|
|
Post by barrak on Feb 3, 2015 11:29:19 GMT -5
I ran Dirac for the first time last night to disappointing results. All of my speakers showed a 10dB drop from 7K to 15K. Using REW, I see no such issues. Plus, it added so much below 100Hz to my fronts that they suffered excessive movement. Hopefully they were not damaged. Most people say the bass seems weak but your reporting the opposite. Can you post some Dirac screen shots? And did you verify that AGC is turned off? It is probably an unrelated issue, as it could happen due to small LR run at full range or a measurement dip above xover with small LR woofers. This brings to mind another important bit that Tony raised once which could explain weak bass. Imagine a scenario where an underpowered woofer (for the room, in a sub or a full range) is measured. The default measurement signal level is 65dB. At that playback level it will do fine matching the rest of the setup. It is easily possible however that at higher playback volumes, the woofer's spl/volt ratio will gradually lag behind midrange/tweeter drivers. One way to test for this behavior is to observe bass integration post-Dirac at different volume levels. Such an issue would exist pre-Dirac, though it might be overwhelmed by uncorrected room modes. Because I have capable speakers all around, I set my measurement level at 75 dB. Small surrounds might not be able to play this level cleanly or safely as it might lead to distortion or driver clipping. |
|
|
|
Post by Priapulus on Feb 3, 2015 12:01:02 GMT -5
I ran Dirac for the first time last night to disappointing results. All of my speakers showed a 10dB drop from 7K to 15K. Using REW, I see no such issues. Plus, it added so much below 100Hz to my fronts that they suffered excessive movement. Hopefully they were not damaged. Sounds like a bad mike. /b
|
|
|
|
Post by ansat on Feb 3, 2015 20:28:38 GMT -5
I ran Dirac for the first time last night to disappointing results. All of my speakers showed a 10dB drop from 7K to 15K. Using REW, I see no such issues. Plus, it added so much below 100Hz to my fronts that they suffered excessive movement. Hopefully they were not damaged. Hey daleaherman, There are 2 different issues that I have seen in the past. The one that is most prevalent is a problem with the microphone calibration causing weak bass and overdriven highs. This would not be the case in your description. The other issue I see is what I call the "Big Hump". This is where the dirac graph is just a giant hump with really steep drop-offs well below the specifications of the speaker. In these cases, it appears that you have a bad microphone. You will want to call emotiva for a replacement. Tony |
|
|
|
Post by ansat on Feb 3, 2015 23:47:55 GMT -5
Hey all, Finished my analysis and thought that I would provide an update. I have been a little vague in my posting as it is a lot of emails and slow movement.
So what's new? Well I have been able to reproduce the results from the emotiva calibration file through various testing. I also was given access to a forum members calibration file from Cross-Spectrum. A HUGE THANK YOU to socketman as this calibration cost additional money out of his pocket. I would also like to thank all of those who provided their feed back to me privately as this was the driver to keep spending time researching this. But at this point, there is nothing more that I can do other then to wait and see what happens and providing assistance to those who agree with my analysis by providing instruction and the math calibration for their measurements.
Now to the updates. (some of this is old news, but there are bits of information that I left out previously)
It all started with this picture.
And my claim that the emotiva calibration is causing the problem. My argument remains the same as before, But now I am only seeing one trend and have pulled the ansatcal file and all that is available in my signature is the math calibration (read on for why).
Since my claim and the testing that I did was not enough to produce movement I continued my research and stopped posting what I was finding and just focusing on my communication with Dirac (which largely will need to remain private).
