Discussion in 'Measurement Setups, Systems, and Standards' started by Luckbad, Oct 5, 2015.
Impulse response. Does that looks good and informative presented as such for you guys ? :
Looks quite good. When I look at the impulse response I usually look at -0.5ms to 3ms (slightly longer range), though -0.5ms to 2ms makes a lot of sense for the better headphones out there. I also measured the loopback (DAC->amp -> ADC) and the impulse response changes a lot when changing windows audio settings (my output is always at 44.1kHz, I only changed the input sample rate). I prefer the CSDs over the impulse response.
What window rise time do you use for your CSDs, @sorrodje? It looked a bit high in the STAX thread. I use 0.25ms.
For comparision, here is my impulse response for HD600 and HD800 (modded) with the coupler I now use:
HD800 (modded) on creatology foam + 3mm felt disc coupler:
HOLY SHIT, I just realized that I still haven't posted a pic of my coupler. Sorry. I should've done this a long time ago. I just assumed that I already did.
I use a Panasonic WM61-A. I actually bought 10 and measured them all to get "matched pairs". They were all pretty close, considering how cheap they were. I actually used a DBX RTA-M before, which wasn't better. I use this as a phantom power supply and plug it straight into my ASUS Xonar STX soundcard.
I calibrated the volume level by using the specced mic sensitivity (15mV/Pa) and line in sensitivity (2Vrms) in ARTA. (Remember that my setup has no gain between the mic and the input) I don't think this calibration is very accurate at all, but I simply do not have the means to calibrate it more accurately.
My REW settings are:
Spectrogram and CSD impulse response window:
Hamming Left, Blackmann-Harris 4-term right
120db top/49-35db bottom for measurements taken at 85db, this is a 36-55db range, depending on measurement accuracy, measured transducer, etc.
I mostly use either a 36db or 40db range
X 15; Y 340; Z 230
5.21ms time range (so every division is 1ms)
4-7ms window, depending on where the reflections start. I try to place my coupler >0.7m away from reflective surfaces 0.7m is about 4ms.
0.25ms rise time
44.1kHz log sweep from DC to 22 050Hz, -3db
128k sweep for CSDs, 1M sweep for distortion; Idk why but the longer sweeps change the phase (and accordingly: Impulse response/CSD)
I already said that I don't like the creatology foam as a coupling material: I feel that it gives too much of a seal (Marv used a piece of felt to deliberately break the seal) and is too reflective somehow. When I used a V2-type coupler I used a thicker piece of felt (I think it is 3mm), roughly the shape of my ear, to absorb reflections. I still didn't like the results this coupler gave me, at least for the HD800. Most headphones will measure better on the V2 type coupler than the HD800. This is why I then decided to build a coupler closer to V1. I ended up using the following material (about 3cm thick):
(HD600 driver for scale )
(more close up pic, so you can better see the structure of the material - looks to be open cell foam)
I think this is essentially multiple dense open cell foams squashed together. Maybe this is part of the secret (that it technically isn't one material, but multiple materials in one). That I got it for free was simply a bonus.
This is mainly supposed to be a more easy to use Marv-V1-type coupler, without having to use a solid plate for bass measurements. With my open headphones, HD600 and especially HD800 the bass measurements actually more closely match what I hear and are very close to the bass measurements my mic-in-the-ear-canal measurement method gives me.
The catch is that this material will probably not seal with the STAX headphones, so the bass measurements will probably be inaccurate for stats. I can't say for sure as I've never had the chance to measure one.
You could also make the argument that this material is not reflective enough and doesn't show resonances between the side of the head and the earcup properly and to a degree this is correct. I just find this to be the better compromise (between too dead and too reflective).
I know that I forgot something, I just can't think of what it was. Will update this post when I remember again.
EDIT: Updated my REW settings
EDIT 2: Also attached CSDs in sexy white (36db range). I wish REW had a rainbow color option. Note how the HD800 mods get rid of the resonance even on the v2 coupler. Also the open cell foam coupler accentuates the 5kHz ridge on the HD600. These plots are more indicative of what I hear (and the mic in my ear results look similar).
