• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Royal EQ Rumble! Objective measurements of Dirac vs Audyssey vs REW+EAPO vs Nothing

johnp98

Active Member
Joined
Jan 29, 2021
Messages
106
Likes
161
I have always been interested in comparing the results of equalizing my speakers using REW, Audyssey, and Dirac and taking objective measurements for the comparison. I am not someone who is a validated trained listener, but I will also give a brief subjective opinion.

First off, my setup is as follows:

L&R speakers = HTM-12 v2 (Erins Audio Corner should have kippel ‘better than anechoic’ measurements up in the next while of the HTM-12 version 2)

Subs = Two 18” Mach5 IXL-18 subs in VBSS enclosures tuned to ~20hz I believe. Running off a Behringer NX6000D which I have already optimized the time delays and phases and EQed the subs already. Yes I realize I run my subs a little high.

Receiver = Marantz SR7002 which has Audyssey Multi EQ

Room = ~32ft x 12ft. Speakers at appropriate equilateral triangle. Listening distance ~10ft from each speaker.

Measurements = Using the Umic-1 and REW and the moving mic method for the overall frequency response. I put 1/12th smoothing on the measurements as I wanted to keep a lot of the resolution for the comparison, but likely for real life one could use 1/6th or psychoacoustic smoothing. I also took measurements at the main listening position to compare the impulse responses etc.

Anyways, after a long evening of measuring, here are my results!

No EQ:
Baseline / raw measurements with no EQ:
1622154790454-png.3138798


REW + Equalizer APO:
Using the above measurements I then used REWs EQ function and based on variable smoothing I produced L and R filters which I loaded into Equalizer APO on the computer, remeasured and it gave these results post EQ
1622154798893-png.3138799


Listening test = Can tell it is more neutral sounding. Slightly better soundstage across all frequencies.

Audyssey:
I realize that my Marantz SR7002 does not the most recent version of Audyssey
1622154806022-png.3138800

yet I wanted to know how well it worked, and also if it got the same timing and delays and sub integration that I achieved by using REWs time alignment feature and confirming with numerous measurements.

I took the 6 measurements in the exact same position where I would eventually do my Dirac measurements so they theoretically would be making an EQ based on the same data.

Here are the Audyssey results immediately post EQ:
1622154813838-png.3138801


Wow, so it clearly made things worse, so changing back to a crossover of 100hz gave me this:

1622154821233-png.3138802


Ok, clearly the time delays where also off, (it added 5.5ft to the distance of the subs), so I took it back to the delays that I manually calculated and it gave me this (so just keeping the EQ portion, and manually doing the crossover and the delays):
1622154828727-png.3138803

(so it is just keeping the EQ which I could look at and see was: L = 50hz +1, 2k -1, 16k +4 and R = 50hz +1, 4k +1, 16k +4)

Playing around with the AVR EQ settings a bit more, I found that you can manually change the 9 preset filters of 63hz, 125hz, 250hz, 500hz, 1k hz, 2k hz, 4k hz, 8k hz, 16k hz yourself. I am not sure what the Q value on these filters are, but it does seem quite wide and based on moving mic measurements an EQ of seemed to be the best. After playing around using these filters (125hz -1(L) -4 (R), 250hz +2, 500hz +1 , 4k +1 8k = -2) I got:
1622154837288-png.3138804


Hopefully this is just because I have an older version, but it was scary how bad vanilla Audyssey made it, and if I did not have a measurement mic then I would be in the dark with how to improve on what Audyssey gave me… wrong crossovers, wrong delays, and wrong EQ. Scary stuff.

Do new version let you manually see the EQ it is doing or allow you to put in the EQs that you want?

Listening test = I can tell the difference from no EQ, but hard if I was blinded to say which one was actually better.

Dirac:
Ok, now I went back to baseline and then used Dirac. Measured in the same locations that I used for Audyssey (plus a few more as Dirac allows more than 6 measurements).

Here are the result from Dirac.
1622154846608-png.3138805


Now I also know that Dirac uses a combination of FIR and IIR filters but this is where my knowledge becomes a little thin. I don’t know the best way to compare the measurements from the REW + EAPO group vs Dirac. I know that Dirac should help with the phase and the impulse response, but I don’t know what is the standard way to look at these results and publish them. But anyways here is a comparison of the impulse response pre and post:

Pre (No EQ)
1622154855636-png.3138806

1622237155237.png


Post (with Dirac)
1622154863462-png.3138807

1622237061040.png


What else should I be comparing pre and post Dirac?

I activated a 14 day dirac trial, so if people have other suggestions for measurements that I should obtain with Dirac active that would be much appreciated as I can get them asap before my trial runs out.

Listening test vs REW EQ = Has subtle improvements in imaging, overall tone and ‘neutrality’ of speaker not improved beyond what REW+EAPO did.

