Headphone CSD waterfall plots

[PelPix]

Sorry if I'm a bit slow.  I haven't really seen a waterfall plot before.  It seems like a really interesting way to map the data.
So essentially, what we're seeing is driver response with frequency and volume?
So the lines that extend past the rest of the wall (further than so many MS) are ringing resonances?
 
To put it extremely simply, the axis coming towards us is sort of like the impulse response, the axis going from left to right is by frequency, and the axis going up and down is the volume?

 

[purrin]

Oops. The last few sets (in blue lines) are not CSD waterfall plots, but non-linear distortion graphs. Basically I excite a few frequencies (run one to test tones). Ideally, on the RTAs, we should only see those two test tones. Anything else is non-linear distortion with harmonic distortion being an integer multiple of the signals and intermodulation distortion an integer multiple of the difference between two test tones.

 

[purrin]

 
Quote:

Sorry if I'm a bit slow.  I haven't really seen a waterfall plot before.  It seems like a really interesting way to map the data.
So essentially, what we're seeing is driver response with frequency and volume?
So the lines that extend past the rest of the wall (further than so many MS) are ringing resonances?
 

 
Driver response in terms of frequency, volume, and time (the axis coming toward us). Being able to see what happens over time time is more awesome than simply an FR.
 
The only caveat here is that these plots are not optimized for absolute correct FR. The measurement methods I employed attempt to remove the effect of the enclosures somewhat (thus less bass measured for closed phones.) I was more interested in seeing the speed of decay and effects of ringing on the driver than an accurate FR. Relative FR from one headphone to another (assuming that I measured it) can be inferred only if the enclosures are similar.

 

[arnaud]

Hi Purrin,
 
Congratulations again for bringing concepts from loudspeaker design world to headphone testing.
 
Quote:

Oops. The last few sets (in blue lines) are not CSD waterfall plots, but non-linear distortion graphs. Basically I excite a few frequencies (run one to test tones). Ideally, on the RTAs, we should only see those two test tones. Anything else is non-linear distortion with harmonic distortion being an integer multiple of the signals and intermodulation distortion an integer multiple of the difference between two test tones.

I don't remember IMD being a key design parameter when selecting speaker drivers, is it always advertised as part of standard performance sheet? 
 
 
What I don't understand is the meaning of such test for a transducer. For instance, I can see the meaning while looking at the behavior of electronic components. For example, hard clipping from an amplification stage should show up pretty well in the harmonic distortion graph . Or, for low level signals I suppose some issues can occur when crossing the 0 line (when different components take care of the + and - components in the signal. DISCLAIMER: I am about as knowledgeable in electronics as an average turtle and it probably shows up in my post ).
 
 
On the other hand, for a speaker, you don't expect it to generate significant amounts of distortion when driven at a frequency due to the nature of the device:

1. It really only vibrates AND radiates at the excitation frequency
2. It does not suffer from hard clipping like electronics (you simply get some compression effect)

 
Any information you can share about the origins of these distortions for a transducer?
 
Another point of concern in your graphs is how much of the distortion is coming from the apparatus (amplification ...). Maybe you could add a reference plot with the output of clio (or better, that of the amp passed through an attenuator) fed into its input?
 
Quote:

Driver response in terms of frequency, volume, and time (the axis coming toward us). Being able to see what happens over time time is more awesome than simply an FR.
 
The only caveat here is that these plots are not optimized for absolute correct FR. The measurement methods I employed attempt to remove the effect of the enclosures somewhat (thus less bass measured for closed phones.) I was more interested in seeing the speed of decay and effects of ringing on the driver than an accurate FR. Relative FR from one headphone to another (assuming that I measured it) can be inferred only if the enclosures are similar.

Marv, I reckon you started doing these tests on regular dummy head data and then found out you got cleaner results with your current anechoic termination test rig.
 
While I thought this was a great idea at first, I am not sure of the meaning after thinking a bit about it:

• There does not seem to be a fundamental issue with using a standard dummy head (see my CSD post-processing of Tyll's data)
• Headphones are (for the most part, grant you the K-1000 is a different beast) designed to feel an acoustic load forward of the driver (e.g. the acoustic chamber between the driver the ear due to earpad spacing).

 
I see several major issues with the anechoic termination test:

• As you pointed out, the low frequency part of your CSD plot has little to do with actual performance. (that's regardless of open or closed design, being in the acoustic near field of the driver, e.g. it's almost like doing an IEM test without a proper seal)
• The anechoic termination completely changes the acoustic resonances of the enclosure (from a sort of rigid boundary condition to anechoic one, acoustic resonances will shift down in frequency VERY significantly, by a factor of 2 for example). This will be seen up to very high frequency because such acoustic resonances occur above 2kHz.
• The anechoic termination adds significant acoustic absorption in the front chamber (ear to driver, bounded by the earpad) so the acoustic resonance of this region will not only be shifted in frequencies but also significantly damped so you pretty much won't see them in the CSD plot.

