Burst Response! HD800, SR-207, HD650

Think about it with DC or low frequency. As soon as you start DC, you will get a bunch of frequencies due to the transition from off to on, and from on to off. The same happens if you are not looking at DC. Just the harmonics are now shifted in frequency.

 

Got it, but aren't we overthinking the issue?

We are simply trying to compare original and recorded waveforms, or compare recorded waveforms of different headphones to each other, not analyze their spectral components.

Also, won't these extraneous components be seen in the recordings? The waveforms?

 

Now that's a legit question. Do we even need to bother with these if step response does the job.

The answer is that we need a bit more data.

 

One will see such components in recordings. The only deal is that one may not be able to say all those issues are due to single frequency, but a set of them. One could still proly associate things to subjective observations.

 

I don't see a correlation to frequency response. The hd650 and hd800 (very different FRs) sine wave recordings look more similar to each other than the others. Both of these headphones folks might consider less dynamically compressed than the others as well.

 

But this is moot no? Since such components are there in both the original wave and recorded wave. And the fact remains that the initial levels of the crests of the sine waves differ, in some cases substantially, with each headphone. This is something we can actually see.

 

It depends what one wants to do.

If one wants to isolate characterization to a single frequency, and compare to the original recording by looking at: response delay, amplitude and such, then an abrupt start/stop signal needs to be appropriately windowed to avoid other frequencies from contaminating the results.

If one is more interested in the response of a system due to an abrupt change (frequency "contamination" or not), a step response should do well. A modulated step response (i.e. toneburst) should also provide some light IMO.

In fact, in simple control and PLL design, we look at step response for clues about what's going on.

 

What if we started with a low level sine and then abruptly raised the level far away from the edge?

 

Could try a Gaussian envelope or even a Blackman and check results. Also may want to experiment with two tones if it doesn't get too hairy.

 

LOL, I'll send you the data and you can do this for me! Not good with math packages - I tend to program everything from scratch.

 

OK, I pulled up some step responses to compare with the 100Hz bursts.

IMPORTANT NOTE:

1. The ETCs (left) are derived from impulse responses, which in turn are derived from sweeps.
2. The bursts (right) are actual recordings from a microphone into an ADC into Adobe Audition. Sampling rate was 192kHz.

The ETCs follow a similar pattern to the bursts, but the issue, which I should have thought of earlier, is that the ETCs do not give us a idea of what the steady state level of the a sine is supposed to be. Remember, we are trying to ascertain why some headphones seem to hit harder or have stronger dynamics than others. Therefore, knowing the steady state level is necessary so we can see how the signal approaches the steady-state.

In order of "dynamics" or lack of compression, I would subjectively rate the headphones as follows:

#1 Atticus - hits damn hard
#2 HD650 / HD800 (tie) - hits hard
#3 Z1R - reserved dynamically
#4 SR-207 - limp dick




Couple of interesting things. The headphones which I felt exhibited the least dynamics or most compression were the SR-207 and Z1R. Note that the sine waves of these two headphones take their time to reach steady state.

The HD650 and HD800 also have a blunted initial crests, but not to the extent of the SR-207. However, in contrast to the Z1R and SR-207, the second crest actually overshoots slightly, although one can argue that part of this effect with the HD800 could be attributed to a tiny DC shift - see (3)

Personally, I think we are on to something and that any concerns of signal contamination may be overstated.

 

In this case, the signal wouldn't be changing suddenly. The signal still rises at a 10kHz sine, which isn't that fast in the overall scheme of things. The difference is that we are not coming from nothing - no signal to a signal, which is kind of a divided by infinity sort of thing. And even if we did, as we can see in the recordings, it's not gonna f**k up the first sine wave. The recorded sine pattern burst is exactly the same as the original wav file, with the exception that the amplitudes are different.

Read the first line of the post. I'm not stuck at 10kHz and would like to explore other sets of frequencies.

 

Yes. The experiment is see to how the signal reaches the steady state. Not just how long, but how, and what behaviors are exhibited while it does so. Any attempts to reframe it as otherwise is excess mental maturation on the side and severely obfuscates what I've been trying to point out: correlation between subjective dynamics and over/underdamped wave crests on the way to a steady state.

Sure there are other frequency components, but that's not what we are looking at here. And if these components were so bad to the extent that they would be relevant, we are not going to get anywhere near the same cyclic sinosoidal pattern of the original. We'd see weird ass squared waves and spikey shit.

What it comes down to if this is shit. If it's shit and has less than 60% correlation to what people hear as "dynamics", then I won't bother. I'll lock this thread and cease down this road of exploration.