Final side by side comparisons:

Finally I measured the L+R speakers at the same time with the moving mic method and here is a comparison of all the methods:
1622154872557-png.3138808

Green = No EQ
Red = REW EQ
Purple = Audyssey (after changing back the crossover and delays)
Blue = Manual EQ via Audyssey
Orange = Dirac

Ok, and for people who just get way too anxious about seeing all the small imperfections. Here is the same graph but with psychoacoustic smoothing so you realize that some of the changes are not earth shattering:
1622154899263-png.3138809

Green = No EQ
Red = REW EQ
Purple = Audyssey (after changing back the crossover and delays)
Blue = Manual EQ via Audyssey
Orange = Dirac

So it seem like the ranking is:
Dirac > REW + EAPO > Manual Audyssey EQ = No EQ > Audyssey EQ >>> Audyssey sets EQ and crossover and delay


Take home points:
Don’t trust old versions of Audyssey! Like really don’t trust it! Do we trust the new versions?

You need a measurement mic and REW to even get the initial setup done properly.

Using REW for EQs (instead of Dirac) gets you 80-90% of the way there.

Dirac works very well, as many people have previously mentioned, but measurements and listening tests do back that up.

(Also, what else should I measure while I have access to Dirac?)
 
Last edited:

abdo123

Major Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
3,676
Likes
3,197
Location
Brussels, Belgium
You should really wait for the anechoic measurements so Dirac would have an excellent speaker to start with and not have to make any assumptions.

Nevertheless, Dirac seems the obvious winner to me.

try to use Var instead of psychoacoustic smoothing for EQ purposes.
 

Χ Ξ Σ

Senior Member
Joined
Jul 16, 2020
Messages
305
Likes
1,156
Location
UTC-8
Thanks for the comparison! Your room and speakers must be really nice as the No EQ measurements are already really good. If you want to do more measurements while you have access to Dirac, maybe try measuring multiple seats?
 
OP
J

johnp98

Active Member
Joined
Jan 29, 2021
Messages
106
Likes
161
You should really wait for the anechoic measurements so Dirac would have an excellent speaker to start with and not have to make any assumptions.

Nevertheless, Dirac seems the obvious winner to me.

try to use Var instead of psychoacoustic smoothing for EQ purposes.
Yeah I am looking forward to getting the anechoic measurements of the HTM-12 v2 as that will help immensely!
I did use Var smoothing for the EQ (hence why the high end is not as smooth as Dirac I assume). Just showed the results in 1/12th and then psychoacoustic at the end.
 

abdo123

Major Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
3,676
Likes
3,197
Location
Brussels, Belgium
Thanks for the comparison! Your room and speakers must be really nice as the No EQ measurements are already really good. If you want to do more measurements while you have access to Dirac, maybe try measuring multiple seats?

I wouldn’t call a 20 dB error rate above 100 Hz ‘really good’ but the subbass is really really impressive. However, he said he already EQed the subwoofers so no surprise there.
 

abdo123

Major Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
3,676
Likes
3,197
Location
Brussels, Belgium
Can you measure how much boost Dirac is adding for your particular EQ?

I always found it fascinating how much boost Dirac applies (and gets away with it) since most Dirac enabled devices are used for volume control as well.

I was using an analog volume control downstream my Dirac enabled device and i was heavily clipping in the digital domain from the boost.

and when i added the necessary headroom (which was 10 dB) I ran out of amplifier power and nothing sounded dynamic at all.

after some adjustments to the target curve now it needs 6 dB headroom and it’s manageable.
 
OP
J

johnp98

Active Member
Joined
Jan 29, 2021
Messages
106
Likes
161
I wouldn’t call a 20 dB error rate above 100 Hz ‘really good’ but the subbass is really really impressive. However, he said he already EQed the subwoofers so no surprise there.
Yeah I know it's not perfect my any stretch, but it's certainly not 20dB error. Probably +/-5. The 200hz dip has been giving me grief when I look at measurements, and I wonder if it is from speaker boundary interference with the side wall (as it is ~3.5ft from the front wall, but only 2ft from the side wall as the room is quite narrow).
 
OP
J

johnp98

Active Member
Joined
Jan 29, 2021
Messages
106
Likes
161
Can you measure how much boost Dirac is adding for your particular EQ?

I always found it fascinating how much boost Dirac applies (and gets away with it) since most Dirac enabled devices are used for volume control as well.

I was using an analog volume control downstream my Dirac enabled device and i was heavily clipping in the digital domain from the boost.
Sorry I am not sure how I would go about measuring that.
You mean just the overall dB level? As that would be easy.
But individual filter eq boosting I have no idea. But it certainly seems to boost any frequencys (see the 2.5k dip being boosted)
 

abdo123

Major Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
3,676
Likes
3,197
Location
Brussels, Belgium
It’s easy with my MiniDSP since i have a USB digital output (post digital processing).

But it would be difficult via the virtual interface Dirac creates on windows.

You know how when you use PEQ filters with positive gain you need to add some headroom in the digital domain so no clipping would occur?

Well Dirac doesn’t really add headroom, you have to add it yourself, be careful.
 
