How to equalize your headphones: advanced tutorial (in progress) - Page 3

I'm trying to use it for games (I don't actually use VSThost to equalize but to do some post-processing, more specifically Isone & Isone Surround which both simulate speakers amazingly well), but even for video playback it's not too good yet (when watching a blu-ray I still get pops). I've tried messing around with priorities but maybe not enough (yet). I don't actually need it for video/audio playback as I use JRiver MC17, but I'd love to get it working properly with games as it does make some of them sound absolutely jaw-dropping compared to raw stereo or Dolby Headphones. 

FYI, there's a new standalone VST host about, called LiveProfessor. It has some issues with UI of Electri-Q, however it's better at handling ASIO it seems and has far fewer crashers than VSTHost.

However, it will not be free once out of beta - but neither is VAC.

Another recommendation would be to play sine waves at -6 dB or so to have some digital control headroom. Not recommended with 16-bit or otherwise noisy sources.

If you already have it, Foobar2000 has a good tone generator which can be used instead of SineGen.

I use the old and tested "many sines" technique without equal loudness curve and dual eq stuff.

Just tuning with many sines (or a sweep) until the spectrum sounds equally loud at all frequencies.

I do not remember all of it, but the tone generator had many tunables. Basic usage:

File->Add location...

Type in: tone://frequency for infinite length

tone://frequency,length for finite length

Enjoy your playlist of tones.

Applying inverse equal loudness curve just makes most music sound horrible and doesn't match any speaker or live setup I've heard other than cheap computer speakers, headphones and unsealed IEMs. Clearly it's wrong then - and this is what you'll be doing...

No way I'm playing them that loud - the result is normal "radio" talk loudness, which would place it around 30-40 phons.[

(Funny thing: some radio stations seem to use cardioid microphones at too close distances, causing a minor bass boost on voice.)

Perhaps I'm perceiving equal "loudness" differently? Doesn't look like, have a link to the equalized IEMs (SE-5) near-ATH:

I do perceive bass stronger than usual ATH by some 10 dB if the tympanometry (with excess precision: 50 Hz, 100, 500, 1000, 5k and 10 kHz) is to be trusted. I bet it is, since the results agreed with audiometry and couldn't be choked down to any ailment - the responses were all normal, just sensitivities differed. Otoacoustic emission tests were also all fine.

(This suite of tests would've costed a pretty penny if I weren't friends with the audiologist.)

Also, I've had to slightly fix the equalization since, but less than 3 dB max.

Just for pun: I'm what you could actually call... a basshead!

(Or rather, someone who can appreciate distortion-free bass some more. Audiometry was normal at 40 phon, within error bars, so there you go.)

It'd be fun to use constant stimulus method instead of the adaptive one I did using that website. Would be more precise.

Oh, a warning: if you're using BS2b (Bauer Stereophonic to Binaural) as a crossfeed mind that it attenuates high frequencies. (feed frequency is important here)

As I've ran that measurement with IEMs equalized for BS2b, this shouldn't be an issue. But if you want to chain BS2b, make sure you equalize with it on, otherwise you'll end up with dark sound. Other virtualizers may be different in this regard.

Bah, I can't get this bs2b correction straight.

By measurement, it goes 3 dB, BW=2.54, normal high shelf. That'd be 3 dB/oct Bessel. (Default electri-q shelf is 6 dB/oct Bessel.)

So feel free to put that as the compensation.

However, this doesn't match with normal music, as crossfeed is a highly nonlinear effect in low frequencies. Approximately gives a 3 dB low frequency cut. Seems it ends up with a comb-like effect in addition to that.

Final edit: confirmed, it adds a -3 dB comb that's in period of chosen knee frequency. Just reproducing that comb filter gives about 75% of bs2b soundstaging change, but not all of it.

In case anyone is curious, the almost exact fix for BS2b effect. Note that this likely will mess up pure mono signals.

Weirdness all around. I'd like some math master to explain this to me, but this is probably not the right forum for it. PM if you know the answer or want the eq file.

(For 1908 Hz knee.)

Guess I'll need a better crossfeed or will have to resurrect the analog one. (tricky!)

The answer: http://gilmore2.chem.northwestern.edu/projects/meier_prj.htm

Hello nasty math. So, technically it's possible to decomb the crossfeed...

Never mind, it seems that BS2b doesn't have any comb effect, unlike e.g. Head-Fit. However, it does either destroy bass reverb or doesn't crossfeed enough.

Head-Fit at least doesn't destroy bass and its mild comb effect can be equalized out. It's not aggressive enough in centering subbass though, giving slightly weird effects with very low frequencies.

