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:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Thx. That should save some time.
     
  2. philipmorgan

    philipmorgan Member of the month

    Pyrate BWC
    Joined:
    Sep 28, 2015
    Likes Received:
    3,790
    Trophy Points:
    93
    Location:
    In the wind, so to speak
    Home Page:
    How do you set up that measurement and generate that plot, John? I'm a bit of a noob with REW so feel free to point me to the right part of the docs if that's easier.
     
  3. JohnM

    JohnM Author of REW - Rando

    Joined:
    Dec 13, 2015
    Likes Received:
    134
    Trophy Points:
    33
    The measurement of the LCD-2r2 was taken using a miniDSP EARS rig. I generated a 3 cycle 100Hz sine as a WAV file, then used File -> Import -> Import Audio Data to load it into REW. With the measurement and the audio file loaded I used the trace arithmetic A*B function on the All SPL graph to generate the product of the responses, which is equivalent to a convolution. The resulting time domain response can be viewed on the Impulse graph and responses overlaid on the Impulse graph of the Overlays window. That's about all there is to it, other than checking the window selection on the impulse plot for the imported tone burst to make sure the window isn't curtailing it, since the tone burst starts and ends at zero can just choose rectangular windows (using the IR Windows button). I should probably add a tone burst option to REW's signal generator, I'll put that on my todo list.
     
  4. philipmorgan

    philipmorgan Member of the month

    Pyrate BWC
    Joined:
    Sep 28, 2015
    Likes Received:
    3,790
    Trophy Points:
    93
    Location:
    In the wind, so to speak
    Home Page:
    The latest build of REW dropped today and has a tone burst feature in the signal generator now.
    Screen Shot 2018-07-15 at 6.54.54 AM.png
     
  5. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Latest experiments. Needs a lot of cleaning up. This is from actually recorded wave files. Verite wins overall in this three-way contest. Elex is a close second.

    Verite Burst.png
    Elex Burst.png
    Ether burst.png
     
  6. MF_Kitten

    MF_Kitten Banned per own request

    Banned
    Joined:
    Jan 1, 2016
    Likes Received:
    181
    Trophy Points:
    43
    Since you're already here... could REW get a stereo mic oriented mode for things like the EARS, so I can load the calibration files for the left and right mics, and not have to change them each time, and maybe even a feature that does the left and right side in sequence when checked off, instead of having to do it twice?

    Finally, when exporting measurements and EQ filters as IR, I would love a feature that cuts all the silence before and after the impulse before exporting, so you can load it as a convolution reverb or impulse filter without having to mess around with it first.

    Also, with headphones being a closed environment with super minimal micro/nano-acoustics compared to a room, what are the best settings for measuring them? Things like the IR window and stuff I mean. Hell, I have no idea what to do with that stuff when measuring my room! :p

    Edit: here are a couplethings I have wished I could do as well:
    -export calibration file as .wav IR
    -Apply EQ filter to measurement IR before export
    -"time warp" through the CSD to correct for measyrement artifacts. Many headphones have a deep dip in the measured FR that is actually a peak being disguised. In the CSD you'll see it go down and then right back up a couple slices down the line. What if you could tell REW to look at slice X for a certain frequency range, like "look at this slice for 3.2KHz-4.1KHz", with the chosen slice being the one that has the real level.
     
    Last edited: Nov 2, 2018
  7. MF_Kitten

    MF_Kitten Banned per own request

    Banned
    Joined:
    Jan 1, 2016
    Likes Received:
    181
    Trophy Points:
    43
    I can only guess which is which, but I love how you can see the "speed" so clearly. A slower driver doesn't get that first cycle out quite fast enough, and can't stop it quite fast enough. Lower first cycle, higher last cycle.
     
  8. JohnM

    JohnM Author of REW - Rando

    Joined:
    Dec 13, 2015
    Likes Received:
    134
    Trophy Points:
    33
    That's on the todo list.

    Choose the option to export the windowed IR and set the left window start where you want the export to start and the right window end where it should end.

    Defaults should be fine, but I'd suggest that given where this discussion is taking place there are others who could bring greater experience to bear in commenting on that.

    That would be quite a faff. The response would need to be extrapolated to cover DC to half sample rate and corresponding phase info generated. Bit too much guesswork required for my liking.
    That's doable, REW has an IR for the equalised measurement anyway. What would you use it for?
    Not really following you there.
     
  9. MF_Kitten

    MF_Kitten Banned per own request

    Banned
    Joined:
    Jan 1, 2016
    Likes Received:
    181
    Trophy Points:
    43
    Apply EQ filter to measurement would be for stuff like capturing a binaural impulse response of speakers in a specific room, then EQing them to my spec, and then exporting that as an IR that can be used in a DAW to simulate said speakers when using headphones. I can load up the calibration files as target responses and EQ based on those, which is something I do already. But being able to export it that way would be awesome.

    edit: Oh, and the CSD thing is so hard to explain. Just leave it for now, as I don't know enough about it to suggest a viable solution :p
     
  10. zerodeefex

    zerodeefex SBAF's Imelda Marcos

    Staff Member Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    14,051
    Trophy Points:
    113
    Wtf, I need to clean up this thread. This isn't the "REW suggestions thread from someone who shouldn't be using REW."
     
  11. EagleWings

    EagleWings Friend

    Pyrate Contributor
    Joined:
    Jul 20, 2017
    Likes Received:
    1,709
    Trophy Points:
    93
    Location:
    India
    Bringing over the burst response conversation from the SR1A thread.
    This dual tone burst was a revelation to me. I have always wondered how a driver vibrates, when more than one instrument is playing simultaneously and this finally explained. Is it possible to take a piece of actual music track that has multiple instruments playing and measure the burst response?
     