EDIT 3: Now I remember what I forgot: My mic config. Updated.
My whole measurement rig actually cost less than 20€ for 10 mics + phantom supply + wire + coupler. I think everyone should give it a try. Soldering the WM61-A mic capsules also really isn't that hard.
@Serious here are my setttings for waterfall plots
Looks quite good, my Y axis angle is steeper, but otherwise very similar (except that I use a 5.21ms time range). I quite like your settings.
Have you tried changing the impulse response windowing function in REW (under Preferences, in a tab further to the right)?
Considering the above IR plots are satisfying for me I didn"t play that much with other settings
No, they affect the looks of the CSD and Spectrogram plots. Play around with it for a bit, you'll see what I mean. I posted my settings above.
Ah OK. will check that . thks
This post is a copy of my post in the Kennerton ODIN thread. This discussion belongs here and would needlessly derail the other thread.
Yes! This is something that I've been wondering about for a long time: How do other people calibrate their volume levels?
How I did it (probably very inaccurate):
You need ARTA, otherwise it will be more difficult. Under 'Audio Devices Setup' enter values for the Line In and Line Out sensitivity. The values here are mVpeak and not the specced Vrms, so beware of that! The most important thing however is the microphone sensitivity. First, use the specced sensitivity for your microphone capsule. See that you can get a mic preamp with unity gain, otherwise you will have to find out the gain value. The WM-61A is specced at a sensitivity of -35+/-4dB. This translates into 17.7828 mV/Pa for -35db. -39db translates to only 11.2202 mV/Pa and -31db to 28.1838 mv/Pa. Great. As you can see, this will not be accurate enough. 8db max deviation.
Go on innerfidelity and see how much voltage is required for your headphone to reach 90db SPL at 500Hz. This is where things start to get even more inaccurate, as the different measurement systems will also see different volume levels due to different coupling. My IEM measurements show +5db higher level at 500Hz (or below) for the HD800 and HD600, a frequency that isn't affected by the shape of the ears. The voltage should be in the same ballpark, not exactly the same.
Other methods (probably better):
If you bought an expensive mic, you will very likely have a sheet with the measured sensitivity. If you do, then simply using that figure will probably give the most accurate result.
This is probably the easiest: Buy a cheap SPL meter and hope that it is accurate. Its inaccuracies will also depend on the frequency and don't forget that these are often A weighted. 500Hz should probably be fine. Place the SPL meter relatively close to a speaker driver and play a sine tone. Then hold your microphone to exactly the same position and calibrate accordingly.
This might be accurate, if done right: Do essentially the same thing that I described under (2), but with a speaker or speaker driver: Look up the published and 3rd party measured sensitivity of a driver (or speaker) that you just bought and measure it outside at 1m distance. Of course you absolutely need the same voltage on your amp. If you're not sure, use 2Vrms and use math to convert the volume level to 2.83Vrms, or similar. This should probably be done outside, as a 1m distance to your speaker inside will change the measurements a lot.
Maybe try the same thing as (2), but with an IEM. Should help keep coupler differences to a minimum.
Find someone with a calibrated measurement system. Calibrate your mic sensitivity and mic FR against that.
In the end, measurement system calibration is vital. Especially mic sensitivity and (to a lesser extent) mic frequency response, but proper calibration is probably not so easy if you don't have access to expensive tools.
I'm curious to see how the others calibrated their volume levels. Maybe I'm very stupid and overlooked the obvious. Let me know.
(Copied post here )
Oh, I completely forgot about that. I thought that those had to be much more expensive. 110$ is still not cheap.
The problem with this one (for me) is that it's made for 1/2" and 1" capsules. The WM-61A is 1/4". I believe the ones most people use are 1/4" mics in a 1/2" body or so, so that should work (the Behringer, UMIK, ...). This is probably accurate.
To be honest I doubt that the UMIK calibration is that inaccurate but I can't really say. I might buy one myself and calibrate against it. I'd also be very interested to see how distortion compares against my setup.