I started my research into the XTZ microphone calibration found in the emotiva folder and Dirac's requirements. The calibration file appears to be generated by Dirac by design. Below is taken directly from the Dirac website and I have made bold the parts that are critical to my argument
A little digging online, and I found that the calibration file included with Dirac is the same calibration file that XTZ gives by default with their microphone. I started digging for the XTZ manual which is found hereIn the manual I found the following statement. So to recap - Dirac wants a microphone that is either naturally -+ 1db or it wants a microphone + calibration that will result in a -+ 1db The XTZ microphone needs to be pointed towards the speaker to be accurate above 5khz. Given these two - Plus Emotivas recommendation to point the microphone to 90 degrees, I started to dig into the Dirac user manual as it appears there is some conflicting information. On page 16 of the Dirac manual it states that the microphone needs to be pointed at 90 degrees. So it seems that the use of the XTZ microphone with Dirac for a full range measurement could cause some issues in the high end reporting too low due to the usage of a on axis calibration file with a microphone pointed towards the ceiling. This lead me to wonder if the XTZ is the microphone being used as the reference microphone that the Emotiva microphone was calibrated off of as this could be the cause of what I am seeing. I have not been able to confirm what the reference microphone used was, so I continued my research. The next steps were to start taking my measurements again. This time using on axis calibration with on axis measurements 2 different cross spectrum microphones pointed forward towards the middle of the room (so not perfectly on axis which can account for the differences in the 10000hz+) compared to the emotiva microphone using emm-1 calibration file.  And wow. The problems in the high end started to disappear. This is a huge difference between the graph above this one and as I said earlier, if I would pointed the microphone directly at the speaker, I would expect a little bit different results due to the diaphragm size differences between the microphones and in all honesty, that last 2/3 of 1 octave is a small price compared to the 2.5 octaves that are now accurate. From this, I am convinced that the issue is with the calibration file used for the reference microphone that the EMM-1's were calibrated off of. And to be honest - I am thrilled that I can get the emotiva microphone to finally be in agreement with my cross spectrum microphones. (as you can also see the cross spectrum microphones very much agree with each other.) The low end is still a mystery though. But this information was enough to build the math cal. Now some more waiting. I provided my latest data and details to Dirac and the only thing left was the cross spectrum calibration results. (which was terrifying and exciting -- These results from cross spectrum would either provide some validation to all the time spent doing this research or completely invalidate my claim.) I received the data from socketman and immediately started working with the data. And what do you know. I have a bunch of squiggly lines!  Above you will see just a comparison of the calibration files. Red is the emm-1 calibration file generated by Dirac. The Blue line is the on axis calibration file from Cross Spectrum (pointed towards the speaker), the purple is the 90 cross spectrum calibration file (pointed towards the ceiling) and the yellow is the math calibration file that I have been giving out. If you look at the red line vs the blue line - It shows that the emm-1 calibration file is within ~1.5db from 100hz to about 17khz meaning that if we are to believe that the cross spectrum calibration is accurate, then the emm-1 calibration file is a really good 0 degree on axis calibration file. If you compare the red line with the purple line, you will see that from 100hz to 2khz, the emotiva microphone holds it own and starts to go down hill from there (losing its -+ 1.5db around 5khz). The math calibration file in yellow is -+1.5db across the whole scale, with a minor deviation at 38hz. (which is kind of funny as this was the part I corrected without the math...)Given how close the math cal is, I have removed the ansatcal as I no longer see any value in giving it out. So how does this translate in to measurements to Dirac.  Again - Assuming that the Cross- Spectrum calibration is accurate, The red line would show how much the EMM-1 calibration would change the result. And the green line is how much gain or loss that Dirac would add in error because of this. My analysis concludes that the file that Dirac LE is producing for the EMM-1 is a viable 0 degree on axis calibration file and is not suitable for off axis measurements and would need to be changed to provide Dirac LE with the information that it needs to correctly apply EQ (this has no effect on timing or the corrections to the impulse response). The issues below 100hz is still a mystery to me, but I believe it to be inaccurate. While opinions on how the calibration sounds can vary greatly, I firmly believe that the results are not as accurate as they could be. I would like to end my research & analysis with a few personal comments and a few requests from our community. Throughout this process I have seen a lot of data, taken a lot of measurements, had 100s of personal emails and over 500 PM's. I have tried very hard to keep the conversation peaceful. I am not trying to blame anyone, nor do I care to place any blame. In all honesty, I have nothing to gain by this being fixed as I have a work around that makes my EMM read exactly as I believe it should. But I really love emotiva products and am a proud owner of the XMC-1. I also have gained a new friend in Dirac and will be supporting their endeavors where ever it may lead them and would happily donate the time all over again if everyone with Dirac and an XMC-1 can get the correct results and see just how powerful the Dirac program can be (even with most of its customization taken away). I would like to ask that the members here who are in agreement with my results and those who are in disagreement to keep this conversations on target and not to fill this thread with arguments. Both Emotiva and Dirac are watching and they are both looking into this further. I have no timelines, nor any official statements from either company that I can share and will not be pursuing any further information from them about this. From here, I am leaving my math calibration in my signature as a workaround for anyone who wants to use it (It is really close to the results that Cross-Spectrum came up with). I hope that both Dirac and Emotiva can come to a conclusion with the assistance of this data as I feel that it can take bad experiences and turn them into good ones and take the good experiences and turn them into great ones. Thanks all, Tony |
|
klinemj
Emo VIPs 
Official Emofest Scribe
Posts: 15,092
|
Post by klinemj on Feb 4, 2015 5:42:23 GMT -5
Interesting data. Looks like a very good product can indeed be made better. Thanks for your efforts!