OP
J

johnp98

Active Member
Joined
Jan 29, 2021
Messages
106
Likes
161
It’s easy with my MiniDSP since i have a USB digital output (post digital processing).
But it would be difficult via the virtual interface Dirac creates on windows.
You know how when you use PEQ filters with positive gain you need to add some headroom in the digital domain so no clipping would occur?
Well Dirac doesn’t really add headroom, you have to add it yourself, be careful.
Yeah that totally makes sense, and will do!
 

ppataki

Active Member
Joined
Aug 7, 2019
Messages
197
Likes
168
Location
Budapest
Thank you @johnp98, this is really useful!
Can you please also post the Step Response graphs?
Based on my experience you should see a huge difference there for Dirac vs any other room EQ tool
(Dirac will time align the drivers in a multiway speaker using frequency dependent delay and that will be clearly visible in the Step Respons curve)
 

Haint

Active Member
Joined
Jan 26, 2020
Messages
246
Likes
280
Unfortunately the base level Audyssey MultiEQ is well known to be completely worthless. I'm surprised Sound United hasn't dumped it and just rolled out their own 10 or 20 band auto-PEQ system, seems like it'd be trivial to do with their resources and modern processing power. Would save a shit ton on licensing fees. XT32 on the other hand works very well and would produce results much closer to Dirac in raw FR (possibly even better if you're willing to play with the Audyssey app), but it too is terrible at aligning multi-subs and the crossover. You pretty much have to manually align multi-subs and Y-cable them to a single input before running Auyssey, then tune the crossover afterwards with subwoofer distance and volume.
 
Last edited:
OP
J

johnp98

Active Member
Joined
Jan 29, 2021
Messages
106
Likes
161
Thank you @johnp98, this is really useful!
Can you please also post the Step Response graphs?
Based on my experience you should see a huge difference there for Dirac vs any other room EQ tool
(Dirac will time align the drivers in a multiway speaker using frequency dependent delay and that will be clearly visible in the Step Respons curve)

Pre (No EQ):
1622222710420.png


Post (with Dirac):
1622222773810.png


Hummm... I also was expecting a larger response. I wonder if it is having a hard time with the subwoofer? Should I just measure the speakers with no sub? Or am I looking at the wrong thing? I feel quite proficient with measurements and the frequency response (and waterfalls etc), but the impulse and step responses I am not too sure how they are typically displayed and with what settings. Curious why there is not a more significant response. Thoughts?
 
OP
J

johnp98

Active Member
Joined
Jan 29, 2021
Messages
106
Likes
161
Unfortunately the base level Audyssey MultiEQ is well known to be completely worthless. I'm surprised Sound United hasn't dumped it and just rolled out their own 10 or 20 band auto-PEQ system, seems like it'd be trivial to do with their resources and modern processing power. Would save a shit ton on licensing fees. XT32 on the other hand works very well and would produce results much closer to Dirac in raw FR (possibly even better if you're willing to play with the Audyssey app), but it too is terrible at aligning multi-subs and the crossover. You pretty much have to manually align multi-subs and Y-cable them to a single input before running Auyssey, then tune the crossover afterwards with subwoofer distance and volume.

Yeah I assume XT32 is significantly better, or at least I hope it is! But yeah I don't know the licensing fees associated, but just having a PEQ system that one could load their own EQ would be ideal. But receivers don't often allow that, hence why minidsp has found a market I assume.
 

ppataki

Active Member
Joined
Aug 7, 2019
Messages
197
Likes
168
Location
Budapest
Zoom in horizontally to see let's say from -2ms to 2ms and then from -10ms to 10ms
That will show you what is happening with driver alignment, normally you shall see without Dirac spike of the tweeter followed by a huge hump that represents the woofer and with Dirac those shall be aligned
You could compare that like: no EQ, with REW, with Audyssey and with Dirac
 

ppataki

Active Member
Joined
Aug 7, 2019
Messages
197
Likes
168
Location
Budapest
Here are some of my measurements where this is clearly seen:

Nubert nuVero 140:

1622224927157.png


Nubert nuBox 483:

1622224486063.png


and my current DIY FAST system (12" Audio Nirvana fullrange drivers + 12" Dayton RSS315HF4 dual subs)

1622224660582.png


The difference is mind-blowing, especially in case of the multiway systems - and of course you can also hear this difference
You cannot achieve this with simple EQ
 

Attachments

  • 1622224428302.png
    1622224428302.png
    379.1 KB · Views: 28
Last edited:

Χ Ξ Σ

Senior Member
Joined
Jul 16, 2020
Messages
305
Likes
1,156
Location
UTC-8
Pre (No EQ):
View attachment 132424

Post (with Dirac):
View attachment 132425

Hummm... I also was expecting a larger response. I wonder if it is having a hard time with the subwoofer? Should I just measure the speakers with no sub? Or am I looking at the wrong thing? I feel quite proficient with measurements and the frequency response (and waterfalls etc), but the impulse and step responses I am not too sure how they are typically displayed and with what settings. Curious why there is not a more significant response. Thoughts?
I believe you should zoom in on the horizontal scale, the first 7ms or so.
 
Top Bottom