Burst Response! HD800, SR-207, HD650

Discussion in 'Measurement Techniques Discussion' started by purr1n, Jan 8, 2017.

  1. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    This is not a windowed tone, in the sense that there was a tone which was continuously playing, and a gated window was put upon an interval of that tone. This is an actual recording of a burst of X cycles at specific Y frequency of the same amplitude, capturing what happens at the beginning (sudden start) and end (sudden stop) of that burst. This has nothing to do with waterfall plots or impulse responses. We are looking at the raw waveform.
     
    Last edited: Mar 21, 2017
  2. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I think understand.

    My interpretation of what @JohnM is saying is that a rectangular gated toneburst will have an instantaneous turn on and off, which will have high frequency components. I believe the result would be a modulated step response, or in the periodic case, a modulated square wave response. Where modulation would be whatever the frequency response is of the step response or square wave, shifted in frequency by the tone inside the toneburst window.

    I'm not sure about the use of the Hilbert transform though. One can get an analytic signal that way, but I'm not sure about the use of it in this context.

    A Gaussian window may alleviate high frequency components arising from the abrupt windowing. But this would be at the expense of time resolution. It's just the way things work. A wavelet might be able to trade better frequency and time resolution better than straight Fourier given it's kernel. But it's been a while since I've used wavelets.

    EDIT: As far as the relationship with CSDs and impulse responses, the connection I see is that in both cases one is gating a signal by an specific window (impulse response for a CSD, or tone for a burst tone test). The selection of window might have an impact on the results in both cases.
     
    Last edited: Mar 21, 2017
  3. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Yes, but none of this has anything to do with what are looking at - the behaviors of the waveform (and how they deviate from the ideal) at the sudden start and sudden stop of the tone burst at a specific frequency.
     
  4. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Possible I'm confused, but the way I see it is that the behavior of the equipment due to a sudden start and stop of a tone burst is due to more than just the frequency of that tone burst. It's also a function of how sudden that tone was turn on, which itself injects other frequencies there. In other words, suppose the tone in the burst is 1 kHz. The behavior of the equipment, particularly at the transitions, will be not solely due to 1 kHz.
     
  5. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Consider this windowed tone at the measurement side of things:

    windowed.png

    The frequency response will have sidebes (one gets a modulated sinc for a rectangular gating window):

    windowed_frequency.png

    There is therefore always a tradeoff between time domain resolution and frequency domain resolution. Having a more "slow" time transition, say Gaussian window, reduces the side lobes. At the expense of time domain resolution.

    In a CSD, using a Blackman window likely does similar things. The window is not as abrupt in the time domain, loosing some resolution there. But in the frequency domain, resolution may increase.
     
  6. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    We are not looking at frequency response. We are looking at amplitude response over time. We are trying to correlate subjective dynamics and transient attack to certain measurments.

    There is a reason why I didn't throw the burst into the frequency domain.

    You are confusing readers. This isn't that hard. This has nothing to do with mathematical transforms into an FR, CSD, or burst response.
     
    Last edited: Mar 21, 2017
  7. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I understand. The only caveat I see is that it may be difficult to narrow down the results to a particular frequency. Specially at the transitions.

    Other than that, I'm sure different systems will yields different results. And it's very likely these results can be correlated with subjective experience.
     
  8. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I've only explored 10kHz and 100Hz. Would be interesting to see the behavior across a larger set of frequencies. As for correlation to "slam" or "limp-dick", there would appear to be some with the very limited data points we have.

    Another member who heard the Sony Z1R remarked that the "dynamics seem restrained". This was an assessment that I agreed with. Here are the 10kHz burst results compared to HD650:

    z1r burst.png

    Are we on to something? Maybe. We need more data. BTW, the 100Hz results don't seem to correlate as much. I surmise that this is because the honk of a trumpet, slam of a drum, crack of a whip, have high frequency components that contribute to a strong sense of attack / dynamics or lack of or opposite of "soft", "compressed", "restrained dynamics", "limp-dick" sound.
     
