reduce the noise. It is always better to put the power supply in a grounded metal case.

3) If the noise still excist (after 1 and 2) and when it goes away after grounding the input, you can place a resistor (ex.: 10 Mohm) between the input and the ground.
When you use "jack-connectors" with a "build-in switch", you can ground the input automatically to the ground when you pull out the jack.

Best regards,