UMIK calibration is probably accurate. the way my OS ( Linux ) and JAVA + REW manage the hardware is more my usual suspect hence the fact I need to check.
How is the Dayton UMM-6 for this type of thing?
Wouldn't a USB mic be preferable to the XLR version?
The Dayton UMM-6 is great for this type of thing.
An XLR mic may need to be coupled to a mic preamp with a phantom power of about 48V. The mic preamps supplied by a POS PRO audio interface may be superior in paper to what is available in a USB mic (some Focusrite acolytes swear by Focusrite mic pre-amps) . A USB mic may be sufficient and much more convenient.
Furthermore, with a POS PRO audio interface one may be supplied with dual ADCs (or more) which may allow for simultaneous left and right measurements (or more). Such simultaneous measurements may be a bit harder to accomplish with a single channel USB mic (or a bunch of them).
BTW, this is an awesome interface (for your Dayton UMM-6 or similar clone) which I would buy in a heartbeat if I did not already had a Focusrite 2i2:
Guys, I need a little advice: I'm sketching out a 3D printed design for a measurement rig.
Picture two circular couplers, totally flat, that have the microphone sticking out of the middle. The couplers would be 100% solid, as thick as they needed to be to prevent internal resonance/etc. It would be built wide enough that the ear pad is pressed reasonably hard against the rig for a good seal.
Does this sound like it would work? Any anticipated problems? Does it need to have soft material at the point of the interface, or can it just be hard plastic with a microphone sticking out?
I think a solid coupler wouldn't be a good idea. Think CD as a coupler (I tried that).
Instead you should buy all sorts of foam materials and measuring different headphones and trying to find out which measures closest to what you hear. You want a worse seal than a solid material and less reflections, too. Try the creatology foam but don't let that be the only thing you try. In my opinion there are much better materials than it, but it all depends on what you are looking for.
You could 3D print something that holds the material, but keep in mind that the material behind the foam (even if it is 1" thick or thicker) will influence the measurements a lot.
What if the coupler was not much thicker than a CD, and I added foam on the headphone-facing side?
That's essentially what Marvey did. Pics are here (I don't think they are on SBAF):
Try different thicknesses/variants of foam and maybe even experiment with felt on top of the foam.
Hey guys, what you think about the Styrofoam surface on this dummy head as a coupling material for headphones.
It is a full size head so headphones can fit properly, and it has a stand.
I was thinking to insert mics in the ears. .
Honestly I wouldn't use something with an ear shape but you could try. Also I'm not sure if the styrofoam is a good idea. The measurements I get with the materials I have on hand don't look very promising, but you could try.
One thing that I've been doing for the past few days to find out how to improve my coupler design is to measure with a microphone on my skin (on my head, but not on my ears) to see essentially what the skin would behave as a material for a flat plate coupler. I get no ringing with the skin with my HD800. The only other material that I get no ringing with is the one I've been using. It's also extremely similar in terms of bass extension with the HD800. On the other hand the coupling is also more lossy, as my measurements are nearly 5db lower in general with my material.
I don't think i can rely on the level difference being the same for every headphone, which in turn means that I don't know how to calibrate the level for my coupler. This way I can't be sure that distortion measurements for different headphones are comparable. At the same time no other material gave me CSDs similar to my skin, other than that.
The head also had by far the smoothest response below 1kHz with no resonances at all. With my coupler I always get weird tiny bumps that really shouldn't be there. This is also something I want to eliminate. Part of it may be a headband resonance thing (the <100Hz ones), like Dan mentioned with the Ether Flow on Jude's rig, which means that it could be eliminated on a more final coupler design with a headband rest.
Overall I still think that I will stick to this material for now. In terms of resonance characteristics it seems to be the closest to the skin and the "most anechoic" in the treble by far of the ones that I've tried, while also giving me realistic bass measurements. I may want to try Pork skin like @Marvey has, but that's obviously not a permanent solution.
Yep, thanks I actually started to read this thread from beginning and just realized the need for uniform results among setups.
This is really excellent thread good info.
Separate names with a comma.