    Last edited: Mar 21, 2017
  9. Rockin_Zombie

    Rockin_Zombie Facebook Friend

    Joined:
    Jan 2, 2017
    Likes Received:
    117
    Trophy Points:
    43
    Location:
    Toronto
    Are these from the same tone? Amplitude is different.
     
  10. SSL

    SSL Friend

    Pyrate
    Joined:
    Nov 12, 2015
    Likes Received:
    1,239
    Trophy Points:
    93
    I would guess that is because with real instruments, it takes proportionally more time for a proportionally larger/more massive oscillator to get moving. If you look at the frequencies in say a piano note attack, the lower frequencies including the fundamental take a few cycles to get up to full amplitude. So the demands on the driver to accelerate would be less at lower frequencies.
     
  11. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Yup. I set amplitude to about the same, within a few db (around -10db) while recording although I wasn't exact. Differences might be to some level adjustment / normalization in wave editors at different times.
     
  12. JohnM

    JohnM Author of REW - Rando

    Joined:
    Dec 13, 2015
    Likes Received:
    134
    Trophy Points:
    33
    It's tempting to think that playing and stopping a tone at a given frequency and looking at the result tells you how the system behaves at that frequency. Problem is, you only see the behaviour at that particular frequency if the tone never ends (and for that matter, started infinitely long ago). As soon as the tone has a start and an end the frequency content of the signal being applied becomes more complex. It turns into the frequency content of the tone, just a line at the tone's frequency, multiplied by the frequency content of how the tone was modulated - i.e. in this case the frequency content of a pulse of the same length as the tone. As an example, this 10 cycles of 1 kHz:

    [​IMG]

    has this frequency content:

    [​IMG]

    The peaks at nulls at 100 Hz spacing are because the burst lasted 10 ms, the period of a 100 Hz wave. Because the content spans a wide frequency range, the output of the system when that signal is passed through it, i.e. the appearance of the 1k burst when captured, depends on the response of the system across that whole frequency range. There are various ways to find out what that appearance will be. It can be captured directly by playing the tone burst. Or it can be calculated, giving exactly the same result, by convolving the tone burst with the system's impulse response - the impulse response tells us everything about how the system will alter a signal passed through it (within the more or less linear range of the system at least). Equivalently, the frequency response above could be multiplied by the frequency response of the system and the result inverse FFT'd to produce exactly the same signal as convolving with the IR or measuring the output directly.

    All probably teaching grandma to suck eggs, but the point is all the info needed to know what the system will do is already available from the IR. Burst decay is another, and interesting, way to look at that behaviour, but using tones with hard starts and hard stops makes it more difficult to see how the device behaves at the particular frequency of interest because the spectrum of a hard truncated tone, as seen above, is pretty broad so the behaviour at other frequencies (particularly the lowest frequencies) has a significant influence. Using different ways of shaping the toneburst provides a test signal of more controlled bandwidth so the behaviour reflects that of the device over a narrower range of frequencies. ARTA uses a gaussian window for burst decay plots, another signal that might be of interest is REW's CEA-2010 burst. It is normally used for subwoofer power testing, but has a 1/3 octave span at the selected frequency.

    [​IMG]

    The idea of using the Hilbert transform to look at the captured response is indeed to form the analytic signal, the magnitude of the analytic signal is the envelope. Looking at the envelope makes it easier to see the shape of the response without getting distracted by the sinusoidal variation of the test signal.

    Incidentally, ARTA's user manual (http://www.artalabs.hr/download/arta-user-manual.pdf) has a discussion of the burst decay that compares and contrasts it with the CSD for those interested.
     
  13. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Amplitude shouldn't really matter unless we're dealing with super high levels where everything becomes non-linear.