Mark
|
|
cawgijoe
Emo VIPs
"When you come to a fork in the road, take it." - Yogi Berra
Posts: 5,033
|
Post by cawgijoe on Feb 4, 2015 6:17:18 GMT -5
Tony, it's quite obvious that you really enjoy this stuff. Lots of hard work. I haven't paid much attention as Dirac as is sounds wonderful to me.
I do agree with Mark that if this results in improvement, I'm all for it. Thanks.
|
|
|
|
Post by LuisV on Feb 4, 2015 9:34:54 GMT -5
Thanks for all of your hard work Tony. Hopefully, both Emotiva and Dirac, are receptive to your analysis. Its great to see an enthusiast willing to go the extra mile to help others and provide their findings to the vendors involved for scrutiny and potential inclusion into their solutions. Kudos to you and to those that provided a helping hand...
|
|
|
|
Post by hesster on Feb 4, 2015 10:26:17 GMT -5
I certainly agree with these comments on the work done here, and Tony' personal time with my setup issues. I am also a very satisfied Emotiva product user, and absolutely love my XMC-1. Given this, much of this in depth stuff is way beyond many of us novices, but we are smart enough to explicitly follow forum guidelines and tips on Dirac and Mic setup procedure's, including alternative Mic Cals and gain settings to obtain the best available results. Still, we are always looking to any improvements which will simply translate into "do this" software, Cals, procedure's, or whatever revisions/updates may become available.
|
|
|
|
Post by Priapulus on Feb 4, 2015 10:28:08 GMT -5
The deep bass still seems weak Sincerely, /b
I've got my bass back! First time since I started fooling with Dirac.
I was fooling around with Dirac tonight. Then I quit, and reloaded my most recent version, and I've got my bass back. I didn't do any changes, I just took my saved results and made and downloaded to the XMC, new filters (see above, which I liked except for lack of bass).
I'm not aware of any changes that I might have made. Perhaps something previously went wrong with the filter construction or loading, or something wonky in the XMC. But all is well now.
So, if you're Dirac seems off, try building and loading the same filters again. A factory reset and reboot of the XMC might be a good idea too!
Sincerely /b
And the next day, when I booted the XMC-1 up, the bass was gone again!!!!
I think I've figured it out. It seems like if you change loudness levels in "Setup/Speakers/Dirac/Levels/subwoofer/" The changes don't take effect immediately. They take effect after restarting the XMC, after a Dirac run, etc. And least that is how it seems to me. If so, I presume its a bug.
Sincerely blair
Sorry if this is off topic; but it seemed relevant to a workaround to the bass concern
|
|
|
|
Post by socketman on Feb 4, 2015 10:32:44 GMT -5
I am sure there are plenty of people that may not take the time to fully read Tony's very informative post, so here is the take away in short form. I sent my mic to Cross Spectrum labs and had it calibrated, when I got it back I compared the results with Emotiva's provided calibration. Long story short, somehow we were provided with the Odeg correction file rather than for 90deg/straight up. This cause 2 things, first is exaggerated highs and second decreased bass.  You will see in this screen shot that I measured my left from speaker with the EMM without the cal file, another with the cal file from Emotiva and then with the file from Cross Spectrum. You will see that the EMM thinks the highs are not high enough and the lows around 50Hz are too high . Thanks to Tony for all his hard work and helping me along the way. Anyone who wants the Cross Spectrum file can PM me and I will share it since my main objective is the greater good. Richard |
|
|
|
Post by socketman on Feb 4, 2015 11:16:28 GMT -5
I will upload the file and share a link here soon much easier that way.
Richard
|
|
LCSeminole
Global Moderator 
Res firma mitescere nescit.
Posts: 20,858
|
Post by LCSeminole on Feb 4, 2015 11:16:52 GMT -5
My UMIK-1 came from Cross Spectrum with the 3 different calibration files so shouldn't mic position/orientation fix this situation as well? I know Flávio recommended for me to use the 90 degree file with my UMIK-1 and it produced slightly better results with REW, but Cross Spectrum would not have included the other degree files if they didn't work, correct?
|
|
|
|
Post by markc on Feb 4, 2015 11:40:13 GMT -5
Feb 4, 2015 16:16:52 GMT LCSeminole said:
My UMIK-1 came from Cross Spectrum with the 3 different calibration files so shouldn't mic position/orientation fix this situation as well? I know Flávio recommended for me to use the 90 degree file with my UMIK-1 and it produced slightly better results with REW, but Cross Spectrum would not have included the other degree files if they didn't work, correct?
