• 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.

Dirac Live, FIR-filter - Yes, No?

DonH56

Major Contributor
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,719
Likes
2,851
Location
Monument, CO
#41
The variance position-to-position will depend heavily upon the room (and what is in it) and the speakers.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,773
Likes
4,843
#42
Okay, just for Ray here is a vector average. Never looked at it before either. Magenta is vector average and regular amplitude average is in yellow. The vector average takes into account phase. It doesn't look like a good thing to me. Some discussion of it here. Oh, and by not a good thing I meant for measuring response as you hear it. I could be wrong, and it could be useful for other functions. So not complaining it is there.

The discontinuity around 6oo hz would be about right for the cancellation due to changing position by about 1 foot. So if you were looking for some kind of phase change such an average involving only two points would be useful. Maybe great in fact for woofer fine tuning. BTW, you have to time align to get a vector average.

Some discussion of Vector Average here.
https://www.avnirvana.com/threads/spacial-average-measurements.1204/

Vector and regular averagea.png


Here is the front center and rear center measurments. Directly in line with the speaker, but one measure is 1 foot further away. You see both the amplitude average in blue and the vector average in green. Notice the deepest cancellation is about 588 hz which is what you'd expect for moving the microphone 12.5 inches at my temp and humidity. So I missed by a half inch. Sorry for the sloppy work.

I can see vector averaging being great for time aligning multiple subs by only using one measurement position and measuring each sub separately.

Vector vs two centers.png
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
18,970
Likes
19,176
Location
Seattle Area
#43
Wondering about the single point vs multi-point room correction measurement situation. Are you giving up a lot by measuring only at the listening position? As usual it depends.
As part of developing ARCOS room EQ, Harman (Dr. Olive) tested this difference. Here was the outcome:

1547355606445.png


RC1 is optimized for one location, RC2 is optimized for multiple. As you see, RC1 (optimized for one) did better. These still used multiple microphones but for RC1 they were more biased toward the listening position.

Here is the response of each room eq algorithm:



RC1 (in brown) has a much more ideal response than multi-point measurements for the rest.
 

jhaider

Active Member
Joined
Jun 5, 2016
Messages
256
Likes
140
#44
My REW measurement was from a single spot, not multiple (ala Dirac). As I explained I can't repeat a multi-point measurement scheme (without a fixture as Kal explained).
I don't think one should take confirmatory measurements for room correction systems at the same positions. I believe random sampling tests real-world efficacy and stability of a room correction better than repeating the microphone positions used for calibration. My procedure is to sample six points around the listening position. I think Dirac as implemented on the Bryston SP4 performed very well (measurement superimposed on Dirac graph, 1/12 oct. smoothing):

figure 23 – dirac live precision.png


More measurements here: https://hometheaterhifi.com/reviews...ton-sp4-16-channel-surround-processor-review/
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,773
Likes
4,843
#46
sorry if this has been discussed already but has anyone measured with DIRAC applied to see how closely the "after" measurement as claimed by DIRAC matches reality?
No, but that has been in my plans. A friend with a very good system uses Dirac, and I've wanted to measure the result with REW after correction vs with no correction. Obviously something you'd like to see. Disappointed someone has not done this yet.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
18,970
Likes
19,176
Location
Seattle Area
#47
sorry if this has been discussed already but has anyone measured with DIRAC applied to see how closely the "after" measurement as claimed by DIRAC matches reality?
I have and that is what we have been discussing. :)
 

Kal Rubinson

Addicted to Fun and Learning
Joined
Mar 23, 2016
Messages
866
Likes
645
Location
NYC/CT
#48
I don't think one should take confirmatory measurements for room correction systems at the same positions. I believe random sampling tests real-world efficacy and stability of a room correction better than repeating the microphone positions used for calibration.
I do agree with this. Taking measurements at the same positions only confirms that the EQ system has performed its specific function of correcting based on those positions. However, there is an implicit promise in EQ systems that they are sampling enough of the room to make a generally effective solution. It is the latter that we hope to attain.
 