Also, I've retooled my setup to be ~50% quieter (near 35 phon instead of ~45, you have been right about being set too loud), gives better soundstaging with Head-Fit for some reason. (despite linear response being similar and compensated)

I finally got to it and I'm still fine tuning as somehow I'm getting it a little wrong but my 840s are getting more and more well balanced, will post up pics when I'm done :D

Quote:

Originally Posted by Joe Bloggs 

Answer to the top question nobody asks: audio stutters when streamed through Virtual Audio Cable and VSTHost

 

There are a dozen settings in VAC and VSTHost that are claimed to be related to this: MS per int and Stream format limit in VAC, buffer size in VSTHost, priority levels for VSTHost... but none have had as great an effect on fixing things than this fix for me:

 

 

These two audio-related services in windows need to be manually raised to high priority for audio to play smoothly through VAC and VSTHost on any but the most bare-bones system with nothing running in the background.

 

Instructions for Windows Vista / 7

1. Open Task Manager (Ctrl-Alt-Del)

2. Go to the Processes tab

3. At the bottom, there's a button with a shield icon labelled "Show processes from all users".  Click this button and okay any security warning that pops up.  The svchost processes that host these two services will be hidden unless you take this step.

4. Go to the Services tab

5. Click on the Name column to sort services by name, this helps you to find the two audio services easily.

6. Right click on AudioEndpointBuilder, click "Go to Process"

7. You will be taken to the Processes tab where a process called svchost.exe will be highlighted.

8. You may see several identically named svchost.exe.  Right click the one that is currently highlighted, click "Set Priority", choose "High".  Confirm to change priority in the warning window that pops up.

9. Go back to the Services tab and repeat steps 6-8 for AudioSrv to raise its service host priority to high as well.

 

After I made these manual tweaks, VAC and VSTHost have been running as stutter-free as can be expected on my heavily loaded system.

Some observations about these services:

Both Svchost processes serve also other services. You can list the services with command

tasklist /SVC /FI "IMAGENAME eq svchost.exe"

My list is:

Image Name                     PID Services
========================= ======== ============================================
svchost.exe                   1204 DcomLaunch, PlugPlay, Power
svchost.exe                   1284 RpcEptMapper, RpcSs
svchost.exe                   1432 AudioSrv, Dhcp, eventlog,
                                   HomeGroupProvider, lmhosts, wscsvc
svchost.exe                   1476 AudioEndpointBuilder, CscService, hidserv,
                                   HomeGroupListener, Netman, PcaSvc, TrkWks,
                                   UmRdpService, UxSms, wudfsvc
svchost.exe                   1504 AeLookupSvc, Appinfo, BITS, Browser,
                                   CertPropSvc, gpsvc, IKEEXT, iphlpsvc,
                                   LanmanServer, MMCSS, ProfSvc, RasMan,
                                   RemoteAccess, Schedule, SENS, SessionEnv,
                                   SharedAccess, ShellHWDetection, Themes,
                                   Winmgmt, wuauserv
svchost.exe                   1720 EventSystem, fdPHost, netprofm, nsi,
                                   SstpSvc, WdiServiceHost, WebClient
svchost.exe                   1848 CryptSvc, Dnscache, LanmanWorkstation,
                                   NlaSvc, TapiSrv, TermService
svchost.exe                   2052 BFE, DPS, MpsSvc
svchost.exe                   2208 stisvc
svchost.exe                   4284 PolicyAgent
svchost.exe                   4444 FDResPub, FontCache, SSDPSRV, upnphost
svchost.exe                   6992 p2pimsvc, p2psvc, PNRPsvc
 

When you change the priority, you have to change it always after reboot. When you change the priority, you are changing it for all the services in the same process.

The services are configured in Registry key HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Svchost

If I read the registry right, it might be possible just to create new Scvhost processes just for these audio services.

Key:LocalServiceNetworkRestricted

Value:

DHCP
eventlog
AudioSrv
BthHFSrv
LmHosts
wscsvc
homegroupprovider
WPCSvc
 

Key:LocalSystemNetworkRestricted

UxSms
WdiSystemHost
Netman
trkwks
AudioEndpointBuilder
WUDFSvc
IPBusEnum
hidserv
dot3svc
irmon
sysmain
PcaSvc
homegrouplistener
WPDBusEnum
wlansvc
TabletInputService
CscService
UmRdpService
 

and there's sub keys for both of these with additional options.

It might be possible just to make duplicate copies of these and isolate audio services to those.

But I don't know what happens when I get updates for Microsoft if they assume the services are not modified. That might mess my system.

More info:

http://support.microsoft.com/kb/314056

It might be possible to try to automate this priority change:

- list tasks to spool file

- grep with the audio services, get pid

- change priority (somehow) with pid.

And the add this to boot.

Here's a new initial equalization result for Brainwavz B2 with Comply P tips. Very smooth, but bit bandlimited. Can't have everything just with TWFK.