  12. gixxerwimp

    gixxerwimp Professional tricycle rider

    Pyrate
    Joined:
    Sep 30, 2015
    Likes Received:
    5,773
    Trophy Points:
    113
    Location:
    small island claimed by China
    If you open an audio file in Audacity, you'll see the waveform in all it's complexity. Hopefully (ideally?), the driver vibrates correspondingly.

    This is the first note of Get Lucky with the bass and cymbal hit superimposed (hmm, it looks a touch clipped).

    Get Lucky.jpg
     
  13. dietwater

    dietwater Acquaintance

    Joined:
    Jan 17, 2020
    Likes Received:
    66
    Trophy Points:
    18
    Location:
    India
    The behavior of sr207 is like how a capacitor would behave to sine wave input. Not super surprising since electrostats are capacitive loads and even more so in the sr lineup since they are known to have fairly high capacitance. I was able to simulate the texture of L300 and SR009 by just taking a low pass (square wave to sine wave circuit impulse response) and convolving over the input sound.

    I'm really curious what happened to this thread. Happened to take really nice measurements but then got into the same issue that other audio measurements get into. Trying to fit everything into it and trying to get an explanation. Burst is a decent way to check if a headphone suffers from poor attack, not sure if it can be used for other purposes.

    I also read about some of the arguments from REW representative. I understand that doing fft at the edges is going to have consequences, but fft or any other spectral analysis technique mentioned was in the first place meant not really to check the edges of the signals. Whatever explanation given doesn't explain why SR207 alone showed such slow response in initial stages of the burst while the other headphones didn't show such a problem. In short, the post was irrelevant to what was being analysed.
     
    Last edited: Jan 20, 2020
  14. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Still taking these measurements. I convolved a set from IR for a while just to see, but now I'm recording them directly from a capsule mic.

    See page 8 of the SR1a for its burst response at various frequencies.
    https://www.superbestaudiofriends.o...oly-moly-buy-this-now.8041/page-8#post-266742

    They good for assessing decay and energy release too.

    Next steps are to write a visualization program lik how I did with a CSD.
     
  15. dietwater

    dietwater Acquaintance

    Joined:
    Jan 17, 2020
    Likes Received:
    66
    Trophy Points:
    18
    Location:
    India
    I would like to know if I could be of any assistance to you in terms of these simulations. I don't have any measuring gear but I can work my way through coding, given enough time.
     
  16. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Some stuff to ponder on...

    B1696 HD800.wav.jpg

    B1696 Grado RS1.wav.jpg

    B1696 SR1a.wav.jpg
     
  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
    How many dBs per division is that? Also, I forgot to ask yesterday, but are they measured on the V1 or V2 type flat plate coupler? I'm assuming V1. Would V2 look better in the bass? I tend to get more bass extension with my V2 type coupler.

    I find it interesting how little high frequency resonances there are with the 50 and 120Hz bursts. The "stopping" of the signal is essentially a step (so not bandlimited), but maybe those are hidden within the (louder) fundamental decay. The slope is less steep on the LF bursts so there should be less higher frequencies.
     
  18. Boops

    Boops Friend

    Pyrate
    Joined:
    Nov 8, 2015
    Likes Received:
    3,179
    Trophy Points:
    93
    Location:
    New York
    A few other questions/thoughts:

    1. are we just seeing the top half of the burst “wave” in these or did you change the visualization to keep the graph above the X axis?

    2. I like the use of fading/visibility to indicate lower dB. Same question as @Serious: what dB are we talking about? Would it be useful to label the Y axis with values?

    3. Did you consider using the same color coding here as you do on the CSDs where red = higher dB down through blue = lower? Not sure how the actual dBs compare across the two graphs, but it might aid understanding to have data represented in a visually consistent way since both of these are charting dB over time.
     
  19. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    10,961
    Trophy Points:
    113
    Location:
    Philippines, The
    Since we're coming on with questions, I'd like to add that while it's less visually appealing I'm kinda curious to see the actual 10 cycles too for this graphic, since I remember some headphones exhibited a trouble getting fully into the swing of things that was vaguely indicative of performance too (at least for the Sony flagship, based on memory).
     
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    The entire range is approximately 60db+ some depending strength of the recorded signal in the wave file which contains all the bursts. Ideally, I could write an interactive app that played each sine wave, adjusted it to a standard level, and then recorded again. But then again, is this necessary and could this be overthinking the problem? The flat plate coupler is decently representative of the FR (except between 2-6kHz where the situation will vary depending upon headphone, and arguable "better" past 6kHz).

    In recording the 16-bit wave file, I will typically allow 10-13db headroom for most signals. Some treble peakage if bad might only leave a few db. I take the results and normalize them to the top line on each graph, covert to log scale, and then I lop off the bottom quarter (for visualization reasons). I could however just standard the db scale range for each plot or label it. I could decide not to normalize the sine waves, or at least normalize everything to the 500Hz signal, which may be a good idea. Or maybe I can standardize on a db range while still maxing the peaks to the top line. But really, since this is sine waves, albeit applied with log, maybe doesn't matter. Hard to decide, it's give and take.

    V2. I felt V1 would be too problematic. I would have to EQ, and then the artificial ear would cause signals to arrive at the mic at different times. It's not that this would not be interesting, but too many individual differences. Figured we should take that factor away as much as possible.


    Still full wave, but just applied absolute value. I thought about doing energy time curves, but when this is done, some information is lost. Is interesting to see how some headphones maintain the frequency of the burst, and how others do not. I thought about CSD rainbow colors and decided against it. This is just a 2D area under the curve plot. I could have just kept the color solid.
     
    Last edited: Apr 12, 2020

Share This Page