 
Have you thought about going back to standard dummy head measurement now that you got more experience with CSD testing? I understand you might want to isolated the driver resonances from the acoustic chamber but in this case, extracting the driver (like you did with the SR80?) or better yet, doing a laser vibrometer measurement on the diaphragm (can you get your hands on that ? ) would be suitable.
 
cheers,
arnaud

 

[purrin]

 
Quote:

I see several major issues with the anechoic termination test:

• As you pointed out, the low frequency part of your CSD plot has little to do with actual performance. (that's regardless of open or closed design, being in the acoustic near field of the driver, e.g. it's almost like doing an IEM test without a proper seal)
• The anechoic termination completely changes the acoustic resonances of the enclosure (from a sort of rigid boundary condition to anechoic one, acoustic resonances will shift down in frequency VERY significantly, by a factor of 2 for example). This will be seen up to very high frequency because such acoustic resonances occur above 2kHz.
• The anechoic termination adds significant acoustic absorption in the front chamber (ear to driver, bounded by the earpad) so the acoustic resonance of this region will not only be shifted in frequencies but also significantly damped so you pretty much won't see them in the CSD plot.

 
Have you thought about going back to standard dummy head measurement now that you got more experience with CSD testing? I understand you might want to isolated the driver resonances from the acoustic chamber but in this case, extracting the driver (like you did with the SR80?) or better yet, doing a laser vibrometer measurement on the diaphragm (can you get your hands on that ? ) would be suitable.

Definitely, especially for closed phones where a good seal is rather expected from the design. I actually started to gather data from "dummy head" tests, most which I have not published because I'm tweaking a few things. I'll send you some data next week or put up a web resource where you can just grab stuff and analyze. The challenge for the dummy head will be to find a suitable surface that simulates human skin - both its reflective and absorbative effects. I guess I can use own my head, but MLSSA impulses at 90db get tiring.
 
One effect or concern from the dummy head / sealed type tests is that they tend to accentuate peaks and valleys in the treble, and even create their own. You see this in Tyll's data or in my recent Darth Beyer plot. The large range between maximums and the minimums tend to bury the lower level data in with the noise.
 
The second issue or more important question is whether we actually hear these nodes? To what extent do our ears and brains filter the high frequency nodes within the enclosure and our skin? I think quite a lot. For example Tyll's HD800 FR shows significant valleys at 9k, 11k and peaks at 7k, 10k, and 12k. When listening to sine sweeps, I don't hear this behavior - my experience is much closer to that of t=0 on my own "anechoic" CSDs. 
 
I think where I'm getting at is "keep it simple", have comparable relative data, and let each person (and headphone) sort it out. I'm not definitely not an absolutist, because then it becomes like certain religions. The anechoic tests, at least the in way I've implemented them, still consider some effects of the enclosure, which wasn't what I wanted initially. But maybe that isn't a bad thing.
 
Point #2 above is a good one and warrants additional investigation and data collection using both methods. I haven't seen any shifting of enclosure frequencies yet (I may need to look in the right area), only increase in volume and decay time. I am not totally certain, but I believe most of the effect of the enclosure is below 1.5-2k after 0.5ms. And even if there are resonances, they are likely to be very broad and shallow, it other words, it's not going to be something that stabs your ear. Which was one of my original goals, find those bad spots.

 

[purrin]

Which reminds me, I should go back and do a "dummy head" type test on the HD800. I'm still curious whether the 5-6k ringing is attributable to the driver, enclosure or both. I won't be able to get back to this stuff until at least a week later.

 

[purrin]

Fischer Audio FA-003

By request. Plots are not totally reflective of FR, especially toward the bass and the trough at 8k, but driver ringing is identified. Not too bad really.
 

 

[arnaud]

Marv, looking forward to your additional data. I agree on sharp peaks being much more obtrusive than narrow deeps. The skin impedance is an issue with dummy heads, but indeed specialized heads should have that in consideration. The pinna reflections messing up the interpretation of results toward higher frequencies is indeed a concern if you're just after the driver / enclosure dynamics but I assume equalized measurements would deal with this (Tyll does that I believe? You need a calibrated dummy head for this, like a diffuse field HRTF taken with the measuring device).
 
Also, unless it's a joke I did not catch, the word is "anechoic" (without echo) not anecolic (which sounds like without diarrhea in French ).
 
