Help with measurements

[thatoneguy]

Hello.
I'm trying to measure a DIY headphone driver in a pair of headphones I made based on planar but configured in an unusual way.
I'm a total noob at this measurement stuff so I'm looking for guidance and outside interpretation.

I'm using the MiniDSP UMIK-1 held up to one driver where I expect my ear to be.

Not sure if my settings are correct but here it is.






linear waterfall plot




For some reason the log waterfall plot doesn't display under 200hz display under 200hz unless I expanded the window to 60ms



Do my settings look alright?
Also what's the deal with all the valleys in the waterfall response? Is that really the driver decay? I'm in a 7 foot wide 20 foot long shed, perhaps room reflections are causing that mess?
And is there a method for smoothing out all those wrinkles in the FR?
Are those wrinkles audible? Is there something in this mess of data that I should focus on improving?

 

[maverickronin]

There's all kinds of things going on there.

First of all is your coupling. Are you just holding the mic to the driver in the open air?

 

[thatoneguy]

Yes, is that bad? I don't know how else I'm supposed to do it.

 

[maverickronin]

Assuming you aren't going for a K1000/MySphere sort of style that hold the drivers in front of the ears in free space you need to put the driver into a baffle and couple it to something holding the mic to approximate the volume/acoustic impedance/etc the driver will see in actual use in order to get representative measurements.

If you already have a prototype pair of headphones the MiniDSP EARS is a good place to start. If you just have the bare driver and/or don't want to spend more money on testing hardware you can make a flat plate coupler.

Basically just take some 3/4" MDF or a thinner steel or aluminum plate, drill a hole for the microphone, add some supports for it, and seal the end with either blutak or neutral cure silicone depending on how permanent you want it to be. Then cover the top face with some closed cell craft foam to slightly damp reflections and get a surface more approximating human flesh than bare wood or metal.

When I was really into modding my headphones I made a DIY plate coupler like that with Dayton EMM6 and it was great for my purposes of adjusting damping on planar magnetics and 'stats.

Your raw FR looks excellent though. It seems like a very solid base to start tweaking from.

I would guess the the very long decay in the waterfall is wither a gating or noise issue, especially in the mids and treble. Accurate waterfall plots in the bass are hard to get because of background noise. The wavelengths are longer so you need to extend the gating and therefore let in more noise. I was never able to get that to work myself.

 

[thatoneguy]

Assuming you aren't going for a K1000/MySphere sort of style

Actually, that is exactly what it is. Although there is nothing guarding the front or back right now, the driver is completely exposed.
I'm measuring the FR from the outside side of the driver because the mic doesn't fit in between the two drivers and allow me to hold it at a straight angle.

couple it to something holding the mic to approximate the volume/acoustic impedance/etc the driver will see in actual use in order to get representative measurements.

Like something that simulates your head?

If you already have a prototype pair of headphones the MiniDSP EARS

I just bought the UMIK-1, can't afford the EARS.

Your raw FR looks excellent though. It seems like a very solid base to start tweaking from.

You wouldn't believe me if I told you how good it sounds. On well recorded tracks I literally can't tell the difference between reproduced sound and reality.
The sub bass is audible down to 1hz although only starts getting relatively loud at 5hz. After 20hz it reaches normal loudness levels.
Tracks with sub bass make the air feel solid, like nothing I've ever experienced.. The sound stage makes the HD800s sound tiny and the sound quality makes the HD800 sound like a phone speaker.
I've spent time with the majority of other top end headphones and nothing compares.

I would guess the the very long decay in the waterfall is wither a gating or noise issue

I don't understand, what do you mean by gating?

 

[maverickronin]

Actually, that is exactly what it is.

OK. That is quite different. In that case just holding up the mic to it will at lead give you something in the ballpark

Like something that simulates your head?

Not the whole head, but at least the interaction of the driver with the "enclosure" formed by the baffle, pads, and the side of your head.

