They use an audio analyzer, Audio Percision is a popular one, Stereophile measurements always start off stating what gear they used. And the ones used can be $25,000 to $50,000 and higher, it is not cheap to create a piece of test gear that can be very accurate down to insane thresholds.
Let’s walk through the measurements of the Benchmark DAC3 ameasured by Stereophile (AudioScienceReview also measured it), link to measurements:
Figure 1: Testing the impluse and seeing what type of filter is used, normal people don’t really need to care about this.
Figure 2: Testing the filter for 44.1kHz (CD and most digital music). Ideally, there should be nothing after 1/2 the sampling frequency (talking the red line), so 22.05kHz. We can see that by ~24kHz, the red line is down to -120dB, so that’s good. The blue line is also showing aliasing and “images”, there are no spikes above -100dB, so that’s most likely not audible, which is a good thing.
Figure 3: Testing frequency response. We see it is super accurate for the Benchmark, almost no alteration.
Figure 4: Testing noise and whatnot. We see it’s so good it’s almost not on the graph (meaning even >$20,000 equipment can barely keep up).
Figure 5: Testing distortion. We see besides the 1kHz tone, and distortion spikes are super low.
Figure 6: Testing square wave. Should look similar to this , and both red and blue should be close in shape and level.
Figure 7: Should look like a sine wave.
Figure 8: Testing distortion for bass notes.
Figure 9: Testing for IMD (Intermodular distortion) in the treble, which is more audible. Again, besides the two main notes, any distortion is very low (disregard the rise all the way to the left).
Figure 10: Testing for jitter. The levels should be low and any distortion spikes (timing errors in regards to how in sync the 1’s and 0’s are) should not be higher than the green line.
Let’s walk through the measurements of the Benchmark DAC3 ameasured by Stereophile (AudioScienceReview also measured it), link to measurements:
Figure 1: Testing the impluse and seeing what type of filter is used, normal people don’t really need to care about this.
Figure 2: Testing the filter for 44.1kHz (CD and most digital music). Ideally, there should be nothing after 1/2 the sampling frequency (talking the red line), so 22.05kHz. We can see that by ~24kHz, the red line is down to -120dB, so that’s good. The blue line is also showing aliasing and “images”, there are no spikes above -100dB, so that’s most likely not audible, which is a good thing.
Figure 3: Testing frequency response. We see it is super accurate for the Benchmark, almost no alteration.
Figure 4: Testing noise and whatnot. We see it’s so good it’s almost not on the graph (meaning even >$20,000 equipment can barely keep up).
Figure 5: Testing distortion. We see besides the 1kHz tone, and distortion spikes are super low.
Figure 6: Testing square wave. Should look similar to this , and both red and blue should be close in shape and level.
Figure 7: Should look like a sine wave.
Figure 8: Testing distortion for bass notes.
Figure 9: Testing for IMD (Intermodular distortion) in the treble, which is more audible. Again, besides the two main notes, any distortion is very low (disregard the rise all the way to the left).
Figure 10: Testing for jitter. The levels should be low and any distortion spikes (timing errors in regards to how in sync the 1’s and 0’s are) should not be higher than the green line.