Edit: i just tested measuring a battery by both either using only 1 terminal on probe and both. With connecting both the probe and ground i got the accurate voltage, with just the probe itself the result is in the screenshot.
So now my concern is when i cannot use ground to probe the circuit.
Sorry if the following sounds harsh, just trying to be clear:
What you are doing is wrong. Not even close to correct.
The way you're using the scope will frequently result in what you are seeing. Let me try to explain.
Your scope isn't magic. The scope needs *two* connections to the circuit. Just like a light bulb needs two connections to a circuit to turn on, or a voltmeter needs two connections to the circuit to read a voltage (or ohms, or whatever). The oscilloscope is EXACTLY LIKE THAT, it's just a much faster voltmeter which can graph the voltage over time.
What you're effectively doing is plugging your 'voltmeter' onto the circuit by sticking the red lead on the circuit and the black lead into the ground lead on the mains plug. In order for you to get any sort of reading, your device under test has to also be grounded to the mains, and the 'voltmeter' is going to read the combination of the voltage from your circuit and any noise which is on the ground wire. This is basic electronics. Now substitute 'scope' conceptually into this thought process and you'll see why you're getting the noise.
In your 'use just one probe on the battery' example, you'll never get a reading. Do you expect a voltmeter to read if you're only connecting the red lead? You shouldn't expect any more from your scope.
Along the same lines, if you did isolate the ground from the oscilloscope like you were asking, and not connect the ground lead, there is absolutely no way for the oscilloscope to EVER get a reading - because you've just eliminated the one, very crappy, return path for the signal. Just like unhooking the black lead from your voltmeter.
The *CORRECT* method is to ground using the ground clip on the probe that the oscilloscope gods gave you. It's so important that there are numerous scholarly papers on it. Even the 'how to use an oscilloscope' document at
https://learn.sparkfun.com/tutorials/how-to-use-an-oscilloscope/using-an-oscilloscope contains this gem:
The first key to probing a signal is finding a solid, reliable grounding point. Clasp your ground clip to a known ground, sometimes you may have to use a small wire to intermediate between the ground clip and your circuit’s ground point. Then connect your probe tip to the signal under test. Probe tips exist in a variety of form factors – the spring-loaded clip, fine point, hooks, etc. – try to find one that doesn’t require you to hold it in place all the time.
I noted down a ways you mentioned that sometimes you don't feel that you can find a useful, or safe, ground on the device under test. I've probed a lot of devices, and can't say that I've ever encountered a situation where I couldn't figure out a better ground than just hoping that the AC ground is good enough. I have encountered a lot of places where I wished I had a *better* way to connect the ground.
So here are some notes which might help:
The ground wire is short for a reason. In an ideal world, you would ground the clip electrically close to the signal you are probing, and use a separate ground for each signal to help eliminate any ground noise in your measurement. With that firmly in mind, there isn't any reason you can't cheat a little bit where you just care about basic readings. Just be aware that any noise you encounter may be an artifact of how much you've cheated. The problem you're running into is that what you're trying to do (hope the electrical ground will suffice) is way over the line as far as what is likely to work.
Personally I try to ground each oscilloscope probe to a common ground point on the circuit. I will use various alligator clip test leads or similar to extend the cables just as needed. This seems to be the best compromise, especially if I'm watching signals on an ongoing basis, and not just probing around the board.
In the case where I am experiencing difficulty probing using a probe while keeping it's ground clip connected, I'll occasionally make sure I have at least one probe attached to the circuit (and oscilloscope) with a solid ground, and then use a second probe without the ground lead connected to probe signals. This way, the return path is at least 'close' - i.e. through the second scope probe. I know when doing this I WILL end up with more noise on the signal - but this is perfectly fine for where you're looking at signals of a reasonable amplitude, as long as you recognize this as a likely problem.
In your case, I would suggest that a good assortment of useful grabbers and clips and adapters would help greatly - this would help you get a solid ground point. Every device has somewhere which you can connect to. For the larger devices, don't overlook the metal chassis in some cases - although that too may have more noise on it.
Any time you see noise, you should instantly think 'need better ground'.
Some useful links for grabbers:
http://parrot-invent.com/shop/http://muellerelectric.com/product-category/grabbers-plungers/https://www.soselectronic.com/test-probes?brand=HIRSCHMANN-SKSOne other item which I purchased was a new set of 4900 series scope leads and an assortment of adapters from probemaster. See
http://probemaster.com/oscilloscope-probes/4900-series/accessories/ for a nice chart of what you can do with the adapters they make.