DonH56

Major Contributor
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,719
Likes
2,851
Location
Monument, CO
#50
I discussed my results earlier as well, have not posted curves however (busy week/weekend). Reasonably well matched at the MLP.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,911
Likes
1,375
Location
Zg, Cro
#52
"So, controlling how well Dirac Live made corrections by repaeting the measurement at the same 9 points used initially by Dirac Live is misleading and/or out of context? Oh boy.. :facepalm::facepalm: "

Taking 9 measurements and applying a simple average may not yield the same result as taking 9 measurements and applying a proprietary algorithm to determine a corrective amplitude/phase filter.
That is correct. We certainly don't know for sure if Dirac software made a simple average of those 9 measurements, but the point of my post was to repeat the measruements in the same volume of space at which those 9 samples were taken in order to check how good job Dirac have done with the correction.

To do that you don't need to repeat the measurement at those 9 points nor you have to know how exactly Dirac processed those 9 measruements. You can, for example, us RTA moving microphone method and take 80+ samples in that same volume of listening space so you will certainly get a very good idea how Dirac's correction is performing.
 
Last edited:

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,911
Likes
1,375
Location
Zg, Cro
#53
Wondering about the single point vs multi-point room correction measurement situation. Are you giving up a lot by measuring only at the listening position? As usual it depends.

I used REW to measure in six positions. The center LP, one foot right of the LP, and one foot left of the LP. Then moved back one foot and repeated. So 6 measurements on a 3 x 2 foot grid. 12 feet from the speaker is the LP.

First all 6 measures with no smoothing. Which looks rather poor. Even for a given measurement.
View attachment 20253
But our ears don't really hear that. They have some smoothing capability. So I applied 1/6th octave smoothing. It looks better. This is about the kind of smoothing in the mids and above for our ears.
View attachment 20254

Then I used an ERB smoothing filter. It looks not terrible now.
View attachment 20255

That looks better still. Our ears will hear first arrival sound mostly plus do some smoothing themselves. So the last one may be close to our perception. Above 1 khz the ERB smoothed responses fit into a 2 db wide window. Below 500 hz the room modes will have an effect of course. And something of a transition in between.

I don't know what Dirac and other softwares do, but they probably do some sort of averaging or smoothing among other things. It strikes me that any of the ERB smoothed response measurements used to EQ the result isn't going to sound radically different (maybe barely noticeable) to any of the others or even an average of them all. So single point measurements probably work okay for this.
This looks quite good. :) Have your measurements been taken with both speakers playing?

From my experience it usually takes you to make average of sweeps taken at at least 9 points to have a decent match with RTA moving microphone method with 70-80 samples. I'm using 1/6 smoothing as a starting point when making manual corrections with rePhase and 1/12 smoothing for final touch.
 
Last edited:

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,911
Likes
1,375
Location
Zg, Cro
#54
Here are my 7 "across the top of the sofa" measures from two years ago, with "flat" DSP.

Traces 1 through 7 are left to right, with trace 4 (orange) the center.

View attachment 20263

Impulse Response, far left, as an example:

View attachment 20264

Center, Vector Average, and Simple Average: (never looked at vector average before)

View attachment 20266
Ray, what is "center average"? How do you get it and how does it differ from "simple" average? :)
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,911
Likes
1,375
Location
Zg, Cro
#55
I don't think one should take confirmatory measurements for room correction systems at the same positions.
This is contradictory to the Harman's conclusions @amirm posted above.

I optimised response to my LP but when checking if I did it right I was pleasantly suprised to discover that the position 1.5m behind my LP also benefited from the room EQ optimised for LP. The response there was both, more linear and smooth when measured but also much more pleasant to listen although my optimisation was done for LP.
 

RayDunzl

Major Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
6,820
Likes
2,927
Location
Riverview, FL
#57
Ray, what is "center average"? How do you get it and how does it differ from "simple" average?

There is no Center Average, just a measurement from the center (prime listening position)
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
18,970
Likes
19,176
Location
Seattle Area
#59
Sorry if this has been answered already but how does Dirac handle multiple subs?
As a single one. It can't EQ them independently.
 
Top Bottom