Ok, Comply P are wrong tips for them, cause lowpass and bass distortion.

Comply T100 works far, far better with detail, but is less linear. Here's the graph:

Subbass still needs more boost, but it starts to distort. The sound is extremely detailed.

The IEM is not sibilant without the equalization - which proves my theory that it's resonance and not frequency response causing this effect.

Since you're going the DSP way, you should also try Dolby Headphones (dolbyhph.dll not included. You can find it on PowerDVD or any other software using Dolby Headphones. It's not included because it's proprietary).

It's very tricky to set-up in a way that doesn't sound like the singer is in a tin can (or too bassy), but when you do, it sounds great!

If I close my eyes, it really does sound like the music is in front of me. I've never used a crossfeed plug-in that managed to do that so well (but I'll give an honorable mention to xnor's crossfeed plug-in).

Here are my settings for the HD555 (Channel mixer is used to create the multi-channel signal for Dolby Headphones):

the other tabs are default

Different headphones will probably need different settings on the Upmix tab.

Nah, I've checked with everything I had. It's so generally bright that the sibilance doesn't stand out.

It's quite different from VSonic GR07 sibilance, sounds much more like a relatively high wide-band presence boost rather than some resonance...

It does accentuate sharp "c"/"tsk" phoneme. Not other plosives, "s" or "shh". This suggests resonance at much higher frequency.

Similar IEM - AKG K3003? http://en.goldenears.net/index.php?mid=GR_Earphones&document_srl=12318

Quote:

Originally Posted by Joe Bloggs 

Says the only guy in the EQ thread that EQs his subbass up 10+dB for all his phones even after being introduced to equal loudness contours

Fix your trolling gear. 10 dB is nothing next to the +40 to 50 dB of the equal loudness contour. See, it's pretty hard to tune a device to reproduce extreme subbass correctly.

Why do the resulting corrections look nothing like an equal loudness curve? Because they're not it at all.

Dynamic driver IEM would need some huge displacement to move enough air. Bass reflex (vent) helps, but is still not enough. You could try simulating it via an acoustic lowpass - tiny nozzle - in a multi-driver setup. Or you could attempt the use of a driver with large Qts either completely open back or with special sound absorbing material. (FutureSonic Ear Monitor?)

BA IEMs have it harder, With a right specialized balanced armature you could reproduce it perfectly - nobody I know of has done so yet it seems. (Even SE-5 extra bass armature starts to distort, but at least it can reach the right volume unlike most others where I'd rather not try since it could destroy the driver.)

As to the high rolloff/cutoff, well, you need something seriously precise and light to reproduce frequencies that high. A score of IEMs manages to do that already.

Hey, even most speakers roll off < 35 Hz.

Here's an example: http://www.linkwitzlab.com/woofer3.htm and yet another, of a specialized subwoofer: http://www.linkwitzlab.com/thor_splmax.htm

Guess why prof. Linkwitz "invented" the correction circuit - you can actually sometimes compensate for this behavior without introducing too much distortion. Obviously only if it's caused by mechanical dampening (in dynamic drivers) and not excursion limitations. (which is the case in BA)

http://www.linkwitzlab.com/Removed%20pages/x-filters.htm#9

Technically, the perfect driver would be a very light moving armature of large size or perhaps very light and stiff large orthodynamic.

(electrostatics also have major displacement limitations due to voltage required increasing with square root and there's only so much you can do against arcing)

Of course it won't fix decay issues. Many BA actually sound closer to sawtooth wave than sine when given one at low frequencies.

The red graph actually agrees perfectly. There's a - 6 dB at 20 Hz and should be +6 dB for standard subbass compensation.

(According to Golden Ears at least. I'm inclined to agree with them.)

Great thanks for the graph, where is it from?

Note that I'm not playing music at 94 dB SPL, but rather somewhere around 40 dB SPL.

Weird that they measure it with that much pressure, but I suppose their measurement system takes that into account.

Here's a new correction, with Comply P Slim (not P as before, I made a mistake then).

Found a pair, they work better than T100 - make for smoother equalization and bit better isolation, subbass as well as comfort.

I actually haven't even looked at those graphs when making it.

Let's paint the approximate shape of the correction on top of that graph. Please excuse the lousy painting skill.

This matches the Diffuse Field (DF) curve very well...

http://www.head-acoustics.de/downloads/eng/application_notes/Equalization_brochure.pdf

Quote:

Originally Posted by AstralStorm 

This matches the Diffuse Field (DF) curve very well...

http://www.head-acoustics.de/downloads/eng/application_notes/Equalization_brochure.pdf

The graph that was posted is already DF equalized. With the curve above you now have 2x DF equalization - the big dip should be flat. You should boost around 3 kHz.