The following is simplified to hopefully make it easier to grasp:
To understand most of these plots you have to understand the difference between the time and frequency domain.
If we plot a signal in the time domain we have the time on the x-axis and amplitude on the y-axis.
A simple sine wave starts at t=0. The amplitude rises with increasing time until it reaches its peak, then falls again, crosses the zero line, goes below zero until it reaches the negative peak, decreases in amplitude again until it finally reaches zero again. This is a full cycle (360°).
If it takes a sine wave 0.1s for a full cycle the frequency is 1/0.1 = 10 Hz. In other words, we get 10 such cycles per second.
Now let's look at a more complex waveform, a square wave:
A perfect square wave can be represented as an infinite sum of sine tones. The plot above only shows 3 harmonics. If the square wave has a frequency of 10 Hz the 1st harmonic (or fundamental) is a 10 Hz sine wave, the 3rd harmonic is a 30 Hz sine, the 5th a 50 Hz sine etc.
Also note that each higher harmonic has a lower amplitude. If you sum all (1st, 3rd, 5th, 7th, ... infinity) harmonics up you get the square wave that is shown above.
Now let's plot this signal in the frequency domain:
The x-axis now shows frequencies instead of time. The y-axis still shows amplitude, but in decibels.
So you can see there are sinusoids at 10 Hz, with lower amplitude at 30 Hz, with even lower amplitude at 50 Hz ... just like in the time domain plot above, though normally you just see the summed square wave in the time domain.
In other words, going into the frequency domain is like dissecting the timedomain signal and analyzing how much of the signal lies at a given frequency.
Now what do you see if you look e.g. at the Frequency Response measurement graph?
Assume you send a signal into the amp that contains all frequencies, so that you should see a flat line from 0 Hz to infinity in the frequency domain.
The FR measurement shows the frequency domain representation of the signal that is output by the amp.
Since DC (0 Hz, essentially a flat line with an amplitude other than zero in the time domain) is bad for headphones, since it just generates heat, many amps have a DC blocking filter. Depending on the design of this filter the amp may show some low frequency roll-off. That means that as you decrease frequency the amp will output a weaker and weaker signal until it doesn't output anything anymore at DC. If this roll-off starts in the audible range (above 20 Hz) and is severe you could potentially hear weaker sub-bass than it should be with your headphones.
There's also high frequency roll-off, for reasons I won't mention here. The audible range ends at about 20 kHz (20000 Hz), so again if the roll-off is not severe our above that it's not gonna cause audible problems.
A much more audible problem can be channel imbalance. In the Frequency Response graph, if you see one channel (red or blue) a dB higher than the other it means that all sounds will be played slightly louder in that channel.
Edited by xnor - 7/4/13 at 5:22pm