    I think I agree with Gilberto. The problem is that I from the looks of it I think these measurements are still too influenced by FR/Phase/IR. At least that's what I think, or maybe I'm interpreting something into them that's actually not there.

    In a general sense it could also help to look at the efficiency (db/W - not sensitivity, db/V). I find the HD650 to sound compressed compared to the HD800. The STAX estats even more so. The FE83En was also a lot more compressed sounding than the HD800. The Voxativ widebander I feel is actually more dynamic/less compressed sounding than the HD800. I've not taken burst response measurements, but I bet they'd not look nice due to the whizzer cone. At least the phase response seemed good up to 10kHz.

    The HD800 driver is actually very efficient. I think I measured around 95-99db/W when used as a tweeter at 1m (but with the 300 Ohm impedance sensitvity is extremely low). The Voxativ driver seems to be even more efficient.


    I've not seen one measurement that I found to correlate with "transient response" as I hear it. Obviously transients in music are far from actual transients. Measurements also don't seem to capture things like a Be coloration - maybe we would have to extend the FR past 20kHz to look at how break-up is controlled? I'm generally not too convinced stiff cones are really what we want. There often seems to be some hardness to the sound, even if resonances shouldn't be audible.
     
    Last edited: Mar 21, 2017
  14. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    LOL, what does IR (impulse response) have to do with any of this? Or even FR or phase? If this was the case, we would see wow and flutter in the sine plots I've posted. Maybe there is some very minor amount if we do some deep mathematical analysis, but this isn't the point! At least for now in the interests of keeping it simple.

    What Gilberto was talking about is when we put into a window, intervals of signals, and mathematically transform them from the time domain to the frequency domain, we get artifacts, such as in the case of rectangular windows, a opposed to windows were the edges for certain rise and fall times are based on a more complex function.

    This is a nonsensical argument because we are looking at the raw waveforms. We are not looking to visualize things in a frequency response, distortion, or CSD format. There are no fourier transforms on an impulse response (which in turn is derived from a sine sweep or pattern noise) involved where the window shapes would affect the visualization.
     
    Last edited: Mar 21, 2017
  15. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    @Marvey, the artifacts will be present in the time domain toneburst results. It's not a by-product of applying the FFT.

    @JohnM, I will look into it, but I think the analytical signal makes things complex (removes the negative frequencies). It is used to simplify models of things. Also if you remove the "carrier", the envelope is just a pulse or a square wave.
     
  16. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I need to understand.
    1. What is the exact the nature of the artifacts? How would they manifest in the recorded waveforms I've shown above?
    2. What are they caused by?
    3. Do they matter enough for this exercise where we are attempting to correlate start / stop behaviors to perception of certain sonic traits?
     
  17. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    The starting and stopping is more of a square wave than a sine wave. Not just one frequency. ADCs essentially also window the results.
    I do think the weird shape of the SR207 plots are because of its FR. Same for my HD600 and HD800 results. Z1R has treble rolloff so it'll look slower. Maybe we can compare against step response plots and see if they correlate.

    On the other hand it is weird that my SR202 measurements looked like they had a much slower rise-time than the HD800, despite much better treble extension (I measured with higher than 22kHz BW). The step response plots had a much faster rise-time than the sinus step. It's also weird that the rise-time was so slow for the 100Hz tone for the SR207, but that was likely the resonance. HD650 also has a resonance there and also had a slower rise-time than the HD800.
     
    Last edited: Mar 21, 2017
  18. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    The starting and stopping are of an actual sine wave. I played back and recorded a 10kHz sine wave, not a 10kHz square wave.
     
  19. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    1. & 2. The artifacts are due to the measurement gating and/or starting and stopping of the sinusoidal.
    3. It will mater when trying to assign differences in perception due to a particular tone. The perceived differences will not be due to a particular tone alone.
     
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    But I am not gating anything? How am I gating an actual recording?

    I am pressing play on a .wav file, and then recording the result through an ADC from a microphone into Adobe Audition.
     

Share This Page