Keep the good stuff coming!

 

[purrin]

LOL, I knew I spelled it wrong, but couldn't find a spell checker with that word in it.
 
One thing to think about (and this is for everyone!) for the narrow troughs on the typical FR graphs: they are almost areas with stored energy problems. They are a trough for the first fractional millisecond, then they quickly become ridges.

 

[purrin]

Quote:

Originally Posted by arnaud /img/forum/go_quote.gif
 
What I don't understand is the meaning of such test for a transducer...
 
On the other hand, for a speaker, you don't expect it to generate significant amounts of distortion when driven at a frequency due to the nature of the device:

1. It really only vibrates AND radiates at the excitation frequency
2. It does not suffer from hard clipping like electronics (you simply get some compression effect)

 
I did these tests because someone asked and I was curious myself. But anyways, these tests are important for measuring the performance of speaker drivers to determine their optimal ranges (crossover placement). I was actually surprised (not really) these headphones measured so well. Speaker drivers measure worse, but that is probably because of their physical size and power requirements which are a magnitude greater or more.
 
Quote:

Originally Posted by arnaud /img/forum/go_quote.gif
 
Another point of concern in your graphs is how much of the distortion is coming from the apparatus (amplification ...). Maybe you could add a reference plot with the output of clio (or better, that of the amp passed through an attenuator) fed into its input?
 

 
I should do a loopback. But then again I may not take the time because, I'm not too interested in these graphs either. These headphones appear to measure very well, especially since the drivers have to handle the full range. The results are more or less close to each other. Although I have to say that I'm more worried about the noise floor (environmental noise) than the distortion coming from the amp!
 
As for IMD, I don't ever see these on the driver spec sheets either. But it is one of those interesting things to note. IMD is easily more hearable than HD, after all, it's not an integer multiple of any of the fundamentals.

 

[Armaegis]

I would like to see how the effects of different cup damping materials can affect these plots. Tyll's got measurements from various ortho guys; I wonder if any of them would like you to measure their cans as well. 

 

[arnaud]

Quote:

I would like to see how the effects of different cup damping materials can affect these plots. Tyll's got measurements from various ortho guys; I wonder if any of them would like you to measure their cans as well. 

I am trying to investigate this through simulation:

1. I went back to the HD800 model I had made for simulation of the effect of placing additional absorbing material in the earcup ( see link above in this thread ).
2. The comparison below is using Purrin's test data (not a standard dummy head recording) and comparing against the model.
3. "Rigid plate" means that the pinna surface is simply modeled as a rigid flat surface.
4. "Anechoic plate" means that the surface representing the pinna has an absorption equivalent to a thick piece of melamine foam (similar to the test setup).

 
Current observations:

1. Simulation and tests are not perfectly lining up, I plan on revisiting the model later on.
2. As Purrin mentioned and contrary to me previous expectation - we don't see much impact at low frequency.
3. Acoustic resonance predicted at ~2kHz might be an artifact of the model
4. Acoustic resonance at 5.5kHz is well predicted
5. The anechoic termination near the microphone has a large impact on the resonances and add a lot of artificial damping

 
Actual results (SPL comparison and predicted SPL contour plot at 5.4kHz):

 

[purrin]

Let me throw in a rigid plate test when I get back home toward the end of this week. This will be interesting to see!

I'll send you the impulse responses. BTW my HD800s have 2mm on foam stuck on the black plastic portion and on the metal ring around the driver. You may want to model that in. The sticky on the back of the foam is staying on pretty good so I don't want to take the trouble to remove it.

 

[arnaud]

Excellent Marv, it will really help to increase the fidelity of the model as there are so many simplifications now (fine mesh grill performance, driver dynamics, ear pad...). I probably don't expect accuracy within 1dB given the number of assumptions / guess I have to make on the various part properties, but probably can improve the results a bit. It should be not problem to add the 2mm foam, is it melamine foam (like the sponge type I recommended for the mod)?
 
Quote:

Let me throw in a rigid plate test when I get back home toward the end of this week. This will be interesting to see!

I'll send you the impulse responses. BTW my HD800s have 2mm on foam stuck on the black plastic portion and on the metal ring around the driver. You may want to model that in. The sticky on the back of the foam is staying on pretty good so I don't want to take the trouble to remove it.

 

 

[purrin]

I didn't use the melamine foam. Instead I used the adhesive backed 2mm dense foam sheets that kiddies use in school to make whimsical crafts. You may have the impulse response file using the "Anechoic Plate". It's labeled "HD800 Arnaud foam mod" for lack of a better term.