The other calibration files are correct for that mic and in theory will work and it is possible to get correct readings with a horizontal calibrated mic and relevant calibration file pointing directly at a speaker.
But herein lies the problem:
Dirac sequentially checks each of the speakers one after the other and then repeats the first.
It is not possible to point the mic at each and all of the speakers at the same time nor move it from speaker to speaker during the assessment phase.
Hence, a vertically aligned, pointing upwards mic gives the best compromise as all speakers can be measured one after another with the same mic placement |
|
|
|
Post by markc on Feb 4, 2015 11:44:49 GMT -5
I've got my bass back! First time since I started fooling with Dirac.
I was fooling around with Dirac tonight. Then I quit, and reloaded my most recent version, and I've got my bass back. I didn't do any changes, I just took my saved results and made and downloaded to the XMC, new filters (see above, which I liked except for lack of bass).
I'm not aware of any changes that I might have made. Perhaps something previously went wrong with the filter construction or loading, or something wonky in the XMC. But all is well now.
So, if you're Dirac seems off, try building and loading the same filters again. A factory reset and reboot of the XMC might be a good idea too!
Sincerely /b
And the next day, when I booted the XMC-1 up, the bass was gone again!!!!
I think I've figured it out. It seems like if you change loudness levels in "Setup/Speakers/Dirac/Levels/subwoofer/" The changes don't take effect immediately. They take effect after restarting the XMC, after a Dirac run, etc. And least that is how it seems to me. If so, I presume its a bug.
Sincerely blair
Sorry if this is off topic; but it seemed relevant to a workaround to the bass concern
It seems a reboot can remove many setting related problems.
However, in relation to your post about turning "Loudness" on to compensate for Dirac winding down the bass (because it measures it at higher than it is in relation to the higher frequencies), Loudness is not an acceptable workaround because Dirac is not calibrating the bass properly for room response and then Loudness is a non compensatory approximation "fudge" based on assumed hearing thresholds to dial it back up again.
|
|
|
|
Post by geebo on Feb 4, 2015 11:55:33 GMT -5
And the next day, when I booted the XMC-1 up, the bass was gone again!!!!
I think I've figured it out. It seems like if you change loudness levels in "Setup/Speakers/Dirac/Levels/subwoofer/" The changes don't take effect immediately. They take effect after restarting the XMC, after a Dirac run, etc. And least that is how it seems to me. If so, I presume its a bug.
Sincerely blair
Sorry if this is off topic; but it seemed relevant to a workaround to the bass concern
It seems a reboot can remove many setting related problems.
However, in relation to your post about turning "Loudness" on to compensate for Dirac winding down the bass (because it measures it at higher than it is in relation to the higher frequencies), Loudness is not an acceptable workaround because Dirac is not calibrating the bass properly for room response and then Loudness is a non compensatory approximation "fudge" based on assumed hearing thresholds to dial it back up again.
I think he's increasing the subwoofer levels in the Dirac preset which is not the same as the loudness compansation. |
|
LCSeminole
Global Moderator
Res firma mitescere nescit.
Posts: 20,858
|
Post by LCSeminole on Feb 4, 2015 11:56:59 GMT -5
Good point markc, I was picturing myself running REW and not Dirac.
|
|
|
|
Post by socketman on Feb 4, 2015 11:58:25 GMT -5
Here ya go boys, Thanks Tony
narrow band onedrive.live.com/download.aspx?cid=A511D833FA1D5C22&resid=A511D833FA1D5C22%2116693&canary=gESq0Dx%2FeEmPdCOadWSFbPKxnSUBK4hS2AyOb%2BYle6k%3D3
1/3 octave onedrive.live.com/download.aspx?cid=A511D833FA1D5C22&resid=A511D833FA1D5C22%2116692&canary=gESq0Dx%2FeEmPdCOadWSFbPKxnSUBK4hS2AyOb%2BYle6k%3D3
|
|
dubs
Minor Hero
Posts: 69 
|
Post by dubs on Feb 4, 2015 12:17:56 GMT -5
Alright guys, what's a guy, who doesn't understand any of this s--t, supposed to do with his only Emotive mic to make Dirac work it's very best with the XMC-1? Download what? Point where? HELP TONY!!!!!!!!
|
|
dubs
Minor Hero
Posts: 69
|
Post by dubs on Feb 4, 2015 12:20:42 GMT -5
I forgot the how to do it part. Step by step would be very helpful. Thanks to all. Especially Tony for his hard work.
|
|
|
|
Post by ansat on Feb 4, 2015 12:43:27 GMT -5
I will try to get something up over lunch.
|
|
