Can someone explain this phase "glitch" in plain english to me?

[eobet]

I did some EQ on my headphones and at first, everything looked fine. Note how there's a slight drip at 8.1khz which I managed to EQ away:

But then, after a few more rounds, this happened. The dip at 8.1khz is now extreme and refuses to be EQd:

And that's pretty much stayed in place and won't go away (even with EQ turned off). Anyone know why this is?

I'm using a MiniDSP Ears, which I guess isn't the most reliable piece of gear?

 

[ddude003]

You don't say if your boosting your dips or cutting everything else... My guess is your trying to boost your way out of ~15db dip with +15b of boost... Not a good way to go... Also thinking you are using the RME Parametric EQ... If your source is a computer player you might try a convolution engine FIR to EQ or a plugin passive parallel EQ emulation... I am no headphone expert and I have seen some wild looking EQs for headphones... So, you might have a look at https://github.com/jaakkopasanen/AutoEq for some addition information...

 

[sam_adams]

And that's pretty much stayed in place and won't go away. Anyone know why this is?

There's a very good chance that is a resonance. You should generate a minimum phase response from the measurement and then look at the group delay and excess group delay. The one question is; why you are EQing the headphone to flat? You should look at the Harman curves from the link @ddude003 posted (direct link to .csv files).

 

[eobet]

My guess is your trying to boost your way out of ~15db dip with +15b of boost... Not a good way to go... Also thinking you are using the RME Parametric EQ...

I am using the ADI-2, but the most I'm doing is -7db at 4.4khz and at the problem area, 8.1khz, the EQ band is actually set to 0.

Also, I already tried the AutoEQ file for my headphones before I got the RME and it sounded like utter garbage unfortunately. I took a quick look at the Harman cuve, but just from looking at it I don't think I'll like it as there seems to be too much treble for me with Tinnitus. I don't listen to the flat curve anyway, I just like to have it as a reference. I have a more laid-back and fun curve that's my favourite...

EDIT: I guess I should show a graph without any EQ at all...

 

[eobet]

There's a very good chance that is a resonance. You should generate a minimum phase response from the measurement and then look at the group delay and excess group delay.

Thank you for this, I will read up about that, test it and see what I can find out.

Hmm... maybe I haven't found the correct information yet, but the pages I've read so far has told me to look out for sharp dips and peaks in the group delay graph and even though I've increased the zoom a lot, there's nothing coming close to the phase dip & spike (which is still there):



But I now see that that graph is actually called "excess phase", so now I must read about what that means too:



(The above is my preferred listening EQ tuning, btw... pretty far from that Harman curve, I guess.)

 

[sam_adams]

You should take a look at the distortion plot, also.

 

[ddude003]

It might be a good idea to post .mdat files...

 

[eobet]

Here is the mdat.

Now, MiniDSP support replied to me and claimed that this was due to their simulation of the ear canal, and supplied this link... but I'm doubtful because looking at the physical product, it's just a straight hole about 1cm deep with a mic in the bottom of it... it doesn't look very organic (unlike the ear which is quite well made)... but I don't fault them for that since this product cost a fraction of what professional equipment does.

 

[sam_adams]

There's a better than even chance that the response dip at ≈ 7.95 KHz is a quarter-wave cancellation caused by the ear canal of the fixture. Do you have another set of headphones you can measure?

 

[eobet]

I do, but they're closed back so apparently the MiniDSP Ears react quite differently to those and indeed that seems to be the case. So yeah, resonance then? Anything one can do about that in REW?

Also found this, while researching the subject: https://www.reddit.com/r/headphones/comments/8ohrlt/comment/e03pzt7/

 

[sam_adams]

Also found this, while researching the subject

The results of that brief analysis does not seem to bode well for the EARS device as a low-cost measuring device. Most unfortunate.

 

[ddude003]

So maybe its time to go old skool... Stop measuring and looking at the graphs... Play something familiar to you... Close you eyes and listen... Adjust your eq settings to your liking... Then run a sweep and see what that looks like... This is really about your kit + your ears + your brain... Your preferences are ok... If you don't like your on board EQ then try a plugin for your player like the Pulsar 8200 which is a software emulation of George Massenburg's legendary GML 8200... Sometimes the analogue workflow helps get you to audio nirvana...