Since you're making a different kind of headphone you'll need a different setup. A standard flat plate coupler won't do and neither will the EARS. I've never DIYed anything like this before, but off the top of my head, a 6" diameter round flat plate coupler with your UMIK-1 mounted in the middle. Make a stand to mount it vertically, and then another to hold the driver assembly in place relative to the mic plate. Preferably this should all be bolted down with makings so you can get consistent placement across different measurements.

I don't understand, what do you mean by gating?

Gating is basically just just ignoring everything the mic pics up after a brief time window in order to attempt to capture the direct impulse response and not consider reflections. It looks like you're using REW? I haven't used that before, but here's a page from MiniDSP explaining it.

 

[thatoneguy]

What about sticking the mic into a roll of paper towel? Similar size and texture to human head. I know someone that did this once.

 

[thatoneguy]

Gating is basically just just ignoring everything the mic pics up after a brief time window in order to attempt to capture the direct impulse response and not consider reflections

But the step response of my measurement is unusually long and doesn't settle until almost 500ms. My noob eyes don't spot any reflections in it. Isn't the step response supposed to only be a couple ms?

 

[maverickronin]

You mean like unrolling some paper towels and wrapping them back up flush with the end of the mic? That will be an improvement.

 

[maverickronin]

But the step response of my measurement is unusually long and doesn't settle until almost 500ms. My noob eyes don't spot any reflections in it. Isn't the step response supposed to only be a couple ms?

It's hard to say. Gating isn't a perfect solution since it can cut out actual data depending on what and how you're measuring.

Now that I take a second look at your impulse response it does look like it's taking a while to settle down. Without know anything else about your driver my first guess would be that it needs some more rear damping so it will come to a stop sooner.

 

[thatoneguy]

Well right now there is zero rear damping. It's just the raw driver in free air on both sides of the driver.
The headphones have an ESL like quickness despite the step response which doesn't make any sense to me.
I don't know much about mechanical damping or it's implementations in driver development, I just know that when I put my hands or some cloth over the back it only made it sound more veiled.
Perhaps I'm using the wrong material.

 

[RayDunzl]

And is there a method for smoothing out all those wrinkles in the FR?

 

[thatoneguy]

Yeah but that's just cheating the visuals. It doesn't change the reality of the measurement.

 

[RayDunzl]

You asked.

 

[thatoneguy]

I meant to ask if there was a way to smooth out the ripples in real life. I was also referring to the FR ripples, not the waterfall plot reflections or resonances or decay patterns or whatever they are. (if anyone has a clue let me know).
The FR has lots of ripples in the higher frequencies.

 

[maverickronin]

Most other measurements are smoothed a lot more that 1/48 anyway.

 

[thatoneguy]

Is there a standard?

 

[maverickronin]

None at all, especially with headphones.

Many 'stats and extra thin and light planar magnetics get a little ragged in that range on the micro scale, but still sound excellent. You can check the measurement archives of Stax and HiFiMans on InnerFidelity for examples.

https://www.innerfidelity.com/headphone-measurements

IIRC, Tyll used a stepped frequency sweep which would have less resolution than the MLS (I'm assuming anyway...) used by REW but you can still see the general trend when compared to most dynamics.

 

[solderdude]

Most Hifimans have this jagged response:

The HE6 (4 screw) doesn't though:

Audeze's don't (as much)

Ether does this a little:

Fostex T50RP-mk3 as well:

 

[thatoneguy]

Well, I suppose compared to those mine looks pretty good then, especially considering they aren't optimized.

I just realized that most waterfall plots only show roughly 30db of dynamic range, which explains why mine has so many valleys.
Reducing the dynamic range reveals something a bit more normal

Although for some reason the waterfall plot cuts off at 200hz and won't show below.
Also I just realized that impulse response and step response are two different things, although I still can't figure out how they are different.

In any case this is the impulse response

The response seems to be significantly faster than the other high end headphones I've looked at so far.
Although the ripples hang around for a while relative to the initial response, I wonder if those are reflections?